Loading...
1// SPDX-License-Identifier: GPL-2.0
2/* Bareudp: UDP tunnel encasulation for different Payload types like
3 * MPLS, NSH, IP, etc.
4 * Copyright (c) 2019 Nokia, Inc.
5 * Authors: Martin Varghese, <martin.varghese@nokia.com>
6 */
7
8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10#include <linux/kernel.h>
11#include <linux/module.h>
12#include <linux/etherdevice.h>
13#include <linux/hash.h>
14#include <net/dst_metadata.h>
15#include <net/gro_cells.h>
16#include <net/rtnetlink.h>
17#include <net/protocol.h>
18#include <net/ip6_tunnel.h>
19#include <net/ip_tunnels.h>
20#include <net/udp_tunnel.h>
21#include <net/bareudp.h>
22
23#define BAREUDP_BASE_HLEN sizeof(struct udphdr)
24#define BAREUDP_IPV4_HLEN (sizeof(struct iphdr) + \
25 sizeof(struct udphdr))
26#define BAREUDP_IPV6_HLEN (sizeof(struct ipv6hdr) + \
27 sizeof(struct udphdr))
28
29static bool log_ecn_error = true;
30module_param(log_ecn_error, bool, 0644);
31MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
32
33/* per-network namespace private data for this module */
34
35static unsigned int bareudp_net_id;
36
37struct bareudp_net {
38 struct list_head bareudp_list;
39};
40
41struct bareudp_conf {
42 __be16 ethertype;
43 __be16 port;
44 u16 sport_min;
45 bool multi_proto_mode;
46};
47
48/* Pseudo network device */
49struct bareudp_dev {
50 struct net *net; /* netns for packet i/o */
51 struct net_device *dev; /* netdev for bareudp tunnel */
52 __be16 ethertype;
53 __be16 port;
54 u16 sport_min;
55 bool multi_proto_mode;
56 struct socket __rcu *sock;
57 struct list_head next; /* bareudp node on namespace list */
58 struct gro_cells gro_cells;
59};
60
61static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
62{
63 struct metadata_dst *tun_dst = NULL;
64 IP_TUNNEL_DECLARE_FLAGS(key) = { };
65 struct bareudp_dev *bareudp;
66 unsigned short family;
67 unsigned int len;
68 __be16 proto;
69 void *oiph;
70 int err;
71 int nh;
72
73 bareudp = rcu_dereference_sk_user_data(sk);
74 if (!bareudp)
75 goto drop;
76
77 if (skb->protocol == htons(ETH_P_IP))
78 family = AF_INET;
79 else
80 family = AF_INET6;
81
82 if (bareudp->ethertype == htons(ETH_P_IP)) {
83 __u8 ipversion;
84
85 if (skb_copy_bits(skb, BAREUDP_BASE_HLEN, &ipversion,
86 sizeof(ipversion))) {
87 dev_core_stats_rx_dropped_inc(bareudp->dev);
88 goto drop;
89 }
90 ipversion >>= 4;
91
92 if (ipversion == 4) {
93 proto = htons(ETH_P_IP);
94 } else if (ipversion == 6 && bareudp->multi_proto_mode) {
95 proto = htons(ETH_P_IPV6);
96 } else {
97 dev_core_stats_rx_dropped_inc(bareudp->dev);
98 goto drop;
99 }
100 } else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
101 struct iphdr *tunnel_hdr;
102
103 tunnel_hdr = (struct iphdr *)skb_network_header(skb);
104 if (tunnel_hdr->version == 4) {
105 if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
106 proto = bareudp->ethertype;
107 } else if (bareudp->multi_proto_mode &&
108 ipv4_is_multicast(tunnel_hdr->daddr)) {
109 proto = htons(ETH_P_MPLS_MC);
110 } else {
111 dev_core_stats_rx_dropped_inc(bareudp->dev);
112 goto drop;
113 }
114 } else {
115 int addr_type;
116 struct ipv6hdr *tunnel_hdr_v6;
117
118 tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
119 addr_type =
120 ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
121 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
122 proto = bareudp->ethertype;
123 } else if (bareudp->multi_proto_mode &&
124 (addr_type & IPV6_ADDR_MULTICAST)) {
125 proto = htons(ETH_P_MPLS_MC);
126 } else {
127 dev_core_stats_rx_dropped_inc(bareudp->dev);
128 goto drop;
129 }
130 }
131 } else {
132 proto = bareudp->ethertype;
133 }
134
135 if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
136 proto,
137 !net_eq(bareudp->net,
138 dev_net(bareudp->dev)))) {
139 dev_core_stats_rx_dropped_inc(bareudp->dev);
140 goto drop;
141 }
142
143 __set_bit(IP_TUNNEL_KEY_BIT, key);
144
145 tun_dst = udp_tun_rx_dst(skb, family, key, 0, 0);
146 if (!tun_dst) {
147 dev_core_stats_rx_dropped_inc(bareudp->dev);
148 goto drop;
149 }
150 skb_dst_set(skb, &tun_dst->dst);
151 skb->dev = bareudp->dev;
152 skb_reset_mac_header(skb);
153
154 /* Save offset of outer header relative to skb->head,
155 * because we are going to reset the network header to the inner header
156 * and might change skb->head.
