1 // SPDX-License-Identifier: GPL-2.0
2 /* Bareudp: UDP tunnel encasulation for different Payload types like
4 * Copyright (c) 2019 Nokia, Inc.
5 * Authors: Martin Varghese, <martin.varghese@nokia.com>
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
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>
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))
29 static bool log_ecn_error = true;
30 module_param(log_ecn_error, bool, 0644);
31 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
33 /* per-network namespace private data for this module */
35 static unsigned int bareudp_net_id;
38 struct list_head bareudp_list;
41 /* Pseudo network device */
43 struct net *net; /* netns for packet i/o */
44 struct net_device *dev; /* netdev for bareudp tunnel */
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;
54 static int bareudp_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
56 struct metadata_dst *tun_dst = NULL;
57 struct bareudp_dev *bareudp;
58 unsigned short family;
64 bareudp = rcu_dereference_sk_user_data(sk);
68 if (skb->protocol == htons(ETH_P_IP))
73 if (bareudp->ethertype == htons(ETH_P_IP)) {
76 if (skb_copy_bits(skb, BAREUDP_BASE_HLEN, &ipversion,
78 bareudp->dev->stats.rx_dropped++;
84 proto = htons(ETH_P_IP);
85 } else if (ipversion == 6 && bareudp->multi_proto_mode) {
86 proto = htons(ETH_P_IPV6);
88 bareudp->dev->stats.rx_dropped++;
91 } else if (bareudp->ethertype == htons(ETH_P_MPLS_UC)) {
92 struct iphdr *tunnel_hdr;
94 tunnel_hdr = (struct iphdr *)skb_network_header(skb);
95 if (tunnel_hdr->version == 4) {
96 if (!ipv4_is_multicast(tunnel_hdr->daddr)) {
97 proto = bareudp->ethertype;
98 } else if (bareudp->multi_proto_mode &&
99 ipv4_is_multicast(tunnel_hdr->daddr)) {
100 proto = htons(ETH_P_MPLS_MC);
102 bareudp->dev->stats.rx_dropped++;
107 struct ipv6hdr *tunnel_hdr_v6;
109 tunnel_hdr_v6 = (struct ipv6hdr *)skb_network_header(skb);
111 ipv6_addr_type((struct in6_addr *)&tunnel_hdr_v6->daddr);
112 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
113 proto = bareudp->ethertype;
114 } else if (bareudp->multi_proto_mode &&
115 (addr_type & IPV6_ADDR_MULTICAST)) {
116 proto = htons(ETH_P_MPLS_MC);
118 bareudp->dev->stats.rx_dropped++;
123 proto = bareudp->ethertype;
126 if (iptunnel_pull_header(skb, BAREUDP_BASE_HLEN,
128 !net_eq(bareudp->net,
129 dev_net(bareudp->dev)))) {
130 bareudp->dev->stats.rx_dropped++;
133 tun_dst = udp_tun_rx_dst(skb, family, TUNNEL_KEY, 0, 0);
135 bareudp->dev->stats.rx_dropped++;
138 skb_dst_set(skb, &tun_dst->dst);
139 skb->dev = bareudp->dev;
140 oiph = skb_network_header(skb);
141 skb_reset_network_header(skb);
143 if (!ipv6_mod_enabled() || family == AF_INET)
144 err = IP_ECN_decapsulate(oiph, skb);
146 err = IP6_ECN_decapsulate(oiph, skb);
150 if (!ipv6_mod_enabled() || family == AF_INET)
151 net_info_ratelimited("non-ECT from %pI4 "
153 &((struct iphdr *)oiph)->saddr,
154 ((struct iphdr *)oiph)->tos);
156 net_info_ratelimited("non-ECT from %pI6\n",
157 &((struct ipv6hdr *)oiph)->saddr);
160 ++bareudp->dev->stats.rx_frame_errors;
161 ++bareudp->dev->stats.rx_errors;
167 err = gro_cells_receive(&bareudp->gro_cells, skb);
168 if (likely(err == NET_RX_SUCCESS))
169 dev_sw_netstats_rx_add(bareudp->dev, len);
173 /* Consume bad packet */
179 static int bareudp_err_lookup(struct sock *sk, struct sk_buff *skb)
184 static int bareudp_init(struct net_device *dev)
186 struct bareudp_dev *bareudp = netdev_priv(dev);
