1 /*
2 * GENEVE: Generic Network Virtualization Encapsulation
3 *
4 * Copyright (c) 2015 Red Hat, 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/netdevice.h>
16 #include <linux/etherdevice.h>
17 #include <linux/hash.h>
18 #include <net/dst_metadata.h>
19 #include <net/gro_cells.h>
20 #include <net/rtnetlink.h>
21 #include <net/geneve.h>
22 #include <net/protocol.h>
23
24 #define GENEVE_NETDEV_VER "0.6"
25
26 #define GENEVE_UDP_PORT 6081
27
28 #define GENEVE_N_VID (1u << 24)
29 #define GENEVE_VID_MASK (GENEVE_N_VID - 1)
30
31 #define VNI_HASH_BITS 10
32 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
33
34 static bool log_ecn_error = true;
35 module_param(log_ecn_error, bool, 0644);
36 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
37
38 #define GENEVE_VER 0
39 #define GENEVE_BASE_HLEN (sizeof(struct udphdr) + sizeof(struct genevehdr))
40
41 /* per-network namespace private data for this module */
42 struct geneve_net {
43 struct list_head geneve_list;
44 struct list_head sock_list;
45 };
46
47 static int geneve_net_id;
48
49 union geneve_addr {
50 struct sockaddr_in sin;
51 struct sockaddr_in6 sin6;
52 struct sockaddr sa;
53 };
54
55 static union geneve_addr geneve_remote_unspec = { .sa.sa_family = AF_UNSPEC, };
56
57 /* Pseudo network device */
58 struct geneve_dev {
59 struct hlist_node hlist; /* vni hash table */
60 struct net *net; /* netns for packet i/o */
61 struct net_device *dev; /* netdev for geneve tunnel */
62 struct geneve_sock *sock4; /* IPv4 socket used for geneve tunnel */
63 #if IS_ENABLED(CONFIG_IPV6)
64 struct geneve_sock *sock6; /* IPv6 socket used for geneve tunnel */
65 #endif
66 u8 vni[3]; /* virtual network ID for tunnel */
67 u8 ttl; /* TTL override */
68 u8 tos; /* TOS override */
69 union geneve_addr remote; /* IP address for link partner */
70 struct list_head next; /* geneve's per namespace list */
71 __be16 dst_port;
72 bool collect_md;
73 struct gro_cells gro_cells;
74 };
75
76 struct geneve_sock {
77 bool collect_md;
78 struct list_head list;
79 struct socket *sock;
80 struct rcu_head rcu;
81 int refcnt;
82 struct udp_offload udp_offloads;
83 struct hlist_head vni_list[VNI_HASH_SIZE];
84 };
85
geneve_net_vni_hash(u8 vni[3])86 static inline __u32 geneve_net_vni_hash(u8 vni[3])
87 {
88 __u32 vnid;
89
90 vnid = (vni[0] << 16) | (vni[1] << 8) | vni[2];
91 return hash_32(vnid, VNI_HASH_BITS);
92 }
93
vni_to_tunnel_id(const __u8 * vni)94 static __be64 vni_to_tunnel_id(const __u8 *vni)
95 {
96 #ifdef __BIG_ENDIAN
97 return (vni[0] << 16) | (vni[1] << 8) | vni[2];
98 #else
99 return (__force __be64)(((__force u64)vni[0] << 40) |
100 ((__force u64)vni[1] << 48) |
101 ((__force u64)vni[2] << 56));
102 #endif
103 }
104
geneve_lookup(struct geneve_sock * gs,__be32 addr,u8 vni[])105 static struct geneve_dev *geneve_lookup(struct geneve_sock *gs,
106 __be32 addr, u8 vni[])
107 {
108 struct hlist_head *vni_list_head;
109 struct geneve_dev *geneve;
110 __u32 hash;
111
112 /* Find the device for this VNI */
113 hash = geneve_net_vni_hash(vni);
114 vni_list_head = &gs->vni_list[hash];
115 hlist_for_each_entry_rcu(geneve, vni_list_head, hlist) {
116 if (!memcmp(vni, geneve->vni, sizeof(geneve->vni)) &&
117 addr == geneve->remote.sin.sin_addr.s_addr)
118 return geneve;
119 }
120 return NULL;
121 }
122
123 #if IS_ENABLED(CONFIG_IPV6)
geneve6_lookup(struct geneve_sock * gs,struct in6_addr addr6,u8 vni[])124 static struct geneve_dev *geneve6_lookup(struct geneve_sock *gs,
125 struct in6_addr addr6, u8 vni[])
126 {
127 struct hlist_head *vni_list_head;
128 struct geneve_dev *geneve;
129 __u32 hash;
130
131 /* Find the device for this VNI */
132 hash = geneve_net_vni_hash(vni);
133 vni_list_head = &gs->vni_list[hash];
134 hlist_for_each_entry_rcu(geneve, vni_list_head, hlist) {
135 if (!memcmp(vni, geneve->vni, sizeof(geneve->vni)) &&
136 ipv6_addr_equal(&addr6, &geneve->remote.sin6.sin6_addr))
137 return geneve;
138 }
139 return NULL;
140 }
141 #endif
142
geneve_hdr(const struct sk_buff * skb)143 static inline struct genevehdr *geneve_hdr(const struct sk_buff *skb)
144 {
145 return (struct genevehdr *)(udp_hdr(skb) + 1);
146 }
147
148 /* geneve receive/decap routine */
geneve_rx(struct geneve_sock * gs,struct sk_buff * skb)149 static void geneve_rx(struct geneve_sock *gs, struct sk_buff *skb)
150 {
151 struct genevehdr *gnvh = geneve_hdr(skb);
152 struct metadata_dst *tun_dst = NULL;
153 struct geneve_dev *geneve = NULL;
154 struct pcpu_sw_netstats *stats;
155 struct iphdr *iph = NULL;
156 __be32 addr;
157 static u8 zero_vni[3];
158 u8 *vni;
159 int err = 0;
160 sa_family_t sa_family;
161 #if IS_ENABLED(CONFIG_IPV6)
162 struct ipv6hdr *ip6h = NULL;
163 struct in6_addr addr6;
164 static struct in6_addr zero_addr6;
165 #endif
166
167 sa_family = gs->sock->sk->sk_family;
168
169 if (sa_family == AF_INET) {
170 iph = ip_hdr(skb); /* outer IP header... */
171
172 if (gs->collect_md) {
173 vni = zero_vni;
174 addr = 0;
175 } else {
176 vni = gnvh->vni;
177
178 addr = iph->saddr;
179 }
180
181 geneve = geneve_lookup(gs, addr, vni);
182 #if IS_ENABLED(CONFIG_IPV6)
183 } else if (sa_family == AF_INET6) {
184 ip6h = ipv6_hdr(skb); /* outer IPv6 header... */
185
186 if (gs->collect_md) {
187 vni = zero_vni;
188 addr6 = zero_addr6;
189 } else {
190 vni = gnvh->vni;
191
192 addr6 = ip6h->saddr;
193 }
194
195 geneve = geneve6_lookup(gs, addr6, vni);
196 #endif
197 }
198 if (!geneve)
199 goto drop;
200
201 if (ip_tunnel_collect_metadata() || gs->collect_md) {
202 __be16 flags;
203
204 flags = TUNNEL_KEY | TUNNEL_GENEVE_OPT |
205 (gnvh->oam ? TUNNEL_OAM : 0) |
206 (gnvh->critical ? TUNNEL_CRIT_OPT : 0);
207
208 tun_dst = udp_tun_rx_dst(skb, sa_family, flags,
209 vni_to_tunnel_id(gnvh->vni),
210 gnvh->opt_len * 4);
211 if (!tun_dst)
212 goto drop;
213 /* Update tunnel dst according to Geneve options. */
214 ip_tunnel_info_opts_set(&tun_dst->u.tun_info,
215 gnvh->options, gnvh->opt_len * 4);
216 } else {
217 /* Drop packets w/ critical options,
218 * since we don't support any...