157 */
158 nh = skb_network_header(skb) - skb->head;
159
160 skb_reset_network_header(skb);
161
162 if (!pskb_inet_may_pull(skb)) {
163 DEV_STATS_INC(bareudp->dev, rx_length_errors);
164 DEV_STATS_INC(bareudp->dev, rx_errors);
165 goto drop;
166 }
167
168 /* Get the outer header. */
169 oiph = skb->head + nh;
170
171 if (!ipv6_mod_enabled() || family == AF_INET)
172 err = IP_ECN_decapsulate(oiph, skb);
173 else
174 err = IP6_ECN_decapsulate(oiph, skb);
175
176 if (unlikely(err)) {
177 if (log_ecn_error) {
178 if (!ipv6_mod_enabled() || family == AF_INET)
179 net_info_ratelimited("non-ECT from %pI4 "
180 "with TOS=%#x\n",
181 &((struct iphdr *)oiph)->saddr,
182 ((struct iphdr *)oiph)->tos);
183 else
184 net_info_ratelimited("non-ECT from %pI6\n",
185 &((struct ipv6hdr *)oiph)->saddr);
186 }
187 if (err > 1) {
188 DEV_STATS_INC(bareudp->dev, rx_frame_errors);
189 DEV_STATS_INC(bareudp->dev, rx_errors);
190 goto drop;
191 }
192 }
193
194 len = skb->len;
195 err = gro_cells_receive(&bareudp->gro_cells, skb);
196 if (likely(err == NET_RX_SUCCESS))
197 dev_sw_netstats_rx_add(bareudp->dev, len);
198
199 return 0;
200drop:
201 /* Consume bad packet */
202 kfree_skb(skb);
203
204 return 0;
205}
206
207static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
208{
209 return 0;
210}
211
212static int bareudp_init(struct net_device *dev)
213{
214 struct bareudp_dev *bareudp = netdev_priv(dev);
215 int err;
216
217 err = gro_cells_init(&bareudp->gro_cells, dev);
218 if (err)
219 return err;
220
221 return 0;
222}
223
224static void bareudp_uninit(struct net_device *dev)
225{
226 struct bareudp_dev *bareudp = netdev_priv(dev);
227
228 gro_cells_destroy(&bareudp->gro_cells);
229}
230
231static struct socket *bareudp_create_sock(struct net *net, __be16 port)
232{
233 struct udp_port_cfg udp_conf;
234 struct socket *sock;
235 int err;
236
237 memset(&udp_conf, 0, sizeof(udp_conf));
238
239 if (ipv6_mod_enabled())
240 udp_conf.family = AF_INET6;
241 else
242 udp_conf.family = AF_INET;
243
244 udp_conf.local_udp_port = port;
245 /* Open UDP socket */
246 err = udp_sock_create(net, &udp_conf, &sock);
247 if (err < 0)
248 return ERR_PTR(err);
249
250 udp_allow_gso(sock->sk);
251 return sock;
252}
253
254/* Create new listen socket if needed */
255static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
256{
257 struct udp_tunnel_sock_cfg tunnel_cfg;
258 struct socket *sock;
259
260 sock = bareudp_create_sock(bareudp->net, port);
261 if (IS_ERR(sock))
262 return PTR_ERR(sock);
263
264 /* Mark socket as an encapsulation socket */
265 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
266 tunnel_cfg.sk_user_data = bareudp;
267 tunnel_cfg.encap_type = 1;
268 tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
269 tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
270 tunnel_cfg.encap_destroy = NULL;
271 setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
272
273 rcu_assign_pointer(bareudp->sock, sock);
274 return 0;
275}
276
277static int bareudp_open(struct net_device *dev)
278{
279 struct bareudp_dev *bareudp = netdev_priv(dev);
280 int ret = 0;
281
282 ret = bareudp_socket_create(bareudp, bareudp->port);
283 return ret;
284}
285
286static void bareudp_sock_release(struct bareudp_dev *bareudp)
287{
288 struct socket *sock;
289
290 sock = bareudp->sock;
291 rcu_assign_pointer(bareudp->sock, NULL);
292 synchronize_net();
293 udp_tunnel_sock_release(sock);
294}
295
296static int bareudp_stop(struct net_device *dev)
297{
298 struct bareudp_dev *bareudp = netdev_priv(dev);
299
300 bareudp_sock_release(bareudp);
301 return 0;
302}
303
304static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
305 struct bareudp_dev *bareudp,
306 const struct ip_tunnel_info *info)
307{
308 bool udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
309 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
310 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
311 struct socket *sock = rcu_dereference(bareudp->sock);
312 const struct ip_tunnel_key *key = &info->key;
313 struct rtable *rt;
314 __be16 sport, df;
315 int min_headroom;
316 __u8 tos, ttl;
317 __be32 saddr;
318 int err;
319
320 if (skb_vlan_inet_prepare(skb, skb->protocol != htons(ETH_P_TEB)))
321 return -EINVAL;
322
323 if (!sock)
324 return -ESHUTDOWN;
325
326 sport = udp_flow_src_port(bareudp->net, skb,
327 bareudp->sport_min, USHRT_MAX,
328 true);
329 rt = udp_tunnel_dst_lookup(skb, dev, bareudp->net, 0, &saddr, &info->key,
330 sport, bareudp->port, key->tos,
331 use_cache ?
332 (struct dst_cache *)&info->dst_cache : NULL);
333
334 if (IS_ERR(rt))
335 return PTR_ERR(rt);
336
337 skb_tunnel_check_pmtu(skb, &rt->dst,
338 BAREUDP_IPV4_HLEN + info->options_len, false);
339
340 tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
341 ttl = key->ttl;
342 df = test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, key->tun_flags) ?
343 htons(IP_DF) : 0;
344 skb_scrub_packet(skb, xnet);
345
346 err = -ENOSPC;
347 if (!skb_pull(skb, skb_network_offset(skb)))
348 goto free_dst;
349
350 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
351 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
352
353 err = skb_cow_head(skb, min_headroom);
354 if (unlikely(err))
355 goto free_dst;
356
357 err = udp_tunnel_handle_offloads(skb, udp_sum);
358 if (err)
359 goto free_dst;
360
361 skb_set_inner_protocol(skb, bareudp->ethertype);
362 udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
363 tos, ttl, df, sport, bareudp->port,
364 !net_eq(bareudp->net, dev_net(bareudp->dev)),
365 !test_bit(IP_TUNNEL_CSUM_BIT,
366 info->key.tun_flags));
367 return 0;
368
369free_dst:
370 dst_release(&rt->dst);
371 return err;
372}
373
374static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
375 struct bareudp_dev *bareudp,
376 const struct ip_tunnel_info *info)
377{
378 bool udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
379 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
380 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
381 struct socket *sock = rcu_dereference(bareudp->sock);
382 const struct ip_tunnel_key *key = &info->key;
383 struct dst_entry *dst = NULL;
384 struct in6_addr saddr, daddr;
385 int min_headroom;
386 __u8 prio, ttl;
387 __be16 sport;
388 int err;
389
390 if (skb_vlan_inet_prepare(skb, skb->protocol != htons(ETH_P_TEB)))
391 return -EINVAL;
392
393 if (!sock)
394 return -ESHUTDOWN;
395
396 sport = udp_flow_src_port(bareudp->net, skb,
397 bareudp->sport_min, USHRT_MAX,
398 true);
399 dst = udp_tunnel6_dst_lookup(skb, dev, bareudp->net, sock, 0, &saddr,
400 key, sport, bareudp->port, key->tos,
401 use_cache ?