189 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
193 err = gro_cells_init(&bareudp->gro_cells, dev);
195 free_percpu(dev->tstats);
201 static void bareudp_uninit(struct net_device *dev)
203 struct bareudp_dev *bareudp = netdev_priv(dev);
205 gro_cells_destroy(&bareudp->gro_cells);
206 free_percpu(dev->tstats);
209 static struct socket *bareudp_create_sock(struct net *net, __be16 port)
211 struct udp_port_cfg udp_conf;
215 memset(&udp_conf, 0, sizeof(udp_conf));
217 if (ipv6_mod_enabled())
218 udp_conf.family = AF_INET6;
220 udp_conf.family = AF_INET;
222 udp_conf.local_udp_port = port;
223 /* Open UDP socket */
224 err = udp_sock_create(net, &udp_conf, &sock);
231 /* Create new listen socket if needed */
232 static int bareudp_socket_create(struct bareudp_dev *bareudp, __be16 port)
234 struct udp_tunnel_sock_cfg tunnel_cfg;
237 sock = bareudp_create_sock(bareudp->net, port);
239 return PTR_ERR(sock);
241 /* Mark socket as an encapsulation socket */
242 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
243 tunnel_cfg.sk_user_data = bareudp;
244 tunnel_cfg.encap_type = 1;
245 tunnel_cfg.encap_rcv = bareudp_udp_encap_recv;
246 tunnel_cfg.encap_err_lookup = bareudp_err_lookup;
247 tunnel_cfg.encap_destroy = NULL;
248 setup_udp_tunnel_sock(bareudp->net, sock, &tunnel_cfg);
250 rcu_assign_pointer(bareudp->sock, sock);
254 static int bareudp_open(struct net_device *dev)
256 struct bareudp_dev *bareudp = netdev_priv(dev);
259 ret = bareudp_socket_create(bareudp, bareudp->port);
263 static void bareudp_sock_release(struct bareudp_dev *bareudp)
267 sock = bareudp->sock;
268 rcu_assign_pointer(bareudp->sock, NULL);
270 udp_tunnel_sock_release(sock);
273 static int bareudp_stop(struct net_device *dev)
275 struct bareudp_dev *bareudp = netdev_priv(dev);
277 bareudp_sock_release(bareudp);
281 static int bareudp_xmit_skb(struct sk_buff *skb, struct net_device *dev,
282 struct bareudp_dev *bareudp,
283 const struct ip_tunnel_info *info)
285 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
286 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
287 struct socket *sock = rcu_dereference(bareudp->sock);
288 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
289 const struct ip_tunnel_key *key = &info->key;
300 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr, info,
301 IPPROTO_UDP, use_cache);
306 skb_tunnel_check_pmtu(skb, &rt->dst,
307 BAREUDP_IPV4_HLEN + info->options_len, false);
309 sport = udp_flow_src_port(bareudp->net, skb,
310 bareudp->sport_min, USHRT_MAX,
312 tos = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
314 df = key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
315 skb_scrub_packet(skb, xnet);
318 if (!skb_pull(skb, skb_network_offset(skb)))
321 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len +
322 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct iphdr);
324 err = skb_cow_head(skb, min_headroom);
328 err = udp_tunnel_handle_offloads(skb, udp_sum);
332 skb_set_inner_protocol(skb, bareudp->ethertype);
333 udp_tunnel_xmit_skb(rt, sock->sk, skb, saddr, info->key.u.ipv4.dst,
334 tos, ttl, df, sport, bareudp->port,
335 !net_eq(bareudp->net, dev_net(bareudp->dev)),
336 !(info->key.tun_flags & TUNNEL_CSUM));
340 dst_release(&rt->dst);
344 static int bareudp6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
345 struct bareudp_dev *bareudp,
346 const struct ip_tunnel_info *info)
348 bool xnet = !net_eq(bareudp->net, dev_net(bareudp->dev));