219 */
220 if (gnvh->critical)
221 goto drop;
222 }
223
224 skb_reset_mac_header(skb);
225 skb_scrub_packet(skb, !net_eq(geneve->net, dev_net(geneve->dev)));
226 skb->protocol = eth_type_trans(skb, geneve->dev);
227 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
228
229 if (tun_dst)
230 skb_dst_set(skb, &tun_dst->dst);
231
232 /* Ignore packet loops (and multicast echo) */
233 if (ether_addr_equal(eth_hdr(skb)->h_source, geneve->dev->dev_addr))
234 goto drop;
235
236 skb_reset_network_header(skb);
237
238 if (iph)
239 err = IP_ECN_decapsulate(iph, skb);
240 #if IS_ENABLED(CONFIG_IPV6)
241 if (ip6h)
242 err = IP6_ECN_decapsulate(ip6h, skb);
243 #endif
244
245 if (unlikely(err)) {
246 if (log_ecn_error) {
247 if (iph)
248 net_info_ratelimited("non-ECT from %pI4 "
249 "with TOS=%#x\n",
250 &iph->saddr, iph->tos);
251 #if IS_ENABLED(CONFIG_IPV6)
252 if (ip6h)
253 net_info_ratelimited("non-ECT from %pI6\n",
254 &ip6h->saddr);
255 #endif
256 }
257 if (err > 1) {
258 ++geneve->dev->stats.rx_frame_errors;
259 ++geneve->dev->stats.rx_errors;
260 goto drop;
261 }
262 }
263
264 stats = this_cpu_ptr(geneve->dev->tstats);
265 u64_stats_update_begin(&stats->syncp);
266 stats->rx_packets++;
267 stats->rx_bytes += skb->len;
268 u64_stats_update_end(&stats->syncp);
269
270 gro_cells_receive(&geneve->gro_cells, skb);
271 return;
272 drop:
273 /* Consume bad packet */
274 kfree_skb(skb);
275 }
276
277 /* Setup stats when device is created */
geneve_init(struct net_device * dev)278 static int geneve_init(struct net_device *dev)
279 {
280 struct geneve_dev *geneve = netdev_priv(dev);
281 int err;
282
283 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
284 if (!dev->tstats)
285 return -ENOMEM;
286
287 err = gro_cells_init(&geneve->gro_cells, dev);
288 if (err) {
289 free_percpu(dev->tstats);
290 return err;
291 }
292
293 return 0;
294 }
295
geneve_uninit(struct net_device * dev)296 static void geneve_uninit(struct net_device *dev)
297 {
298 struct geneve_dev *geneve = netdev_priv(dev);
299
300 gro_cells_destroy(&geneve->gro_cells);
301 free_percpu(dev->tstats);
302 }
303
304 /* Callback from net/ipv4/udp.c to receive packets */
geneve_udp_encap_recv(struct sock * sk,struct sk_buff * skb)305 static int geneve_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
306 {
307 struct genevehdr *geneveh;
308 struct geneve_sock *gs;
309 int opts_len;
310
311 /* Need Geneve and inner Ethernet header to be present */
312 if (unlikely(!pskb_may_pull(skb, GENEVE_BASE_HLEN)))
313 goto drop;
314
315 /* Return packets with reserved bits set */
316 geneveh = geneve_hdr(skb);
317 if (unlikely(geneveh->ver != GENEVE_VER))
318 goto drop;
319
320 if (unlikely(geneveh->proto_type != htons(ETH_P_TEB)))
321 goto drop;
322
323 opts_len = geneveh->opt_len * 4;
324 if (iptunnel_pull_header(skb, GENEVE_BASE_HLEN + opts_len,
325 htons(ETH_P_TEB)))
326 goto drop;
327
328 gs = rcu_dereference_sk_user_data(sk);
329 if (!gs)
330 goto drop;
331
332 geneve_rx(gs, skb);
333 return 0;
334
335 drop:
336 /* Consume bad packet */
337 kfree_skb(skb);
338 return 0;
339 }
340
geneve_create_sock(struct net * net,bool ipv6,__be16 port)341 static struct socket *geneve_create_sock(struct net *net, bool ipv6,
342 __be16 port)
343 {
344 struct socket *sock;
345 struct udp_port_cfg udp_conf;
346 int err;
347
348 memset(&udp_conf, 0, sizeof(udp_conf));
349
350 if (ipv6) {
351 udp_conf.family = AF_INET6;
352 udp_conf.ipv6_v6only = 1;
353 } else {
354 udp_conf.family = AF_INET;
355 udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
356 }
357
358 udp_conf.local_udp_port = port;
359
360 /* Open UDP socket */
361 err = udp_sock_create(net, &udp_conf, &sock);
362 if (err < 0)
363 return ERR_PTR(err);
364
365 return sock;
366 }
367
geneve_notify_add_rx_port(struct geneve_sock * gs)368 static void geneve_notify_add_rx_port(struct geneve_sock *gs)
369 {
370 struct sock *sk = gs->sock->sk;
371 sa_family_t sa_family = sk->sk_family;
372 int err;
373
374 if (sa_family == AF_INET) {
375 err = udp_add_offload(&gs->udp_offloads);
376 if (err)
377 pr_warn("geneve: udp_add_offload failed with status %d\n",
378 err);
379 }
380 }
381
geneve_hlen(struct genevehdr * gh)382 static int geneve_hlen(struct genevehdr *gh)
383 {
384 return sizeof(*gh) + gh->opt_len * 4;
385 }
386
geneve_gro_receive(struct sk_buff ** head,struct sk_buff * skb,struct udp_offload * uoff)387 static struct sk_buff **geneve_gro_receive(struct sk_buff **head,
388 struct sk_buff *skb,
389 struct udp_offload *uoff)
390 {
391 struct sk_buff *p, **pp = NULL;
392 struct genevehdr *gh, *gh2;
393 unsigned int hlen, gh_len, off_gnv;
394 const struct packet_offload *ptype;
395 __be16 type;
396 int flush = 1;
397
398 off_gnv = skb_gro_offset(skb);
399 hlen = off_gnv + sizeof(*gh);
400 gh = skb_gro_header_fast(skb, off_gnv);
401 if (skb_gro_header_hard(skb, hlen)) {
402 gh = skb_gro_header_slow(skb, hlen, off_gnv);
403 if (unlikely(!gh))
404 goto out;
405 }
406
407 if (gh->ver != GENEVE_VER || gh->oam)
408 goto out;
409 gh_len = geneve_hlen(gh);
410
411 hlen = off_gnv + gh_len;
412 if (skb_gro_header_hard(skb, hlen)) {
413 gh = skb_gro_header_slow(skb, hlen, off_gnv);
414 if (unlikely(!gh))
415 goto out;
416 }
417
418 flush = 0;
419
420 for (p = *head; p; p = p->next) {
421 if (!NAPI_GRO_CB(p)->same_flow)
422 continue;
423
424 gh2 = (struct genevehdr *)(p->data + off_gnv);
425 if (gh->opt_len != gh2->opt_len ||
426 memcmp(gh, gh2, gh_len)) {
427 NAPI_GRO_CB(p)->same_flow = 0;
428 continue;
429 }
430 }
431
432 type = gh->proto_type;
433