402 (struct dst_cache *) &info->dst_cache : NULL);
403 if (IS_ERR(dst))
404 return PTR_ERR(dst);
405
406 skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
407 false);
408
409 prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
410 ttl = key->ttl;
411
412 skb_scrub_packet(skb, xnet);
413
414 err = -ENOSPC;
415 if (!skb_pull(skb, skb_network_offset(skb)))
416 goto free_dst;
417
418 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
419 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
420
421 err = skb_cow_head(skb, min_headroom);
422 if (unlikely(err))
423 goto free_dst;
424
425 err = udp_tunnel_handle_offloads(skb, udp_sum);
426 if (err)
427 goto free_dst;
428
429 daddr = info->key.u.ipv6.dst;
430 udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
431 &saddr, &daddr, prio, ttl,
432 info->key.label, sport, bareudp->port,
433 !test_bit(IP_TUNNEL_CSUM_BIT,
434 info->key.tun_flags));
435 return 0;
436
437free_dst:
438 dst_release(dst);
439 return err;
440}
441
442static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
443{
444 if (bareudp->ethertype == proto)
445 return true;
446
447 if (!bareudp->multi_proto_mode)
448 return false;
449
450 if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
451 proto == htons(ETH_P_MPLS_MC))
452 return true;
453
454 if (bareudp->ethertype == htons(ETH_P_IP) &&
455 proto == htons(ETH_P_IPV6))
456 return true;
457
458 return false;
459}
460
461static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
462{
463 struct bareudp_dev *bareudp = netdev_priv(dev);
464 struct ip_tunnel_info *info = NULL;
465 int err;
466
467 if (!bareudp_proto_valid(bareudp, skb->protocol)) {
468 err = -EINVAL;
469 goto tx_error;
470 }
471
472 info = skb_tunnel_info(skb);
473 if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
474 err = -EINVAL;
475 goto tx_error;
476 }
477
478 rcu_read_lock();
479 if (ipv6_mod_enabled() && info->mode & IP_TUNNEL_INFO_IPV6)
480 err = bareudp6_xmit_skb(skb, dev, bareudp, info);
481 else
482 err = bareudp_xmit_skb(skb, dev, bareudp, info);
483
484 rcu_read_unlock();
485
486 if (likely(!err))
487 return NETDEV_TX_OK;
488tx_error:
489 dev_kfree_skb(skb);
490
491 if (err == -ELOOP)
492 DEV_STATS_INC(dev, collisions);
493 else if (err == -ENETUNREACH)
494 DEV_STATS_INC(dev, tx_carrier_errors);
495
496 DEV_STATS_INC(dev, tx_errors);
497 return NETDEV_TX_OK;
498}
499
500static int bareudp_fill_metadata_dst(struct net_device *dev,
501 struct sk_buff *skb)
502{
503 struct ip_tunnel_info *info = skb_tunnel_info(skb);
504 struct bareudp_dev *bareudp = netdev_priv(dev);
505 bool use_cache;
506 __be16 sport;
507
508 use_cache = ip_tunnel_dst_cache_usable(skb, info);
509 sport = udp_flow_src_port(bareudp->net, skb,
510 bareudp->sport_min, USHRT_MAX,
511 true);
512
513 if (!ipv6_mod_enabled() || ip_tunnel_info_af(info) == AF_INET) {
514 struct rtable *rt;
515 __be32 saddr;
516
517 rt = udp_tunnel_dst_lookup(skb, dev, bareudp->net, 0, &saddr,
518 &info->key, sport, bareudp->port,
519 info->key.tos,
520 use_cache ? &info->dst_cache : NULL);
521 if (IS_ERR(rt))
522 return PTR_ERR(rt);
523
524 ip_rt_put(rt);
525 info->key.u.ipv4.src = saddr;
526 } else if (ip_tunnel_info_af(info) == AF_INET6) {
527 struct dst_entry *dst;
528 struct in6_addr saddr;
529 struct socket *sock = rcu_dereference(bareudp->sock);
530
531 dst = udp_tunnel6_dst_lookup(skb, dev, bareudp->net, sock,
532 0, &saddr, &info->key,
533 sport, bareudp->port, info->key.tos,
534 use_cache ? &info->dst_cache : NULL);
535 if (IS_ERR(dst))
536 return PTR_ERR(dst);
537
538 dst_release(dst);
539 info->key.u.ipv6.src = saddr;
540 } else {
541 return -EINVAL;
542 }
543
544 info->key.tp_src = sport;
545 info->key.tp_dst = bareudp->port;
546 return 0;
547}
548
549static const struct net_device_ops bareudp_netdev_ops = {
550 .ndo_init = bareudp_init,
551 .ndo_uninit = bareudp_uninit,
552 .ndo_open = bareudp_open,
553 .ndo_stop = bareudp_stop,
554 .ndo_start_xmit = bareudp_xmit,
555 .ndo_fill_metadata_dst = bareudp_fill_metadata_dst,
556};
557
558static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
559 [IFLA_BAREUDP_PORT] = { .type = NLA_U16 },
560 [IFLA_BAREUDP_ETHERTYPE] = { .type = NLA_U16 },
561 [IFLA_BAREUDP_SRCPORT_MIN] = { .type = NLA_U16 },
562 [IFLA_BAREUDP_MULTIPROTO_MODE] = { .type = NLA_FLAG },
563};
564
565/* Info for udev, that this is a virtual tunnel endpoint */
566static const struct device_type bareudp_type = {
567 .name = "bareudp",
568};
569
570/* Initialize the device structure. */
571static void bareudp_setup(struct net_device *dev)
572{
573 dev->netdev_ops = &bareudp_netdev_ops;
574 dev->needs_free_netdev = true;
575 SET_NETDEV_DEVTYPE(dev, &bareudp_type);
576 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
577 dev->features |= NETIF_F_RXCSUM;
578 dev->features |= NETIF_F_GSO_SOFTWARE;
579 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
580 dev->hw_features |= NETIF_F_RXCSUM;
581 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
582 dev->hard_header_len = 0;
583 dev->addr_len = 0;
584 dev->mtu = ETH_DATA_LEN;
585 dev->min_mtu = IPV4_MIN_MTU;
586 dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
587 dev->type = ARPHRD_NONE;
588 netif_keep_dst(dev);
589 dev->priv_flags |= IFF_NO_QUEUE;
590 dev->lltx = true;
591 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
592 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
593}
594
595static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
596 struct netlink_ext_ack *extack)
597{
598 if (!data) {
599 NL_SET_ERR_MSG(extack,
600 "Not enough attributes provided to perform the operation");
601 return -EINVAL;
602 }
603 return 0;
604}
605
606static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
607 struct netlink_ext_ack *extack)
608{
609 memset(conf, 0, sizeof(*conf));
610
611 if (!data[IFLA_BAREUDP_PORT]) {
612 NL_SET_ERR_MSG(extack, "port not specified");
613 return -EINVAL;
614 }
615 if (!data[IFLA_BAREUDP_ETHERTYPE]) {
616 NL_SET_ERR_MSG(extack, "ethertype not specified");