349 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
350 struct socket *sock = rcu_dereference(bareudp->sock);
351 bool udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
352 const struct ip_tunnel_key *key = &info->key;
353 struct dst_entry *dst = NULL;
354 struct in6_addr saddr, daddr;
363 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock, &saddr, info,
364 IPPROTO_UDP, use_cache);
368 skb_tunnel_check_pmtu(skb, dst, BAREUDP_IPV6_HLEN + info->options_len,
371 sport = udp_flow_src_port(bareudp->net, skb,
372 bareudp->sport_min, USHRT_MAX,
374 prio = ip_tunnel_ecn_encap(key->tos, ip_hdr(skb), skb);
377 skb_scrub_packet(skb, xnet);
380 if (!skb_pull(skb, skb_network_offset(skb)))
383 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len +
384 BAREUDP_BASE_HLEN + info->options_len + sizeof(struct ipv6hdr);
386 err = skb_cow_head(skb, min_headroom);
390 err = udp_tunnel_handle_offloads(skb, udp_sum);
394 daddr = info->key.u.ipv6.dst;
395 udp_tunnel6_xmit_skb(dst, sock->sk, skb, dev,
396 &saddr, &daddr, prio, ttl,
397 info->key.label, sport, bareudp->port,
398 !(info->key.tun_flags & TUNNEL_CSUM));
406 static bool bareudp_proto_valid(struct bareudp_dev *bareudp, __be16 proto)
408 if (bareudp->ethertype == proto)
411 if (!bareudp->multi_proto_mode)
414 if (bareudp->ethertype == htons(ETH_P_MPLS_UC) &&
415 proto == htons(ETH_P_MPLS_MC))
418 if (bareudp->ethertype == htons(ETH_P_IP) &&
419 proto == htons(ETH_P_IPV6))
425 static netdev_tx_t bareudp_xmit(struct sk_buff *skb, struct net_device *dev)
427 struct bareudp_dev *bareudp = netdev_priv(dev);
428 struct ip_tunnel_info *info = NULL;
431 if (!bareudp_proto_valid(bareudp, skb->protocol)) {
436 info = skb_tunnel_info(skb);
437 if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
443 if (ipv6_mod_enabled() && info->mode & IP_TUNNEL_INFO_IPV6)
444 err = bareudp6_xmit_skb(skb, dev, bareudp, info);
446 err = bareudp_xmit_skb(skb, dev, bareudp, info);
456 dev->stats.collisions++;
457 else if (err == -ENETUNREACH)
458 dev->stats.tx_carrier_errors++;
460 dev->stats.tx_errors++;
464 static int bareudp_fill_metadata_dst(struct net_device *dev,
467 struct ip_tunnel_info *info = skb_tunnel_info(skb);
468 struct bareudp_dev *bareudp = netdev_priv(dev);
471 use_cache = ip_tunnel_dst_cache_usable(skb, info);
473 if (!ipv6_mod_enabled() || ip_tunnel_info_af(info) == AF_INET) {
477 rt = ip_route_output_tunnel(skb, dev, bareudp->net, &saddr,
478 info, IPPROTO_UDP, use_cache);
483 info->key.u.ipv4.src = saddr;
484 } else if (ip_tunnel_info_af(info) == AF_INET6) {
485 struct dst_entry *dst;
486 struct in6_addr saddr;
487 struct socket *sock = rcu_dereference(bareudp->sock);
489 dst = ip6_dst_lookup_tunnel(skb, dev, bareudp->net, sock,
490 &saddr, info, IPPROTO_UDP,
496 info->key.u.ipv6.src = saddr;
501 info->key.tp_src = udp_flow_src_port(bareudp->net, skb,
504 info->key.tp_dst = bareudp->port;
508 static const struct net_device_ops bareudp_netdev_ops = {
509 .ndo_init = bareudp_init,
510 .ndo_uninit = bareudp_uninit,
511 .ndo_open = bareudp_open,
512 .ndo_stop = bareudp_stop,
513 .ndo_start_xmit = bareudp_xmit,
514 .ndo_get_stats64 = ip_tunnel_get_stats64,
515 .ndo_fill_metadata_dst = bareudp_fill_metadata_dst,
518 static const struct nla_policy bareudp_policy[IFLA_BAREUDP_MAX + 1] = {
519 [IFLA_BAREUDP_PORT] = { .type = NLA_U16 },
520 [IFLA_BAREUDP_ETHERTYPE] = { .type = NLA_U16 },