434 rcu_read_lock();
435 ptype = gro_find_receive_by_type(type);
436 if (!ptype) {
437 flush = 1;
438 goto out_unlock;
439 }
440
441 skb_gro_pull(skb, gh_len);
442 skb_gro_postpull_rcsum(skb, gh, gh_len);
443 pp = ptype->callbacks.gro_receive(head, skb);
444
445 out_unlock:
446 rcu_read_unlock();
447 out:
448 NAPI_GRO_CB(skb)->flush |= flush;
449
450 return pp;
451 }
452
geneve_gro_complete(struct sk_buff * skb,int nhoff,struct udp_offload * uoff)453 static int geneve_gro_complete(struct sk_buff *skb, int nhoff,
454 struct udp_offload *uoff)
455 {
456 struct genevehdr *gh;
457 struct packet_offload *ptype;
458 __be16 type;
459 int gh_len;
460 int err = -ENOSYS;
461
462 udp_tunnel_gro_complete(skb, nhoff);
463
464 gh = (struct genevehdr *)(skb->data + nhoff);
465 gh_len = geneve_hlen(gh);
466 type = gh->proto_type;
467
468 rcu_read_lock();
469 ptype = gro_find_complete_by_type(type);
470 if (ptype)
471 err = ptype->callbacks.gro_complete(skb, nhoff + gh_len);
472
473 rcu_read_unlock();
474 return err;
475 }
476
477 /* Create new listen socket if needed */
geneve_socket_create(struct net * net,__be16 port,bool ipv6)478 static struct geneve_sock *geneve_socket_create(struct net *net, __be16 port,
479 bool ipv6)
480 {
481 struct geneve_net *gn = net_generic(net, geneve_net_id);
482 struct geneve_sock *gs;
483 struct socket *sock;
484 struct udp_tunnel_sock_cfg tunnel_cfg;
485 int h;
486
487 gs = kzalloc(sizeof(*gs), GFP_KERNEL);
488 if (!gs)
489 return ERR_PTR(-ENOMEM);
490
491 sock = geneve_create_sock(net, ipv6, port);
492 if (IS_ERR(sock)) {
493 kfree(gs);
494 return ERR_CAST(sock);
495 }
496
497 gs->sock = sock;
498 gs->refcnt = 1;
499 for (h = 0; h < VNI_HASH_SIZE; ++h)
500 INIT_HLIST_HEAD(&gs->vni_list[h]);
501
502 /* Initialize the geneve udp offloads structure */
503 gs->udp_offloads.port = port;
504 gs->udp_offloads.callbacks.gro_receive = geneve_gro_receive;
505 gs->udp_offloads.callbacks.gro_complete = geneve_gro_complete;
506 geneve_notify_add_rx_port(gs);
507
508 /* Mark socket as an encapsulation socket */
509 tunnel_cfg.sk_user_data = gs;
510 tunnel_cfg.encap_type = 1;
511 tunnel_cfg.encap_rcv = geneve_udp_encap_recv;
512 tunnel_cfg.encap_destroy = NULL;
513 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
514 list_add(&gs->list, &gn->sock_list);
515 return gs;
516 }
517
geneve_notify_del_rx_port(struct geneve_sock * gs)518 static void geneve_notify_del_rx_port(struct geneve_sock *gs)
519 {
520 struct sock *sk = gs->sock->sk;
521 sa_family_t sa_family = sk->sk_family;
522
523 if (sa_family == AF_INET)
524 udp_del_offload(&gs->udp_offloads);
525 }
526
__geneve_sock_release(struct geneve_sock * gs)527 static void __geneve_sock_release(struct geneve_sock *gs)
528 {
529 if (!gs || --gs->refcnt)
530 return;
531
532 list_del(&gs->list);
533 geneve_notify_del_rx_port(gs);
534 udp_tunnel_sock_release(gs->sock);
535 kfree_rcu(gs, rcu);
536 }
537
geneve_sock_release(struct geneve_dev * geneve)538 static void geneve_sock_release(struct geneve_dev *geneve)
539 {
540 __geneve_sock_release(geneve->sock4);
541 #if IS_ENABLED(CONFIG_IPV6)
542 __geneve_sock_release(geneve->sock6);
543 #endif
544 }
545
geneve_find_sock(struct geneve_net * gn,sa_family_t family,__be16 dst_port)546 static struct geneve_sock *geneve_find_sock(struct geneve_net *gn,
547 sa_family_t family,
548 __be16 dst_port)
549 {
550 struct geneve_sock *gs;
551
552 list_for_each_entry(gs, &gn->sock_list, list) {
553 if (inet_sk(gs->sock->sk)->inet_sport == dst_port &&
554 inet_sk(gs->sock->sk)->sk.sk_family == family) {
555 return gs;
556 }
557 }
558 return NULL;
559 }
560
geneve_sock_add(struct geneve_dev * geneve,bool ipv6)561 static int geneve_sock_add(struct geneve_dev *geneve, bool ipv6)
562 {
563 struct net *net = geneve->net;
564 struct geneve_net *gn = net_generic(net, geneve_net_id);
565 struct geneve_sock *gs;
566 __u32 hash;
567
568 gs = geneve_find_sock(gn, ipv6 ? AF_INET6 : AF_INET, geneve->dst_port);
569 if (gs) {
570 gs->refcnt++;
571 goto out;
572 }
573
574 gs = geneve_socket_create(net, geneve->dst_port, ipv6);
575 if (IS_ERR(gs))
576 return PTR_ERR(gs);
577
578 out:
579 gs->collect_md = geneve->collect_md;
580 #if IS_ENABLED(CONFIG_IPV6)
581 if (ipv6)
582 geneve->sock6 = gs;
583 else
584 #endif
585 geneve->sock4 = gs;
586
587 hash = geneve_net_vni_hash(geneve->vni);
588 hlist_add_head_rcu(&geneve->hlist, &gs->vni_list[hash]);
589 return 0;
590 }
591
geneve_open(struct net_device * dev)592 static int geneve_open(struct net_device *dev)
593 {
594 struct geneve_dev *geneve = netdev_priv(dev);
595 bool ipv6 = geneve->remote.sa.sa_family == AF_INET6;
596 bool metadata = geneve->collect_md;
597 int ret = 0;
598
599 geneve->sock4 = NULL;
600 #if IS_ENABLED(CONFIG_IPV6)
601 geneve->sock6 = NULL;
602 if (ipv6 || metadata)
603 ret = geneve_sock_add(geneve, true);
604 #endif
605 if (!ret && (!ipv6 || metadata))
606 ret = geneve_sock_add(geneve, false);
607 if (ret < 0)
608 geneve_sock_release(geneve);
609
610 return ret;
611 }
612
geneve_stop(struct net_device * dev)613 static int geneve_stop(struct net_device *dev)
614 {
615 struct geneve_dev *geneve = netdev_priv(dev);
616
617 if (!hlist_unhashed(&geneve->hlist))
618 hlist_del_rcu(&geneve->hlist);
619 geneve_sock_release(geneve);
620 return 0;
621 }
622
geneve_build_header(struct genevehdr * geneveh,__be16 tun_flags,u8 vni[3],u8 options_len,u8 * options)623 static void geneve_build_header(struct genevehdr *geneveh,
624 __be16 tun_flags, u8 vni[3],
625 u8 options_len, u8 *options)
626 {
627 geneveh->ver = GENEVE_VER;