617 return -EINVAL;
618 }
619
620 conf->port = nla_get_u16(data[IFLA_BAREUDP_PORT]);
621 conf->ethertype = nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
622
623 if (data[IFLA_BAREUDP_SRCPORT_MIN])
624 conf->sport_min = nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
625
626 if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
627 conf->multi_proto_mode = true;
628
629 return 0;
630}
631
632static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
633 const struct bareudp_conf *conf)
634{
635 struct bareudp_dev *bareudp, *t = NULL;
636
637 list_for_each_entry(bareudp, &bn->bareudp_list, next) {
638 if (conf->port == bareudp->port)
639 t = bareudp;
640 }
641 return t;
642}
643
644static int bareudp_configure(struct net *net, struct net_device *dev,
645 struct bareudp_conf *conf,
646 struct netlink_ext_ack *extack)
647{
648 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
649 struct bareudp_dev *t, *bareudp = netdev_priv(dev);
650 int err;
651
652 bareudp->net = net;
653 bareudp->dev = dev;
654 t = bareudp_find_dev(bn, conf);
655 if (t) {
656 NL_SET_ERR_MSG(extack, "Another bareudp device using the same port already exists");
657 return -EBUSY;
658 }
659
660 if (conf->multi_proto_mode &&
661 (conf->ethertype != htons(ETH_P_MPLS_UC) &&
662 conf->ethertype != htons(ETH_P_IP))) {
663 NL_SET_ERR_MSG(extack, "Cannot set multiproto mode for this ethertype (only IPv4 and unicast MPLS are supported)");
664 return -EINVAL;
665 }
666
667 bareudp->port = conf->port;
668 bareudp->ethertype = conf->ethertype;
669 bareudp->sport_min = conf->sport_min;
670 bareudp->multi_proto_mode = conf->multi_proto_mode;
671
672 err = register_netdevice(dev);
673 if (err)
674 return err;
675
676 list_add(&bareudp->next, &bn->bareudp_list);
677 return 0;
678}
679
680static int bareudp_link_config(struct net_device *dev,
681 struct nlattr *tb[])
682{
683 int err;
684
685 if (tb[IFLA_MTU]) {
686 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
687 if (err)
688 return err;
689 }
690 return 0;
691}
692
693static void bareudp_dellink(struct net_device *dev, struct list_head *head)
694{
695 struct bareudp_dev *bareudp = netdev_priv(dev);
696
697 list_del(&bareudp->next);
698 unregister_netdevice_queue(dev, head);
699}
700
701static int bareudp_newlink(struct net *net, struct net_device *dev,
702 struct nlattr *tb[], struct nlattr *data[],
703 struct netlink_ext_ack *extack)
704{
705 struct bareudp_conf conf;
706 int err;
707
708 err = bareudp2info(data, &conf, extack);
709 if (err)
710 return err;
711
712 err = bareudp_configure(net, dev, &conf, extack);
713 if (err)
714 return err;
715
716 err = bareudp_link_config(dev, tb);
717 if (err)
718 goto err_unconfig;
719
720 return 0;
721
722err_unconfig:
723 bareudp_dellink(dev, NULL);
724 return err;
725}
726
727static size_t bareudp_get_size(const struct net_device *dev)
728{
729 return nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_PORT */
730 nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_ETHERTYPE */
731 nla_total_size(sizeof(__u16)) + /* IFLA_BAREUDP_SRCPORT_MIN */
732 nla_total_size(0) + /* IFLA_BAREUDP_MULTIPROTO_MODE */
733 0;
734}
735
736static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
737{
738 struct bareudp_dev *bareudp = netdev_priv(dev);
739
740 if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
741 goto nla_put_failure;
742 if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
743 goto nla_put_failure;
744 if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
745 goto nla_put_failure;
746 if (bareudp->multi_proto_mode &&
747 nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
748 goto nla_put_failure;
749
750 return 0;
751
752nla_put_failure:
753 return -EMSGSIZE;
754}
755
756static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
757 .kind = "bareudp",
758 .maxtype = IFLA_BAREUDP_MAX,
759 .policy = bareudp_policy,
760 .priv_size = sizeof(struct bareudp_dev),
761 .setup = bareudp_setup,
762 .validate = bareudp_validate,
763 .newlink = bareudp_newlink,
764 .dellink = bareudp_dellink,
765 .get_size = bareudp_get_size,
766 .fill_info = bareudp_fill_info,
767};
768
769static __net_init int bareudp_init_net(struct net *net)
770{
771 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
772
773 INIT_LIST_HEAD(&bn->bareudp_list);
774 return 0;
775}
776
777static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
778{
779 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
780 struct bareudp_dev *bareudp, *next;
781
782 list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
783 unregister_netdevice_queue(bareudp->dev, head);
784}
785
786static void __net_exit bareudp_exit_batch_rtnl(struct list_head *net_list,
787 struct list_head *dev_kill_list)
788{
789 struct net *net;
790
791 list_for_each_entry(net, net_list, exit_list)
792 bareudp_destroy_tunnels(net, dev_kill_list);
793}
794
795static struct pernet_operations bareudp_net_ops = {
796 .init = bareudp_init_net,
797 .exit_batch_rtnl = bareudp_exit_batch_rtnl,
798 .id = &bareudp_net_id,
799 .size = sizeof(struct bareudp_net),
800};
801
802static int __init bareudp_init_module(void)
803{
804 int rc;
805
806 rc = register_pernet_subsys(&bareudp_net_ops);
807 if (rc)
808 goto out1;
809
810 rc = rtnl_link_register(&bareudp_link_ops);
811 if (rc)
812 goto out2;
813
814 return 0;
815out2:
816 unregister_pernet_subsys(&bareudp_net_ops);
817out1:
818 return rc;
819}
820late_initcall(bareudp_init_module);
821
822static void __exit bareudp_cleanup_module(void)
823{
824 rtnl_link_unregister(&bareudp_link_ops);
825 unregister_pernet_subsys(&bareudp_net_ops);
826}
827module_exit(bareudp_cleanup_module);
828
829MODULE_ALIAS_RTNL_LINK("bareudp");
830MODULE_LICENSE("GPL");
831MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
832MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");
1// SPDX-License-Identifier: GPL-2.0
2/* Bareudp: UDP tunnel encasulation for different Payload types like
3 * MPLS, NSH, IP, etc.