521 [IFLA_BAREUDP_SRCPORT_MIN] = { .type = NLA_U16 },
522 [IFLA_BAREUDP_MULTIPROTO_MODE] = { .type = NLA_FLAG },
525 /* Info for udev, that this is a virtual tunnel endpoint */
526 static struct device_type bareudp_type = {
530 /* Initialize the device structure. */
531 static void bareudp_setup(struct net_device *dev)
533 dev->netdev_ops = &bareudp_netdev_ops;
534 dev->needs_free_netdev = true;
535 SET_NETDEV_DEVTYPE(dev, &bareudp_type);
536 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
537 dev->features |= NETIF_F_RXCSUM;
538 dev->features |= NETIF_F_LLTX;
539 dev->features |= NETIF_F_GSO_SOFTWARE;
540 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
541 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
542 dev->hard_header_len = 0;
544 dev->mtu = ETH_DATA_LEN;
545 dev->min_mtu = IPV4_MIN_MTU;
546 dev->max_mtu = IP_MAX_MTU - BAREUDP_BASE_HLEN;
547 dev->type = ARPHRD_NONE;
549 dev->priv_flags |= IFF_NO_QUEUE;
550 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
553 static int bareudp_validate(struct nlattr *tb[], struct nlattr *data[],
554 struct netlink_ext_ack *extack)
557 NL_SET_ERR_MSG(extack,
558 "Not enough attributes provided to perform the operation");
564 static int bareudp2info(struct nlattr *data[], struct bareudp_conf *conf,
565 struct netlink_ext_ack *extack)
567 memset(conf, 0, sizeof(*conf));
569 if (!data[IFLA_BAREUDP_PORT]) {
570 NL_SET_ERR_MSG(extack, "port not specified");
573 if (!data[IFLA_BAREUDP_ETHERTYPE]) {
574 NL_SET_ERR_MSG(extack, "ethertype not specified");
578 if (data[IFLA_BAREUDP_PORT])
579 conf->port = nla_get_u16(data[IFLA_BAREUDP_PORT]);
581 if (data[IFLA_BAREUDP_ETHERTYPE])
582 conf->ethertype = nla_get_u16(data[IFLA_BAREUDP_ETHERTYPE]);
584 if (data[IFLA_BAREUDP_SRCPORT_MIN])
585 conf->sport_min = nla_get_u16(data[IFLA_BAREUDP_SRCPORT_MIN]);
587 if (data[IFLA_BAREUDP_MULTIPROTO_MODE])
588 conf->multi_proto_mode = true;
593 static struct bareudp_dev *bareudp_find_dev(struct bareudp_net *bn,
594 const struct bareudp_conf *conf)
596 struct bareudp_dev *bareudp, *t = NULL;
598 list_for_each_entry(bareudp, &bn->bareudp_list, next) {
599 if (conf->port == bareudp->port)
605 static int bareudp_configure(struct net *net, struct net_device *dev,
606 struct bareudp_conf *conf)
608 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
609 struct bareudp_dev *t, *bareudp = netdev_priv(dev);
614 t = bareudp_find_dev(bn, conf);
618 if (conf->multi_proto_mode &&
619 (conf->ethertype != htons(ETH_P_MPLS_UC) &&
620 conf->ethertype != htons(ETH_P_IP)))
623 bareudp->port = conf->port;
624 bareudp->ethertype = conf->ethertype;
625 bareudp->sport_min = conf->sport_min;
626 bareudp->multi_proto_mode = conf->multi_proto_mode;
628 err = register_netdevice(dev);
632 list_add(&bareudp->next, &bn->bareudp_list);
636 static int bareudp_link_config(struct net_device *dev,
642 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
649 static void bareudp_dellink(struct net_device *dev, struct list_head *head)
651 struct bareudp_dev *bareudp = netdev_priv(dev);
653 list_del(&bareudp->next);
654 unregister_netdevice_queue(dev, head);
657 static int bareudp_newlink(struct net *net, struct net_device *dev,
658 struct nlattr *tb[], struct nlattr *data[],
659 struct netlink_ext_ack *extack)
661 struct bareudp_conf conf;
662 LIST_HEAD(list_kill);
665 err = bareudp2info(data, &conf, extack);
669 err = bareudp_configure(net, dev, &conf);
673 err = bareudp_link_config(dev, tb);