628 geneveh->opt_len = options_len / 4;
629 geneveh->oam = !!(tun_flags & TUNNEL_OAM);
630 geneveh->critical = !!(tun_flags & TUNNEL_CRIT_OPT);
631 geneveh->rsvd1 = 0;
632 memcpy(geneveh->vni, vni, 3);
633 geneveh->proto_type = htons(ETH_P_TEB);
634 geneveh->rsvd2 = 0;
635
636 memcpy(geneveh->options, options, options_len);
637 }
638
geneve_build_skb(struct rtable * rt,struct sk_buff * skb,__be16 tun_flags,u8 vni[3],u8 opt_len,u8 * opt,bool csum,bool xnet)639 static int geneve_build_skb(struct rtable *rt, struct sk_buff *skb,
640 __be16 tun_flags, u8 vni[3], u8 opt_len, u8 *opt,
641 bool csum, bool xnet)
642 {
643 struct genevehdr *gnvh;
644 int min_headroom;
645 int err;
646
647 skb_scrub_packet(skb, xnet);
648
649 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
650 + GENEVE_BASE_HLEN + opt_len + sizeof(struct iphdr);
651 err = skb_cow_head(skb, min_headroom);
652 if (unlikely(err)) {
653 kfree_skb(skb);
654 goto free_rt;
655 }
656
657 skb = udp_tunnel_handle_offloads(skb, csum);
658 if (IS_ERR(skb)) {
659 err = PTR_ERR(skb);
660 goto free_rt;
661 }
662
663 gnvh = (struct genevehdr *)__skb_push(skb, sizeof(*gnvh) + opt_len);
664 geneve_build_header(gnvh, tun_flags, vni, opt_len, opt);
665
666 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
667 return 0;
668
669 free_rt:
670 ip_rt_put(rt);
671 return err;
672 }
673
674 #if IS_ENABLED(CONFIG_IPV6)
geneve6_build_skb(struct dst_entry * dst,struct sk_buff * skb,__be16 tun_flags,u8 vni[3],u8 opt_len,u8 * opt,bool csum,bool xnet)675 static int geneve6_build_skb(struct dst_entry *dst, struct sk_buff *skb,
676 __be16 tun_flags, u8 vni[3], u8 opt_len, u8 *opt,
677 bool csum, bool xnet)
678 {
679 struct genevehdr *gnvh;
680 int min_headroom;
681 int err;
682
683 skb_scrub_packet(skb, xnet);
684
685 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
686 + GENEVE_BASE_HLEN + opt_len + sizeof(struct ipv6hdr);
687 err = skb_cow_head(skb, min_headroom);
688 if (unlikely(err)) {
689 kfree_skb(skb);
690 goto free_dst;
691 }
692
693 skb = udp_tunnel_handle_offloads(skb, csum);
694 if (IS_ERR(skb)) {
695 err = PTR_ERR(skb);
696 goto free_dst;
697 }
698
699 gnvh = (struct genevehdr *)__skb_push(skb, sizeof(*gnvh) + opt_len);
700 geneve_build_header(gnvh, tun_flags, vni, opt_len, opt);
701
702 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
703 return 0;
704
705 free_dst:
706 dst_release(dst);
707 return err;
708 }
709 #endif
710
geneve_get_v4_rt(struct sk_buff * skb,struct net_device * dev,struct flowi4 * fl4,struct ip_tunnel_info * info)711 static struct rtable *geneve_get_v4_rt(struct sk_buff *skb,
712 struct net_device *dev,
713 struct flowi4 *fl4,
714 struct ip_tunnel_info *info)
715 {
716 struct geneve_dev *geneve = netdev_priv(dev);
717 struct rtable *rt = NULL;
718 __u8 tos;
719
720 memset(fl4, 0, sizeof(*fl4));
721 fl4->flowi4_mark = skb->mark;
722 fl4->flowi4_proto = IPPROTO_UDP;
723
724 if (info) {
725 fl4->daddr = info->key.u.ipv4.dst;
726 fl4->saddr = info->key.u.ipv4.src;
727 fl4->flowi4_tos = RT_TOS(info->key.tos);
728 } else {
729 tos = geneve->tos;
730 if (tos == 1) {
731 const struct iphdr *iip = ip_hdr(skb);
732
733 tos = ip_tunnel_get_dsfield(iip, skb);
734 }
735
736 fl4->flowi4_tos = RT_TOS(tos);
737 fl4->daddr = geneve->remote.sin.sin_addr.s_addr;
738 }
739
740 rt = ip_route_output_key(geneve->net, fl4);
741 if (IS_ERR(rt)) {
742 netdev_dbg(dev, "no route to %pI4\n", &fl4->daddr);
743 return ERR_PTR(-ENETUNREACH);
744 }
745 if (rt->dst.dev == dev) { /* is this necessary? */
746 netdev_dbg(dev, "circular route to %pI4\n", &fl4->daddr);
747 ip_rt_put(rt);
748 return ERR_PTR(-ELOOP);
749 }
750 return rt;
751 }
752
753 #if IS_ENABLED(CONFIG_IPV6)
geneve_get_v6_dst(struct sk_buff * skb,struct net_device * dev,struct flowi6 * fl6,struct ip_tunnel_info * info)754 static struct dst_entry *geneve_get_v6_dst(struct sk_buff *skb,
755 struct net_device *dev,
756 struct flowi6 *fl6,
757 struct ip_tunnel_info *info)
758 {
759 struct geneve_dev *geneve = netdev_priv(dev);
760 struct geneve_sock *gs6 = geneve->sock6;
761 struct dst_entry *dst = NULL;
762 __u8 prio;
763
764 memset(fl6, 0, sizeof(*fl6));
765 fl6->flowi6_mark = skb->mark;
766 fl6->flowi6_proto = IPPROTO_UDP;
767
768 if (info) {
769 fl6->daddr = info->key.u.ipv6.dst;
770 fl6->saddr = info->key.u.ipv6.src;
771 fl6->flowi6_tos = RT_TOS(info->key.tos);
772 } else {
773 prio = geneve->tos;
774 if (prio == 1) {
775 const struct iphdr *iip = ip_hdr(skb);
776
777 prio = ip_tunnel_get_dsfield(iip, skb);
778 }
779
780 fl6->flowi6_tos = RT_TOS(prio);
781 fl6->daddr = geneve->remote.sin6.sin6_addr;
782 }
783
784 if (ipv6_stub->ipv6_dst_lookup(geneve->net, gs6->sock->sk, &dst, fl6)) {
785 netdev_dbg(dev, "no route to %pI6\n", &fl6->daddr);
786 return ERR_PTR(-ENETUNREACH);
787 }
788 if (dst->dev == dev) { /* is this necessary? */
789 netdev_dbg(dev, "circular route to %pI6\n", &fl6->daddr);
790 dst_release(dst);
791 return ERR_PTR(-ELOOP);
792 }
793
794 return dst;
795 }
796 #endif
797
798 /* Convert 64 bit tunnel ID to 24 bit VNI. */
tunnel_id_to_vni(__be64 tun_id,__u8 * vni)799 static void tunnel_id_to_vni(__be64 tun_id, __u8 *vni)
800 {
801 #ifdef __BIG_ENDIAN
802 vni[0] = (__force __u8)(tun_id >> 16);
803 vni[1] = (__force __u8)(tun_id >> 8);
804 vni[2] = (__force __u8)tun_id;
805 #else
806 vni[0] = (__force __u8)((__force u64)tun_id >> 40);
807 vni[1] = (__force __u8)((__force u64)tun_id >> 48);
808 vni[2] = (__force __u8)((__force u64)tun_id >> 56);
809 #endif
810 }
811