4 * Copyright (c) 2019 Nokia, Inc.
5 * Authors: Martin Varghese, <martin.varghese@nokia.com>
6 */
7
8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10#include <linux/kernel.h>
11#include <linux/module.h>
12#include <linux/etherdevice.h>
13#include <linux/hash.h>
14#include <net/dst_metadata.h>
15#include <net/gro_cells.h>
16#include <net/rtnetlink.h>
17#include <net/protocol.h>
18#include <net/ip6_tunnel.h>
19#include <net/ip_tunnels.h>
20#include <net/udp_tunnel.h>
21#include <net/bareudp.h>
22
23#define BAREUDP_BASE_HLEN sizeof(struct udphdr)
24#define BAREUDP_IPV4_HLEN (sizeof(struct iphdr) + \
25 sizeof(struct udphdr))
26#define BAREUDP_IPV6_HLEN (sizeof(struct ipv6hdr) + \
27 sizeof(struct udphdr))
28
29static bool log_ecn_error = true;
30module_param(log_ecn_error, bool, 0644);
31MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
32
33/* per-network namespace private data for this module */
34
35static unsigned int bareudp_net_id;
36
37struct bareudp_net {
38 struct list_head bareudp_list;
39};
40
41/* Pseudo network device */
42struct bareudp_dev {
43 struct net *net; /* netns for packet i/o */
44 struct net_device *dev; /* netdev for bareudp tunnel */
45 __be16 ethertype;
46 __be16 port;
47 u16 sport_min;
48 bool multi_proto_mode;
49 struct socket __rcu *sock;
50 struct list_head next; /* bareudp node on namespace list */
51 struct gro_cells gro_cells;
52};
53
54static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
55{
56 struct metadata_dst *tun_dst = NULL;
57 struct pcpu_sw_netstats *stats;
58 struct bareudp_dev *bareudp;
59 unsigned short family;
60 unsigned int len;
61 __be16 proto;
62 void *oiph;
63 int err;
64
65 bareudp = rcu_dereference_sk_user_data(sk);
66 if (!bareudp)
67 goto drop;
68
69 if (skb->protocol == htons(ETH_P_IP))
70 family = AF_INET;
71 else
72 family = AF_INET6;
73
74 if (bareudp->ethertype == htons(ETH_P_IP)) {
75 struct iphdr *iphdr;
76
77 iphdr = (struct iphdr *)(skb->data + BAREUDP_BASE_HLEN);
78 if (iphdr->version == 4) {
79 proto = bareudp->ethertype;
80 } else if (bareudp->multi_proto_mode && (iphdr->version == 6)) {
81 proto = htons(ETH_P_IPV6);
82 } else {
83 bareudp->dev->stats.rx_dropped++;
84 goto drop;
85 }
86 } else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
87 struct iphdr *tunnel_hdr;
88
89 tunnel_hdr = (struct iphdr *)skb_network_header(skb);
90 if (tunnel_hdr->version == 4) {
91 if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
92 proto = bareudp->ethertype;
93 } else if (bareudp->multi_proto_mode &&
94 ipv4_is_multicast(tunnel_hdr->daddr)) {
95 proto = htons(ETH_P_MPLS_MC);
96 } else {
97 bareudp->dev->stats.rx_dropped++;
98 goto drop;
99 }
100 } else {
101 int addr_type;
102 struct ipv6hdr *tunnel_hdr_v6;
103
104 tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
105 addr_type =
106 ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
107 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
108 proto = bareudp->ethertype;
109 } else if (bareudp->multi_proto_mode &&
110 (addr_type & IPV6_ADDR_MULTICAST)) {
111 proto = htons(ETH_P_MPLS_MC);
112 } else {
113 bareudp->dev->stats.rx_dropped++;
114 goto drop;
115 }
116 }
117 } else {
118 proto = bareudp->ethertype;
119 }
120
121 if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
122 proto,
123 !net_eq(bareudp->net,
124 dev_net(bareudp->dev)))) {
125 bareudp->dev->stats.rx_dropped++;
126 goto drop;
127 }
128 tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
129 if (!tun_dst) {
130 bareudp->dev->stats.rx_dropped++;
131 goto drop;
132 }
133 skb_dst_set(skb, &tun_dst->dst);
134 skb->dev = bareudp->dev;
135 oiph = skb_network_header(skb);
136 skb_reset_network_header(skb);
137
138 if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
139 err = IP_ECN_decapsulate(oiph, skb);
140 else
141 err = IP6_ECN_decapsulate(oiph, skb);
142
143 if (unlikely(err)) {
144 if (log_ecn_error) {
145 if (!IS_ENABLED(CONFIG_IPV6) || family == AF_INET)
146 net_info_ratelimited("non-ECT from %pI4 "
147 "with TOS=%#x\n",
148 &((struct iphdr *)oiph)->saddr,
149 ((struct iphdr *)oiph)->tos);
150 else
151 net_info_ratelimited("non-ECT from %pI6\n",
152 &((struct ipv6hdr *)oiph)->saddr);
153 }
154 if (err > 1) {
155 ++bareudp->dev->stats.rx_frame_errors;
156 ++bareudp->dev->stats.rx_errors;
157 goto drop;
158 }
159 }
160
161 len = skb->len;
162 err = gro_cells_receive(&bareudp->gro_cells, skb);
163 if (likely(err == NET_RX_SUCCESS)) {
164 stats = this_cpu_ptr(bareudp->dev->tstats);
165 u64_stats_update_begin(&stats->syncp);
166 stats->rx_packets++;
167 stats->rx_bytes += len;
168 u64_stats_update_end(&stats->syncp);
169 }
170 return 0;
171drop:
172 /* Consume bad packet */
173 kfree_skb(skb);
174
175 return 0;
176}
177
178static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
179{
180 return 0;
181}
182
183static int bareudp_init(struct net_device *dev)
184{
185 struct bareudp_dev *bareudp = netdev_priv(dev);
186 int err;
187
188 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
189 if (!dev->tstats)
190 return -ENOMEM;
191
192 err = gro_cells_init(&bareudp->gro_cells, dev);
193 if (err) {
194 free_percpu(dev->tstats);
195 return err;
196 }
197 return 0;
198}
199
200static void bareudp_uninit(struct net_device *dev)
201{
202 struct bareudp_dev *bareudp = netdev_priv(dev);
203
204 gro_cells_destroy(&bareudp->gro_cells);
205 free_percpu(dev->tstats);
206}
207
208static struct socket *bareudp_create_sock(struct net *net, __be16 port)
209{
210 struct udp_port_cfg udp_conf;
211 struct socket *sock;
212 int err;
213
214 memset(&udp_conf, 0, sizeof(udp_conf));
