1 #include <linux/types.h>
2 #include <linux/skbuff.h>
3 #include <linux/socket.h>
4 #include <linux/sysctl.h>
6 #include <linux/module.h>
7 #include <linux/if_arp.h>
8 #include <linux/ipv6.h>
9 #include <linux/mpls.h>
10 #include <linux/netconf.h>
11 #include <linux/nospec.h>
12 #include <linux/vmalloc.h>
13 #include <linux/percpu.h>
18 #include <net/ip_fib.h>
19 #include <net/netevent.h>
20 #include <net/ip_tunnels.h>
21 #include <net/netns/generic.h>
22 #if IS_ENABLED(CONFIG_IPV6)
25 #include <net/ipv6_stubs.h>
29 /* max memory we will use for mpls_route */
30 #define MAX_MPLS_ROUTE_MEM 4096
32 /* Maximum number of labels to look ahead at when selecting a path of
35 #define MAX_MP_SELECT_LABELS 4
37 #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1)
41 static int label_limit = (1 << 20) - 1;
42 static int ttl_max = 255;
44 #if IS_ENABLED(CONFIG_NET_IP_TUNNEL)
45 static size_t ipgre_mpls_encap_hlen(struct ip_tunnel_encap *e)
47 return sizeof(struct mpls_shim_hdr);
50 static const struct ip_tunnel_encap_ops mpls_iptun_ops = {
51 .encap_hlen = ipgre_mpls_encap_hlen,
54 static int ipgre_tunnel_encap_add_mpls_ops(void)
56 return ip_tunnel_encap_add_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
59 static void ipgre_tunnel_encap_del_mpls_ops(void)
61 ip_tunnel_encap_del_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
64 static int ipgre_tunnel_encap_add_mpls_ops(void)
69 static void ipgre_tunnel_encap_del_mpls_ops(void)
74 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
75 struct nlmsghdr *nlh, struct net *net, u32 portid,
76 unsigned int nlm_flags);
78 static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
80 struct mpls_route *rt = NULL;
82 if (index < net->mpls.platform_labels) {
83 struct mpls_route __rcu **platform_label =
84 rcu_dereference(net->mpls.platform_label);
85 rt = rcu_dereference(platform_label[index]);
90 bool mpls_output_possible(const struct net_device *dev)
92 return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev);
94 EXPORT_SYMBOL_GPL(mpls_output_possible);
96 static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh)
98 return (u8 *)nh + rt->rt_via_offset;
101 static const u8 *mpls_nh_via(const struct mpls_route *rt,
102 const struct mpls_nh *nh)
104 return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh);
107 static unsigned int mpls_nh_header_size(const struct mpls_nh *nh)
109 /* The size of the layer 2.5 labels to be added for this route */
110 return nh->nh_labels * sizeof(struct mpls_shim_hdr);
113 unsigned int mpls_dev_mtu(const struct net_device *dev)
115 /* The amount of data the layer 2 frame can hold */
118 EXPORT_SYMBOL_GPL(mpls_dev_mtu);
120 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
125 if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
130 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
132 void mpls_stats_inc_outucastpkts(struct net_device *dev,
133 const struct sk_buff *skb)
135 struct mpls_dev *mdev;
137 if (skb->protocol == htons(ETH_P_MPLS_UC)) {
138 mdev = mpls_dev_get(dev);
140 MPLS_INC_STATS_LEN(mdev, skb->len,
143 } else if (skb->protocol == htons(ETH_P_IP)) {
144 IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
145 #if IS_ENABLED(CONFIG_IPV6)
146 } else if (skb->protocol == htons(ETH_P_IPV6)) {
147 struct inet6_dev *in6dev = __in6_dev_get(dev);
150 IP6_UPD_PO_STATS(dev_net(dev), in6dev,
151 IPSTATS_MIB_OUT, skb->len);
155 EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts);
157 static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb)
159 struct mpls_entry_decoded dec;
160 unsigned int mpls_hdr_len = 0;
161 struct mpls_shim_hdr *hdr;
162 bool eli_seen = false;
166 for (label_index = 0; label_index < MAX_MP_SELECT_LABELS;
168 mpls_hdr_len += sizeof(*hdr);
169 if (!pskb_may_pull(skb, mpls_hdr_len))
172 /* Read and decode the current label */
173 hdr = mpls_hdr(skb) + label_index;
174 dec = mpls_entry_decode(hdr);
176 /* RFC6790 - reserved labels MUST NOT be used as keys
177 * for the load-balancing function
179 if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) {
180 hash = jhash_1word(dec.label, hash);
182 /* The entropy label follows the entropy label
183 * indicator, so this means that the entropy
184 * label was just added to the hash - no need to
185 * go any deeper either in the label stack or in the
190 } else if (dec.label == MPLS_LABEL_ENTROPY) {
197 /* found bottom label; does skb have room for a header? */
198 if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) {
199 const struct iphdr *v4hdr;
201 v4hdr = (const struct iphdr *)(hdr + 1);
202 if (v4hdr->version == 4) {
203 hash = jhash_3words(ntohl(v4hdr->saddr),
205 v4hdr->protocol, hash);
206 } else if (v4hdr->version == 6 &&
207 pskb_may_pull(skb, mpls_hdr_len +
208 sizeof(struct ipv6hdr))) {
209 const struct ipv6hdr *v6hdr;
211 v6hdr = (const struct ipv6hdr *)(hdr + 1);
212 hash = __ipv6_addr_jhash(&v6hdr->saddr, hash);
213 hash = __ipv6_addr_jhash(&v6hdr->daddr, hash);
214 hash = jhash_1word(v6hdr->nexthdr, hash);
224 static struct mpls_nh *mpls_get_nexthop(struct mpls_route *rt, u8 index)
226 return (struct mpls_nh *)((u8 *)rt->rt_nh + index * rt->rt_nh_size);
229 /* number of alive nexthops (rt->rt_nhn_alive) and the flags for
230 * a next hop (nh->nh_flags) are modified by netdev event handlers.