geneve_xmit_skb(struct sk_buff * skb,struct net_device * dev,struct ip_tunnel_info * info)812 static netdev_tx_t geneve_xmit_skb(struct sk_buff *skb, struct net_device *dev,
813 struct ip_tunnel_info *info)
814 {
815 struct geneve_dev *geneve = netdev_priv(dev);
816 struct geneve_sock *gs4 = geneve->sock4;
817 struct rtable *rt = NULL;
818 const struct iphdr *iip; /* interior IP header */
819 int err = -EINVAL;
820 struct flowi4 fl4;
821 __u8 tos, ttl;
822 __be16 sport;
823 bool udp_csum;
824 __be16 df;
825 bool xnet = !net_eq(geneve->net, dev_net(geneve->dev));
826
827 if (geneve->collect_md) {
828 if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
829 netdev_dbg(dev, "no tunnel metadata\n");
830 goto tx_error;
831 }
832 if (info && ip_tunnel_info_af(info) != AF_INET)
833 goto tx_error;
834 }
835
836 rt = geneve_get_v4_rt(skb, dev, &fl4, info);
837 if (IS_ERR(rt)) {
838 err = PTR_ERR(rt);
839 goto tx_error;
840 }
841
842 sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true);
843 skb_reset_mac_header(skb);
844
845 iip = ip_hdr(skb);
846
847 if (info) {
848 const struct ip_tunnel_key *key = &info->key;
849 u8 *opts = NULL;
850 u8 vni[3];
851
852 tunnel_id_to_vni(key->tun_id, vni);
853 if (key->tun_flags & TUNNEL_GENEVE_OPT)
854 opts = ip_tunnel_info_opts(info);
855
856 udp_csum = !!(key->tun_flags & TUNNEL_CSUM);
857 err = geneve_build_skb(rt, skb, key->tun_flags, vni,
858 info->options_len, opts, udp_csum, xnet);
859 if (unlikely(err))
860 goto err;
861
862 tos = ip_tunnel_ecn_encap(key->tos, iip, skb);
863 ttl = key->ttl;
864 df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
865 } else {
866 udp_csum = false;
867 err = geneve_build_skb(rt, skb, 0, geneve->vni,
868 0, NULL, udp_csum, xnet);
869 if (unlikely(err))
870 goto err;
871
872 tos = ip_tunnel_ecn_encap(fl4.flowi4_tos, iip, skb);
873 ttl = geneve->ttl;
874 if (!ttl && IN_MULTICAST(ntohl(fl4.daddr)))
875 ttl = 1;
876 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
877 df = 0;
878 }
879 err = udp_tunnel_xmit_skb(rt, gs4->sock->sk, skb, fl4.saddr, fl4.daddr,
880 tos, ttl, df, sport, geneve->dst_port,
881 !net_eq(geneve->net, dev_net(geneve->dev)),
882 !udp_csum);
883
884 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
885 return NETDEV_TX_OK;
886
887 tx_error:
888 dev_kfree_skb(skb);
889 err:
890 if (err == -ELOOP)
891 dev->stats.collisions++;
892 else if (err == -ENETUNREACH)
893 dev->stats.tx_carrier_errors++;
894 else
895 dev->stats.tx_errors++;
896 return NETDEV_TX_OK;
897 }
898
899 #if IS_ENABLED(CONFIG_IPV6)
geneve6_xmit_skb(struct sk_buff * skb,struct net_device * dev,struct ip_tunnel_info * info)900 static netdev_tx_t geneve6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
901 struct ip_tunnel_info *info)
902 {
903 struct geneve_dev *geneve = netdev_priv(dev);
904 struct geneve_sock *gs6 = geneve->sock6;
905 struct dst_entry *dst = NULL;
906 const struct iphdr *iip; /* interior IP header */
907 int err = -EINVAL;
908 struct flowi6 fl6;
909 __u8 prio, ttl;
910 __be16 sport;
911 bool udp_csum;
912 bool xnet = !net_eq(geneve->net, dev_net(geneve->dev));
913
914 if (geneve->collect_md) {
915 if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
916 netdev_dbg(dev, "no tunnel metadata\n");
917 goto tx_error;
918 }
919 }
920
921 dst = geneve_get_v6_dst(skb, dev, &fl6, info);
922 if (IS_ERR(dst)) {
923 err = PTR_ERR(dst);
924 goto tx_error;
925 }
926
927 sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true);
928 skb_reset_mac_header(skb);
929
930 iip = ip_hdr(skb);
931
932 if (info) {
933 const struct ip_tunnel_key *key = &info->key;
934 u8 *opts = NULL;
935 u8 vni[3];
936
937 tunnel_id_to_vni(key->tun_id, vni);
938 if (key->tun_flags & TUNNEL_GENEVE_OPT)
939 opts = ip_tunnel_info_opts(info);
940
941 udp_csum = !!(key->tun_flags & TUNNEL_CSUM);
942 err = geneve6_build_skb(dst, skb, key->tun_flags, vni,
943 info->options_len, opts,
944 udp_csum, xnet);
945 if (unlikely(err))
946 goto err;
947
948 prio = ip_tunnel_ecn_encap(key->tos, iip, skb);
949 ttl = key->ttl;
950 } else {
951 udp_csum = false;
952 err = geneve6_build_skb(dst, skb, 0, geneve->vni,
953 0, NULL, udp_csum, xnet);
954 if (unlikely(err))
955 goto err;
956
957 prio = ip_tunnel_ecn_encap(fl6.flowi6_tos, iip, skb);
958 ttl = geneve->ttl;
959 if (!ttl && ipv6_addr_is_multicast(&fl6.daddr))
960 ttl = 1;
961 ttl = ttl ? : ip6_dst_hoplimit(dst);
962 }
963 err = udp_tunnel6_xmit_skb(dst, gs6->sock->sk, skb, dev,
964 &fl6.saddr, &fl6.daddr, prio, ttl,
965 sport, geneve->dst_port, !udp_csum);
966 return NETDEV_TX_OK;
967
968 tx_error:
969 dev_kfree_skb(skb);
970 err:
971 if (err == -ELOOP)
972 dev->stats.collisions++;
973 else if (err == -ENETUNREACH)
974 dev->stats.tx_carrier_errors++;
975 else
976 dev->stats.tx_errors++;
977 return NETDEV_TX_OK;
978 }
979 #endif
980
geneve_xmit(struct sk_buff * skb,struct net_device * dev)981 static netdev_tx_t geneve_xmit(struct sk_buff *skb, struct net_device *dev)
982 {
983 struct geneve_dev *geneve = netdev_priv(dev);
984 struct ip_tunnel_info *info = NULL;
985
986 if (geneve->collect_md)
987 info = skb_tunnel_info(skb);
988
989 #if IS_ENABLED(CONFIG_IPV6)
990 if ((info && ip_tunnel_info_af(info) == AF_INET6) ||
991 (!info && geneve->remote.sa.sa_family == AF_INET6))
992 return geneve6_xmit_skb(skb, dev, info);
993 #endif
994 return geneve_xmit_skb(skb, dev, info);
995 }
996
geneve_change_mtu(struct net_device * dev,int new_mtu)997 static int geneve_change_mtu(struct net_device *dev, int new_mtu)
998 {
999 /* GENEVE overhead is not fixed, so we can't enforce a more
1000 * precise max MTU.