215#if IS_ENABLED(CONFIG_IPV6)
216 udp_conf.family = AF_INET6;
217#else
218 udp_conf.family = AF_INET;
219#endif
220 udp_conf.local_udp_port = port;
221 /* Open UDP socket */
222 err = udp_sock_create(net, &udp_conf, &sock);
223 if (err < 0)
224 return ERR_PTR(err);
225
226 return sock;
227}
228
229/* Create new listen socket if needed */
230static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
231{
232 struct udp_tunnel_sock_cfg tunnel_cfg;
233 struct socket *sock;
234
235 sock = bareudp_create_sock(bareudp->net, port);
236 if (IS_ERR(sock))
237 return PTR_ERR(sock);
238
239 /* Mark socket as an encapsulation socket */
240 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
241 tunnel_cfg.sk_user_data = bareudp;
242 tunnel_cfg.encap_type = 1;
243 tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
244 tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
245 tunnel_cfg.encap_destroy = NULL;
246 setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
247
248 /* As the setup_udp_tunnel_sock does not call udp_encap_enable if the
249 * socket type is v6 an explicit call to udp_encap_enable is needed.
250 */
251 if (sock->sk->sk_family == AF_INET6)
252 udp_encap_enable();
253
254 rcu_assign_pointer(bareudp->sock, sock);
255 return 0;
256}
257
258static int bareudp_open(struct net_device *dev)
259{
260 struct bareudp_dev *bareudp = netdev_priv(dev);
261 int ret = 0;
262
263 ret = bareudp_socket_create(bareudp, bareudp->port);
264 return ret;
265}
266
267static void bareudp_sock_release(struct bareudp_dev *bareudp)
268{
269 struct socket *sock;
270
271 sock = bareudp->sock;
272 rcu_assign_pointer(bareudp->sock, NULL);
273 synchronize_net();
274 udp_tunnel_sock_release(sock);
275}
276
277static int bareudp_stop(struct net_device *dev)
278{
279 struct bareudp_dev *bareudp = netdev_priv(dev);
280
281 bareudp_sock_release(bareudp);
282 return 0;
283}
284
285static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
286 struct bareudp_dev *bareudp,
287 const struct ip_tunnel_info *info)
288{
289 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
290 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
291 struct socket *sock = rcu_dereference(bareudp->sock);
292 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
293 const struct ip_tunnel_key *key = &info->key;
294 struct rtable *rt;
295 __be16 sport, df;
296 int min_headroom;
297 __u8 tos, ttl;
298 __be32 saddr;
299 int err;
300
301 if (!sock)
302 return -ESHUTDOWN;
303
304 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
305 IPPROTO_UDP, use_cache);
306
307 if (IS_ERR(rt))
308 return PTR_ERR(rt);
309
310 skb_tunnel_check_pmtu(skb, &rt->dst,
311 BAREUDP_IPV4_HLEN + info->options_len, false);
312
313 sport = udp_flow_src_port(bareudp->net, skb,
314 bareudp->sport_min, USHRT_MAX,
315 true);
316 tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
317 ttl = key->ttl;
318 df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
319 skb_scrub_packet(skb, xnet);
320
321 err = -ENOSPC;
322 if (!skb_pull(skb, skb_network_offset(skb)))
323 goto free_dst;
324
325 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
326 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
327
328 err = skb_cow_head(skb, min_headroom);
329 if (unlikely(err))
330 goto free_dst;
331
332 err = udp_tunnel_handle_offloads(skb, udp_sum);
333 if (err)
334 goto free_dst;
335
336 skb_set_inner_protocol(skb, bareudp->ethertype);
337 udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
338 tos, ttl, df, sport, bareudp->port,
339 !net_eq(bareudp->net, dev_net(bareudp->dev)),
340 !(info->key.tun_flags & TUNNEL_CSUM));
341 return 0;
342
343free_dst:
344 dst_release(&rt->dst);
345 return err;
346}
347
348static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
349 struct bareudp_dev *bareudp,
350 const struct ip_tunnel_info *info)
351{
352 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
353 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
354 struct socket *sock = rcu_dereference(bareudp->sock);
355 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
356 const struct ip_tunnel_key *key = &info->key;
357 struct dst_entry *dst = NULL;
358 struct in6_addr saddr, daddr;
359 int min_headroom;
360 __u8 prio, ttl;
361 __be16 sport;
362 int err;
363
364 if (!sock)
365 return -ESHUTDOWN;
366
367 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
368 IPPROTO_UDP, use_cache);
369 if (IS_ERR(dst))
370 return PTR_ERR(dst);
371
372 skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
373 false);
374
375 sport = udp_flow_src_port(bareudp->net, skb,
376 bareudp->sport_min, USHRT_MAX,
377 true);
378 prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
379 ttl = key->ttl;
380
381 skb_scrub_packet(skb, xnet);
382
383 err = -ENOSPC;
384 if (!skb_pull(skb, skb_network_offset(skb)))
385 goto free_dst;
386
387 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
388 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
389
390 err = skb_cow_head(skb, min_headroom);
391 if (unlikely(err))
392 goto free_dst;
393
394 err = udp_tunnel_handle_offloads(skb, udp_sum);
395 if (err)
396 goto free_dst;
397
398 daddr = info->key.u.ipv6.dst;
399 udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
400 &saddr, &daddr, prio, ttl,
401 info->key.label, sport, bareudp->port,
402 !(info->key.tun_flags & TUNNEL_CSUM));
403 return 0;
404
405free_dst:
406 dst_release(dst);
407 return err;
408}
409
410static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
411{
412 if (bareudp->ethertype == proto)
413 return true;
414