680 bareudp_dellink(dev, &list_kill);
681 unregister_netdevice_many(&list_kill);
685 static size_t bareudp_get_size(const struct net_device *dev)
687 return nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_PORT */
688 nla_total_size(sizeof(__be16)) + /* IFLA_BAREUDP_ETHERTYPE */
689 nla_total_size(sizeof(__u16)) + /* IFLA_BAREUDP_SRCPORT_MIN */
690 nla_total_size(0) + /* IFLA_BAREUDP_MULTIPROTO_MODE */
694 static int bareudp_fill_info(struct sk_buff *skb, const struct net_device *dev)
696 struct bareudp_dev *bareudp = netdev_priv(dev);
698 if (nla_put_be16(skb, IFLA_BAREUDP_PORT, bareudp->port))
699 goto nla_put_failure;
700 if (nla_put_be16(skb, IFLA_BAREUDP_ETHERTYPE, bareudp->ethertype))
701 goto nla_put_failure;
702 if (nla_put_u16(skb, IFLA_BAREUDP_SRCPORT_MIN, bareudp->sport_min))
703 goto nla_put_failure;
704 if (bareudp->multi_proto_mode &&
705 nla_put_flag(skb, IFLA_BAREUDP_MULTIPROTO_MODE))
706 goto nla_put_failure;
714 static struct rtnl_link_ops bareudp_link_ops __read_mostly = {
716 .maxtype = IFLA_BAREUDP_MAX,
717 .policy = bareudp_policy,
718 .priv_size = sizeof(struct bareudp_dev),
719 .setup = bareudp_setup,
720 .validate = bareudp_validate,
721 .newlink = bareudp_newlink,
722 .dellink = bareudp_dellink,
723 .get_size = bareudp_get_size,
724 .fill_info = bareudp_fill_info,
727 struct net_device *bareudp_dev_create(struct net *net, const char *name,
729 struct bareudp_conf *conf)
731 struct nlattr *tb[IFLA_MAX + 1];
732 struct net_device *dev;
733 LIST_HEAD(list_kill);
736 memset(tb, 0, sizeof(tb));
737 dev = rtnl_create_link(net, name, name_assign_type,
738 &bareudp_link_ops, tb, NULL);
742 err = bareudp_configure(net, dev, conf);
747 err = dev_set_mtu(dev, IP_MAX_MTU - BAREUDP_BASE_HLEN);
751 err = rtnl_configure_link(dev, NULL);
757 bareudp_dellink(dev, &list_kill);
758 unregister_netdevice_many(&list_kill);
761 EXPORT_SYMBOL_GPL(bareudp_dev_create);
763 static __net_init int bareudp_init_net(struct net *net)
765 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
767 INIT_LIST_HEAD(&bn->bareudp_list);
771 static void bareudp_destroy_tunnels(struct net *net, struct list_head *head)
773 struct bareudp_net *bn = net_generic(net, bareudp_net_id);
774 struct bareudp_dev *bareudp, *next;
776 list_for_each_entry_safe(bareudp, next, &bn->bareudp_list, next)
777 unregister_netdevice_queue(bareudp->dev, head);
780 static void __net_exit bareudp_exit_batch_net(struct list_head *net_list)
786 list_for_each_entry(net, net_list, exit_list)
787 bareudp_destroy_tunnels(net, &list);
789 /* unregister the devices gathered above */
790 unregister_netdevice_many(&list);
794 static struct pernet_operations bareudp_net_ops = {
795 .init = bareudp_init_net,
796 .exit_batch = bareudp_exit_batch_net,
797 .id = &bareudp_net_id,
798 .size = sizeof(struct bareudp_net),
801 static int __init bareudp_init_module(void)
805 rc = register_pernet_subsys(&bareudp_net_ops);
809 rc = rtnl_link_register(&bareudp_link_ops);
815 unregister_pernet_subsys(&bareudp_net_ops);
819 late_initcall(bareudp_init_module);
821 static void __exit bareudp_cleanup_module(void)
823 rtnl_link_unregister(&bareudp_link_ops);
824 unregister_pernet_subsys(&bareudp_net_ops);
826 module_exit(bareudp_cleanup_module);
828 MODULE_ALIAS_RTNL_LINK("bareudp");
829 MODULE_LICENSE("GPL");
830 MODULE_AUTHOR("Martin Varghese <martin.varghese@nokia.com>");
831 MODULE_DESCRIPTION("Interface driver for UDP encapsulated traffic");