1001 */
1002 if (new_mtu < 68 || new_mtu > IP_MAX_MTU)
1003 return -EINVAL;
1004 dev->mtu = new_mtu;
1005 return 0;
1006 }
1007
geneve_fill_metadata_dst(struct net_device * dev,struct sk_buff * skb)1008 static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
1009 {
1010 struct ip_tunnel_info *info = skb_tunnel_info(skb);
1011 struct geneve_dev *geneve = netdev_priv(dev);
1012 struct rtable *rt;
1013 struct flowi4 fl4;
1014 #if IS_ENABLED(CONFIG_IPV6)
1015 struct dst_entry *dst;
1016 struct flowi6 fl6;
1017 #endif
1018
1019 if (ip_tunnel_info_af(info) == AF_INET) {
1020 rt = geneve_get_v4_rt(skb, dev, &fl4, info);
1021 if (IS_ERR(rt))
1022 return PTR_ERR(rt);
1023
1024 ip_rt_put(rt);
1025 info->key.u.ipv4.src = fl4.saddr;
1026 #if IS_ENABLED(CONFIG_IPV6)
1027 } else if (ip_tunnel_info_af(info) == AF_INET6) {
1028 dst = geneve_get_v6_dst(skb, dev, &fl6, info);
1029 if (IS_ERR(dst))
1030 return PTR_ERR(dst);
1031
1032 dst_release(dst);
1033 info->key.u.ipv6.src = fl6.saddr;
1034 #endif
1035 } else {
1036 return -EINVAL;
1037 }
1038
1039 info->key.tp_src = udp_flow_src_port(geneve->net, skb,
1040 1, USHRT_MAX, true);
1041 info->key.tp_dst = geneve->dst_port;
1042 return 0;
1043 }
1044
1045 static const struct net_device_ops geneve_netdev_ops = {
1046 .ndo_init = geneve_init,
1047 .ndo_uninit = geneve_uninit,
1048 .ndo_open = geneve_open,
1049 .ndo_stop = geneve_stop,
1050 .ndo_start_xmit = geneve_xmit,
1051 .ndo_get_stats64 = ip_tunnel_get_stats64,
1052 .ndo_change_mtu = geneve_change_mtu,
1053 .ndo_validate_addr = eth_validate_addr,
1054 .ndo_set_mac_address = eth_mac_addr,
1055 .ndo_fill_metadata_dst = geneve_fill_metadata_dst,
1056 };
1057
geneve_get_drvinfo(struct net_device * dev,struct ethtool_drvinfo * drvinfo)1058 static void geneve_get_drvinfo(struct net_device *dev,
1059 struct ethtool_drvinfo *drvinfo)
1060 {
1061 strlcpy(drvinfo->version, GENEVE_NETDEV_VER, sizeof(drvinfo->version));
1062 strlcpy(drvinfo->driver, "geneve", sizeof(drvinfo->driver));
1063 }
1064
1065 static const struct ethtool_ops geneve_ethtool_ops = {
1066 .get_drvinfo = geneve_get_drvinfo,
1067 .get_link = ethtool_op_get_link,
1068 };
1069
1070 /* Info for udev, that this is a virtual tunnel endpoint */
1071 static struct device_type geneve_type = {
1072 .name = "geneve",
1073 };
1074
1075 /* Initialize the device structure. */
geneve_setup(struct net_device * dev)1076 static void geneve_setup(struct net_device *dev)
1077 {
1078 ether_setup(dev);
1079
1080 dev->netdev_ops = &geneve_netdev_ops;
1081 dev->ethtool_ops = &geneve_ethtool_ops;
1082 dev->destructor = free_netdev;
1083
1084 SET_NETDEV_DEVTYPE(dev, &geneve_type);
1085
1086 dev->features |= NETIF_F_LLTX;
1087 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
1088 dev->features |= NETIF_F_RXCSUM;
1089 dev->features |= NETIF_F_GSO_SOFTWARE;
1090
1091 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
1092 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1093
1094 netif_keep_dst(dev);
1095 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
1096 eth_hw_addr_random(dev);
1097 }
1098
1099 static const struct nla_policy geneve_policy[IFLA_GENEVE_MAX + 1] = {
1100 [IFLA_GENEVE_ID] = { .type = NLA_U32 },
1101 [IFLA_GENEVE_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
1102 [IFLA_GENEVE_REMOTE6] = { .len = sizeof(struct in6_addr) },
1103 [IFLA_GENEVE_TTL] = { .type = NLA_U8 },
1104 [IFLA_GENEVE_TOS] = { .type = NLA_U8 },
1105 [IFLA_GENEVE_PORT] = { .type = NLA_U16 },
1106 [IFLA_GENEVE_COLLECT_METADATA] = { .type = NLA_FLAG },
1107 };
1108
geneve_validate(struct nlattr * tb[],struct nlattr * data[])1109 static int geneve_validate(struct nlattr *tb[], struct nlattr *data[])
1110 {
1111 if (tb[IFLA_ADDRESS]) {
1112 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1113 return -EINVAL;
1114
1115 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1116 return -EADDRNOTAVAIL;
1117 }
1118
1119 if (!data)
1120 return -EINVAL;
1121
1122 if (data[IFLA_GENEVE_ID]) {
1123 __u32 vni = nla_get_u32(data[IFLA_GENEVE_ID]);
1124
1125 if (vni >= GENEVE_VID_MASK)
1126 return -ERANGE;
1127 }
1128
1129 return 0;
1130 }
1131
geneve_find_dev(struct geneve_net * gn,__be16 dst_port,union geneve_addr * remote,u8 vni[],bool * tun_on_same_port,bool * tun_collect_md)1132 static struct geneve_dev *geneve_find_dev(struct geneve_net *gn,
1133 __be16 dst_port,
1134 union geneve_addr *remote,
1135 u8 vni[],
1136 bool *tun_on_same_port,
1137 bool *tun_collect_md)
1138 {
1139 struct geneve_dev *geneve, *t;
1140
1141 *tun_on_same_port = false;
1142 *tun_collect_md = false;
1143 t = NULL;
1144 list_for_each_entry(geneve, &gn->geneve_list, next) {
1145 if (geneve->dst_port == dst_port) {
1146 *tun_collect_md = geneve->collect_md;