415 if (!bareudp->multi_proto_mode)
416 return false;
417
418 if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
419 proto == htons(ETH_P_MPLS_MC))
420 return true;
421
422 if (bareudp->ethertype == htons(ETH_P_IP) &&
423 proto == htons(ETH_P_IPV6))
424 return true;
425
426 return false;
427}
428
429static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
430{
431 struct bareudp_dev *bareudp = netdev_priv(dev);
432 struct ip_tunnel_info *info = NULL;
433 int err;
434
435 if (!bareudp_proto_valid(bareudp, skb->protocol)) {
436 err = -EINVAL;
437 goto tx_error;
438 }
439
440 info = skb_tunnel_info(skb);
441 if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
442 err = -EINVAL;
443 goto tx_error;
444 }
445
446 rcu_read_lock();
447 if (IS_ENABLED(CONFIG_IPV6) && info->mode & IP_TUNNEL_INFO_IPV6)
448 err = bareudp6_xmit_skb(skb, dev, bareudp, info);
449 else
450 err = bareudp_xmit_skb(skb, dev, bareudp, info);
451
452 rcu_read_unlock();
453
454 if (likely(!err))
455 return NETDEV_TX_OK;
456tx_error:
457 dev_kfree_skb(skb);
458
459 if (err == -ELOOP)
460 dev->stats.collisions++;
461 else if (err == -ENETUNREACH)
462 dev->stats.tx_carrier_errors++;
463
464 dev->stats.tx_errors++;
465 return NETDEV_TX_OK;
466}
467
468static int bareudp_fill_metadata_dst(struct net_device *dev,
469 struct sk_buff *skb)
470{
471 struct ip_tunnel_info *info = skb_tunnel_info(skb);
472 struct bareudp_dev *bareudp = netdev_priv(dev);
473 bool use_cache;
474
475 use_cache = ip_tunnel_dst_cache_usable(skb, info);
476
477 if (!IS_ENABLED(CONFIG_IPV6) || ip_tunnel_info_af(info) == AF_INET) {
478 struct rtable *rt;
479 __be32 saddr;
480
481 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
482 info, IPPROTO_UDP, use_cache);
483 if (IS_ERR(rt))
484 return PTR_ERR(rt);
485
486 ip_rt_put(rt);
487 info->key.u.ipv4.src = saddr;
488 } else if (ip_tunnel_info_af(info) == AF_INET6) {
489 struct dst_entry *dst;
490 struct in6_addr saddr;
491 struct socket *sock = rcu_dereference(bareudp->sock);
492
493 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
494 &saddr, info, IPPROTO_UDP,
495 use_cache);
496 if (IS_ERR(dst))
497 return PTR_ERR(dst);
498
499 dst_release(dst);
500 info->key.u.ipv6.src = saddr;
501 } else {
502 return -EINVAL;
503 }
504
505 info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
506 bareudp->sport_min,
507 USHRT_MAX, true);
508 info->key.tp_dst = bareudp->port;
509 return 0;
510}
511
512static const struct net_device_ops bareudp_netdev_ops = {
513 .ndo_init = bareudp_init,
514 .ndo_uninit = bareudp_uninit,
515 .ndo_open = bareudp_open,
516 .ndo_stop = bareudp_stop,
517 .ndo_start_xmit = bareudp_xmit,
518 .ndo_get_stats64 = ip_tunnel_get_stats64,
519 .ndo_fill_metadata_dst = bareudp_fill_metadata_dst,
520};
521
522static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
523 [IFLA_BAREUDP_PORT] = { .type = NLA_U16 },
524 [IFLA_BAREUDP_ETHERTYPE] = { .type = NLA_U16 },
525 [IFLA_BAREUDP_SRCPORT_MIN] = { .type = NLA_U16 },
526 [IFLA_BAREUDP_MULTIPROTO_MODE] = { .type = NLA_FLAG },
527};
528
529/* Info for udev, that this is a virtual tunnel endpoint */
530static struct device_type bareudp_type = {
531 .name = "bareudp",
532};
533
534/* Initialize the device structure. */
535static void bareudp_setup(struct net_device *dev)
536{
537 dev->netdev_ops = &bareudp_netdev_ops;
538 dev->needs_free_netdev = true;
539 SET_NETDEV_DEVTYPE(dev, &bareudp_type);
540 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
541 dev->features |= NETIF_F_RXCSUM;
542 dev->features |= NETIF_F_GSO_SOFTWARE;
543 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
544 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
545 dev->hard_header_len = 0;
546 dev->addr_len = 0;
547 dev->mtu = ETH_DATA_LEN;
548 dev->min_mtu = IPV4_MIN_MTU;
549 dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
550 dev->type = ARPHRD_NONE;
551 netif_keep_dst(dev);
552 dev->priv_flags |= IFF_NO_QUEUE;
553 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
554}
555
556static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
557 struct netlink_ext_ack *extack)
558{
559 if (!data) {
560 NL_SET_ERR_MSG(extack,
561 "Not enough attributes provided to perform the operation");
562 return -EINVAL;
563 }
564 return 0;
565}
566
567static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
568 struct netlink_ext_ack *extack)
569{
570 memset(conf, 0, sizeof(*conf));
571
572 if (!data[IFLA_BAREUDP_PORT]) {
573 NL_SET_ERR_MSG(extack, "port not specified");
574 return -EINVAL;
575 }
576 if (!data[IFLA_BAREUDP_ETHERTYPE]) {
577 NL_SET_ERR_MSG(extack, "ethertype not specified");
578 return -EINVAL;
579 }
580
581 if (data[IFLA_BAREUDP_PORT])
582 conf->port = nla_get_u16(data[IFLA_BAREUDP_PORT]);
583
584 if (data[IFLA_BAREUDP_ETHERTYPE])
585 conf->ethertype = nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
586
587 if (data[IFLA_BAREUDP_SRCPORT_MIN])
588 conf->sport_min = nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
589
590 if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
591 conf->multi_proto_mode = true;
592
593 return 0;
594}
595
596static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
597 const struct bareudp_conf *conf)
598{
599 struct bareudp_dev *bareudp, *t = NULL;
600
601 list_for_each_entry(bareudp, &bn->bareudp_list, next) {
602 if (conf->port == bareudp->port)
603 t = bareudp;
604 }
605 return t;
606}
607
608static int bareudp_configure(struct net *net, struct net_device *dev,
609 struct bareudp_conf *conf)
610{
611 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