1147 *tun_on_same_port = true;
1148 }
1149 if (!memcmp(vni, geneve->vni, sizeof(geneve->vni)) &&
1150 !memcmp(remote, &geneve->remote, sizeof(geneve->remote)) &&
1151 dst_port == geneve->dst_port)
1152 t = geneve;
1153 }
1154 return t;
1155 }
1156
geneve_configure(struct net * net,struct net_device * dev,union geneve_addr * remote,__u32 vni,__u8 ttl,__u8 tos,__be16 dst_port,bool metadata)1157 static int geneve_configure(struct net *net, struct net_device *dev,
1158 union geneve_addr *remote,
1159 __u32 vni, __u8 ttl, __u8 tos, __be16 dst_port,
1160 bool metadata)
1161 {
1162 struct geneve_net *gn = net_generic(net, geneve_net_id);
1163 struct geneve_dev *t, *geneve = netdev_priv(dev);
1164 bool tun_collect_md, tun_on_same_port;
1165 int err, encap_len;
1166
1167 if (!remote)
1168 return -EINVAL;
1169 if (metadata &&
1170 (remote->sa.sa_family != AF_UNSPEC || vni || tos || ttl))
1171 return -EINVAL;
1172
1173 geneve->net = net;
1174 geneve->dev = dev;
1175
1176 geneve->vni[0] = (vni & 0x00ff0000) >> 16;
1177 geneve->vni[1] = (vni & 0x0000ff00) >> 8;
1178 geneve->vni[2] = vni & 0x000000ff;
1179
1180 if ((remote->sa.sa_family == AF_INET &&
1181 IN_MULTICAST(ntohl(remote->sin.sin_addr.s_addr))) ||
1182 (remote->sa.sa_family == AF_INET6 &&
1183 ipv6_addr_is_multicast(&remote->sin6.sin6_addr)))
1184 return -EINVAL;
1185 geneve->remote = *remote;
1186
1187 geneve->ttl = ttl;
1188 geneve->tos = tos;
1189 geneve->dst_port = dst_port;
1190 geneve->collect_md = metadata;
1191
1192 t = geneve_find_dev(gn, dst_port, remote, geneve->vni,
1193 &tun_on_same_port, &tun_collect_md);
1194 if (t)
1195 return -EBUSY;
1196
1197 /* make enough headroom for basic scenario */
1198 encap_len = GENEVE_BASE_HLEN + ETH_HLEN;
1199 if (remote->sa.sa_family == AF_INET)
1200 encap_len += sizeof(struct iphdr);
1201 else
1202 encap_len += sizeof(struct ipv6hdr);
1203 dev->needed_headroom = encap_len + ETH_HLEN;
1204
1205 if (metadata) {
1206 if (tun_on_same_port)
1207 return -EPERM;
1208 } else {
1209 if (tun_collect_md)
1210 return -EPERM;
1211 }
1212
1213 err = register_netdevice(dev);
1214 if (err)
1215 return err;
1216
1217 list_add(&geneve->next, &gn->geneve_list);
1218 return 0;
1219 }
1220
geneve_newlink(struct net * net,struct net_device * dev,struct nlattr * tb[],struct nlattr * data[])1221 static int geneve_newlink(struct net *net, struct net_device *dev,
1222 struct nlattr *tb[], struct nlattr *data[])
1223 {
1224 __be16 dst_port = htons(GENEVE_UDP_PORT);
1225 __u8 ttl = 0, tos = 0;
1226 bool metadata = false;
1227 union geneve_addr remote = geneve_remote_unspec;
1228 __u32 vni = 0;
1229
1230 if (data[IFLA_GENEVE_REMOTE] && data[IFLA_GENEVE_REMOTE6])
1231 return -EINVAL;
1232
1233 if (data[IFLA_GENEVE_REMOTE]) {
1234 remote.sa.sa_family = AF_INET;
1235 remote.sin.sin_addr.s_addr =
1236 nla_get_in_addr(data[IFLA_GENEVE_REMOTE]);
1237 }
1238
1239 if (data[IFLA_GENEVE_REMOTE6]) {
1240 if (!IS_ENABLED(CONFIG_IPV6))
1241 return -EPFNOSUPPORT;
1242
1243 remote.sa.sa_family = AF_INET6;
1244 remote.sin6.sin6_addr =
1245 nla_get_in6_addr(data[IFLA_GENEVE_REMOTE6]);
1246
1247 if (ipv6_addr_type(&remote.sin6.sin6_addr) &
1248 IPV6_ADDR_LINKLOCAL) {
1249 netdev_dbg(dev, "link-local remote is unsupported\n");
1250 return -EINVAL;
1251 }
1252 }
1253
1254 if (data[IFLA_GENEVE_ID])
1255 vni = nla_get_u32(data[IFLA_GENEVE_ID]);
1256
1257 if (data[IFLA_GENEVE_TTL])
1258 ttl = nla_get_u8(data[IFLA_GENEVE_TTL]);
1259
1260 if (data[IFLA_GENEVE_TOS])
1261 tos = nla_get_u8(data[IFLA_GENEVE_TOS]);
1262
1263 if (data[IFLA_GENEVE_PORT])
1264 dst_port = nla_get_be16(data[IFLA_GENEVE_PORT]);
1265
1266 if (data[IFLA_GENEVE_COLLECT_METADATA])
1267 metadata = true;
1268
1269 return geneve_configure(net, dev, &remote, vni, ttl, tos, dst_port,
1270 metadata);
1271 }
1272
geneve_dellink(struct net_device * dev,struct list_head * head)1273 static void geneve_dellink(struct net_device *dev, struct list_head *head)
1274 {
1275 struct geneve_dev *geneve = netdev_priv(dev);
1276
1277 list_del(&geneve->next);
1278 unregister_netdevice_queue(dev, head);
1279 }
1280
geneve_get_size(const struct net_device * dev)1281 static size_t geneve_get_size(const struct net_device *dev)
1282 {
1283 return nla_total_size(sizeof(__u32)) + /* IFLA_GENEVE_ID */
1284 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_GENEVE_REMOTE{6} */
1285 nla_total_size(sizeof(__u8)) + /* IFLA_GENEVE_TTL */
1286 nla_total_size(sizeof(__u8)) + /* IFLA_GENEVE_TOS */
1287 nla_total_size(sizeof(__be16)) + /* IFLA_GENEVE_PORT */
1288 nla_total_size(0) + /* IFLA_GENEVE_COLLECT_METADATA */
1289 0;
1290 }
1291
geneve_fill_info(struct sk_buff * skb,const struct net_device * dev)1292 static int geneve_fill_info(struct sk_buff *skb, const struct net_device *dev)