612 struct bareudp_dev *t, *bareudp = netdev_priv(dev);
613 int err;
614
615 bareudp->net = net;
616 bareudp->dev = dev;
617 t = bareudp_find_dev(bn, conf);
618 if (t)
619 return -EBUSY;
620
621 if (conf->multi_proto_mode &&
622 (conf->ethertype != htons(ETH_P_MPLS_UC) &&
623 conf->ethertype != htons(ETH_P_IP)))
624 return -EINVAL;
625
626 bareudp->port = conf->port;
627 bareudp->ethertype = conf->ethertype;
628 bareudp->sport_min = conf->sport_min;
629 bareudp->multi_proto_mode = conf->multi_proto_mode;
630
631 err = register_netdevice(dev);
632 if (err)
633 return err;
634
635 list_add(&bareudp->next, &bn->bareudp_list);
636 return 0;
637}
638
639static int bareudp_link_config(struct net_device *dev,
640 struct nlattr *tb[])
641{
642 int err;
643
644 if (tb[IFLA_MTU]) {
645 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
646 if (err)
647 return err;
648 }
649 return 0;
650}
651
652static int bareudp_newlink(struct net *net, struct net_device *dev,
653 struct nlattr *tb[], struct nlattr *data[],
654 struct netlink_ext_ack *extack)
655{
656 struct bareudp_conf conf;
657 int err;
658
659 err = bareudp2info(data, &conf, extack);
660 if (err)
661 return err;
662
663 err = bareudp_configure(net, dev, &conf);
664 if (err)
665 return err;
666
667 err = bareudp_link_config(dev, tb);
668 if (err)
669 return err;
670
671 return 0;
672}
673
674static void bareudp_dellink(struct net_device *dev, struct list_head *head)
675{
676 struct bareudp_dev *bareudp = netdev_priv(dev);
677
678 list_del(&bareudp->next);
679 unregister_netdevice_queue(dev, head);
680}
681
682static size_t bareudp_get_size(const struct net_device *dev)
683{
684 return nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_PORT */
685 nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_ETHERTYPE */
686 nla_total_size(sizeof(__u16)) + /* IFLA_BAREUDP_SRCPORT_MIN */
687 nla_total_size(0) + /* IFLA_BAREUDP_MULTIPROTO_MODE */
688 0;
689}
690
691static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
692{
693 struct bareudp_dev *bareudp = netdev_priv(dev);
694
695 if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
696 goto nla_put_failure;
697 if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
698 goto nla_put_failure;
699 if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
700 goto nla_put_failure;
701 if (bareudp->multi_proto_mode &&
702 nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
703 goto nla_put_failure;
704
705 return 0;
706
707nla_put_failure:
708 return -EMSGSIZE;
709}
710
711static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
712 .kind = "bareudp",
713 .maxtype = IFLA_BAREUDP_MAX,
714 .policy = bareudp_policy,
715 .priv_size = sizeof(struct bareudp_dev),
716 .setup = bareudp_setup,
717 .validate = bareudp_validate,
718 .newlink = bareudp_newlink,
719 .dellink = bareudp_dellink,
720 .get_size = bareudp_get_size,
721 .fill_info = bareudp_fill_info,
722};
723
724struct net_device *bareudp_dev_create(struct net *net, const char *name,
725 u8 name_assign_type,
726 struct bareudp_conf *conf)
727{
728 struct nlattr *tb[IFLA_MAX + 1];
729 struct net_device *dev;
730 LIST_HEAD(list_kill);
731 int err;
732
733 memset(tb, 0, sizeof(tb));
734 dev = rtnl_create_link(net, name, name_assign_type,
735 &bareudp_link_ops, tb, NULL);
736 if (IS_ERR(dev))
737 return dev;
738
739 err = bareudp_configure(net, dev, conf);
740 if (err) {
741 free_netdev(dev);
742 return ERR_PTR(err);
743 }
744 err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
745 if (err)
746 goto err;
747
748 err = rtnl_configure_link(dev, NULL);
749 if (err < 0)
750 goto err;
751
752 return dev;
753err:
754 bareudp_dellink(dev, &list_kill);
755 unregister_netdevice_many(&list_kill);
756 return ERR_PTR(err);
757}
758EXPORT_SYMBOL_GPL(bareudp_dev_create);
759
760static __net_init int bareudp_init_net(struct net *net)
761{
762 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
763
764 INIT_LIST_HEAD(&bn->bareudp_list);
765 return 0;
766}
767
768static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
769{
770 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
771 struct bareudp_dev *bareudp, *next;
772
773 list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
774 unregister_netdevice_queue(bareudp->dev, head);
775}
776
777static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
778{
779 struct net *net;
780 LIST_HEAD(list);
781
782 rtnl_lock();
783 list_for_each_entry(net, net_list, exit_list)
784 bareudp_destroy_tunnels(net, &list);
785
786 /* unregister the devices gathered above */
787 unregister_netdevice_many(&list);
788 rtnl_unlock();
789}
790
791static struct pernet_operations bareudp_net_ops = {
792 .init = bareudp_init_net,
793 .exit_batch = bareudp_exit_batch_net,
794 .id = &bareudp_net_id,
795 .size = sizeof(struct bareudp_net),
796};
797
798static int __init bareudp_init_module(void)
799{
800 int rc;
801
802 rc = register_pernet_subsys(&bareudp_net_ops);
803 if (rc)
804 goto out1;
805
806 rc = rtnl_link_register(&bareudp_link_ops);
807 if (rc)
808 goto out2;
809
810 return 0;
811out2:
812 unregister_pernet_subsys(&bareudp_net_ops);
813out1:
814 return rc;
815}
816late_initcall(bareudp_init_module);
817
818static void __exit bareudp_cleanup_module(void)
819{
820 rtnl_link_unregister(&bareudp_link_ops);
821 unregister_pernet_subsys(&bareudp_net_ops);
822}
823module_exit(bareudp_cleanup_module);
824
825MODULE_ALIAS_RTNL_LINK("bareudp");
826MODULE_LICENSE("GPL");
827MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
828MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");