1293 {
1294 struct geneve_dev *geneve = netdev_priv(dev);
1295 __u32 vni;
1296
1297 vni = (geneve->vni[0] << 16) | (geneve->vni[1] << 8) | geneve->vni[2];
1298 if (nla_put_u32(skb, IFLA_GENEVE_ID, vni))
1299 goto nla_put_failure;
1300
1301 if (geneve->remote.sa.sa_family == AF_INET) {
1302 if (nla_put_in_addr(skb, IFLA_GENEVE_REMOTE,
1303 geneve->remote.sin.sin_addr.s_addr))
1304 goto nla_put_failure;
1305 #if IS_ENABLED(CONFIG_IPV6)
1306 } else {
1307 if (nla_put_in6_addr(skb, IFLA_GENEVE_REMOTE6,
1308 &geneve->remote.sin6.sin6_addr))
1309 goto nla_put_failure;
1310 #endif
1311 }
1312
1313 if (nla_put_u8(skb, IFLA_GENEVE_TTL, geneve->ttl) ||
1314 nla_put_u8(skb, IFLA_GENEVE_TOS, geneve->tos))
1315 goto nla_put_failure;
1316
1317 if (nla_put_be16(skb, IFLA_GENEVE_PORT, geneve->dst_port))
1318 goto nla_put_failure;
1319
1320 if (geneve->collect_md) {
1321 if (nla_put_flag(skb, IFLA_GENEVE_COLLECT_METADATA))
1322 goto nla_put_failure;
1323 }
1324
1325 return 0;
1326
1327 nla_put_failure:
1328 return -EMSGSIZE;
1329 }
1330
1331 static struct rtnl_link_ops geneve_link_ops __read_mostly = {
1332 .kind = "geneve",
1333 .maxtype = IFLA_GENEVE_MAX,
1334 .policy = geneve_policy,
1335 .priv_size = sizeof(struct geneve_dev),
1336 .setup = geneve_setup,
1337 .validate = geneve_validate,
1338 .newlink = geneve_newlink,
1339 .dellink = geneve_dellink,
1340 .get_size = geneve_get_size,
1341 .fill_info = geneve_fill_info,
1342 };
1343
geneve_dev_create_fb(struct net * net,const char * name,u8 name_assign_type,u16 dst_port)1344 struct net_device *geneve_dev_create_fb(struct net *net, const char *name,
1345 u8 name_assign_type, u16 dst_port)
1346 {
1347 struct nlattr *tb[IFLA_MAX + 1];
1348 struct net_device *dev;
1349 int err;
1350
1351 memset(tb, 0, sizeof(tb));
1352 dev = rtnl_create_link(net, name, name_assign_type,
1353 &geneve_link_ops, tb);
1354 if (IS_ERR(dev))
1355 return dev;
1356
1357 err = geneve_configure(net, dev, &geneve_remote_unspec,
1358 0, 0, 0, htons(dst_port), true);
1359 if (err)
1360 goto err;
1361
1362 /* openvswitch users expect packet sizes to be unrestricted,
1363 * so set the largest MTU we can.
1364 */
1365 err = geneve_change_mtu(dev, IP_MAX_MTU);
1366 if (err)
1367 goto err;
1368
1369 return dev;
1370
1371 err:
1372 free_netdev(dev);
1373 return ERR_PTR(err);
1374 }
1375 EXPORT_SYMBOL_GPL(geneve_dev_create_fb);
1376
geneve_init_net(struct net * net)1377 static __net_init int geneve_init_net(struct net *net)
1378 {
1379 struct geneve_net *gn = net_generic(net, geneve_net_id);
1380
1381 INIT_LIST_HEAD(&gn->geneve_list);
1382 INIT_LIST_HEAD(&gn->sock_list);
1383 return 0;
1384 }
1385
geneve_exit_net(struct net * net)1386 static void __net_exit geneve_exit_net(struct net *net)
1387 {
1388 struct geneve_net *gn = net_generic(net, geneve_net_id);
1389 struct geneve_dev *geneve, *next;
1390 struct net_device *dev, *aux;
1391 LIST_HEAD(list);
1392
1393 rtnl_lock();
1394
1395 /* gather any geneve devices that were moved into this ns */
1396 for_each_netdev_safe(net, dev, aux)
1397 if (dev->rtnl_link_ops == &geneve_link_ops)
1398 unregister_netdevice_queue(dev, &list);
1399
1400 /* now gather any other geneve devices that were created in this ns */
1401 list_for_each_entry_safe(geneve, next, &gn->geneve_list, next) {
1402 /* If geneve->dev is in the same netns, it was already added
1403 * to the list by the previous loop.
1404 */
1405 if (!net_eq(dev_net(geneve->dev), net))
1406 unregister_netdevice_queue(geneve->dev, &list);
1407 }
1408
1409 /* unregister the devices gathered above */
1410 unregister_netdevice_many(&list);
1411 rtnl_unlock();
1412 }
1413
1414 static struct pernet_operations geneve_net_ops = {
1415 .init = geneve_init_net,
1416 .exit = geneve_exit_net,
1417 .id = &geneve_net_id,
1418 .size = sizeof(struct geneve_net),
1419 };
1420
geneve_init_module(void)1421 static int __init geneve_init_module(void)
1422 {
1423 int rc;
1424
1425 rc = register_pernet_subsys(&geneve_net_ops);
1426 if (rc)
1427 goto out1;
1428
1429 rc = rtnl_link_register(&geneve_link_ops);
1430 if (rc)
1431 goto out2;
1432
1433 return 0;
1434 out2:
1435 unregister_pernet_subsys(&geneve_net_ops);
1436 out1:
1437 return rc;
1438 }
1439 late_initcall(geneve_init_module);
1440
geneve_cleanup_module(void)1441 static void __exit geneve_cleanup_module(void)
1442 {
1443 rtnl_link_unregister(&geneve_link_ops);
1444 unregister_pernet_subsys(&geneve_net_ops);
1445 }
1446 module_exit(geneve_cleanup_module);
1447
1448 MODULE_LICENSE("GPL");
1449 MODULE_VERSION(GENEVE_NETDEV_VER);
1450 MODULE_AUTHOR("John W. Linville <linville@tuxdriver.com>");
1451 MODULE_DESCRIPTION("Interface driver for GENEVE encapsulated traffic");
1452 MODULE_ALIAS_RTNL_LINK("geneve");
1453