231 * Since those fields can change at any moment, use READ_ONCE to
234 static struct mpls_nh *mpls_select_multipath(struct mpls_route *rt,
242 /* No need to look further into packet if there's only
248 alive = READ_ONCE(rt->rt_nhn_alive);
252 hash = mpls_multipath_hash(rt, skb);
253 nh_index = hash % alive;
254 if (alive == rt->rt_nhn)
257 unsigned int nh_flags = READ_ONCE(nh->nh_flags);
259 if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
264 } endfor_nexthops(rt);
267 return mpls_get_nexthop(rt, nh_index);
270 static bool mpls_egress(struct net *net, struct mpls_route *rt,
271 struct sk_buff *skb, struct mpls_entry_decoded dec)
273 enum mpls_payload_type payload_type;
274 bool success = false;
276 /* The IPv4 code below accesses through the IPv4 header
277 * checksum, which is 12 bytes into the packet.
278 * The IPv6 code below accesses through the IPv6 hop limit
279 * which is 8 bytes into the packet.
281 * For all supported cases there should always be at least 12
282 * bytes of packet data present. The IPv4 header is 20 bytes
283 * without options and the IPv6 header is always 40 bytes
286 if (!pskb_may_pull(skb, 12))
289 payload_type = rt->rt_payload_type;
290 if (payload_type == MPT_UNSPEC)
291 payload_type = ip_hdr(skb)->version;
293 switch (payload_type) {
295 struct iphdr *hdr4 = ip_hdr(skb);
297 skb->protocol = htons(ETH_P_IP);
299 /* If propagating TTL, take the decremented TTL from
300 * the incoming MPLS header, otherwise decrement the
301 * TTL, but only if not 0 to avoid underflow.
303 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
304 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
305 net->mpls.ip_ttl_propagate))
308 new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0;
310 csum_replace2(&hdr4->check,
311 htons(hdr4->ttl << 8),
312 htons(new_ttl << 8));
318 struct ipv6hdr *hdr6 = ipv6_hdr(skb);
319 skb->protocol = htons(ETH_P_IPV6);
321 /* If propagating TTL, take the decremented TTL from
322 * the incoming MPLS header, otherwise decrement the
323 * hop limit, but only if not 0 to avoid underflow.
325 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
326 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
327 net->mpls.ip_ttl_propagate))
328 hdr6->hop_limit = dec.ttl;
329 else if (hdr6->hop_limit)
330 hdr6->hop_limit = hdr6->hop_limit - 1;
335 /* Should have decided which protocol it is by now */
342 static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
343 struct packet_type *pt, struct net_device *orig_dev)
345 struct net *net = dev_net(dev);
346 struct mpls_shim_hdr *hdr;
347 struct mpls_route *rt;
349 struct mpls_entry_decoded dec;
350 struct net_device *out_dev;
351 struct mpls_dev *out_mdev;
352 struct mpls_dev *mdev;
354 unsigned int new_header_size;
358 /* Careful this entire function runs inside of an rcu critical section */
360 mdev = mpls_dev_get(dev);
364 MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets,
367 if (!mdev->input_enabled) {
368 MPLS_INC_STATS(mdev, rx_dropped);
372 if (skb->pkt_type != PACKET_HOST)
375 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
378 if (!pskb_may_pull(skb, sizeof(*hdr)))
381 /* Read and decode the label */
383 dec = mpls_entry_decode(hdr);
385 rt = mpls_route_input_rcu(net, dec.label);
387 MPLS_INC_STATS(mdev, rx_noroute);
391 nh = mpls_select_multipath(rt, skb);
396 skb_pull(skb, sizeof(*hdr));
397 skb_reset_network_header(skb);
401 if (skb_warn_if_lro(skb))
404 skb_forward_csum(skb);
406 /* Verify ttl is valid */
411 /* Find the output device */
412 out_dev = rcu_dereference(nh->nh_dev);
413 if (!mpls_output_possible(out_dev))
416 /* Verify the destination can hold the packet */
417 new_header_size = mpls_nh_header_size(nh);
418 mtu = mpls_dev_mtu(out_dev);
419 if (mpls_pkt_too_big(skb, mtu - new_header_size))
422 hh_len = LL_RESERVED_SPACE(out_dev);
423 if (!out_dev->header_ops)
426 /* Ensure there is enough space for the headers in the skb */
427 if (skb_cow(skb, hh_len + new_header_size))
431 skb->protocol = htons(ETH_P_MPLS_UC);
433 if (unlikely(!new_header_size && dec.bos)) {
434 /* Penultimate hop popping */
435 if (!mpls_egress(dev_net(out_dev), rt, skb, dec))
440 skb_push(skb, new_header_size);
441 skb_reset_network_header(skb);
442 /* Push the new labels */
445 for (i = nh->nh_labels - 1; i >= 0; i--) {
446 hdr[i] = mpls_entry_encode(nh->nh_label[i],
452 mpls_stats_inc_outucastpkts(out_dev, skb);
454 /* If via wasn't specified then send out using device address */
455 if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
456 err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,
457 out_dev->dev_addr, skb);
459 err = neigh_xmit(nh->nh_via_table, out_dev,
460 mpls_nh_via(rt, nh), skb);
462 net_dbg_ratelimited("%s: packet transmission failed: %d\n",
467 out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
469 MPLS_INC_STATS(out_mdev, tx_errors);
472 MPLS_INC_STATS(mdev, rx_errors);
478 static struct packet_type mpls_packet_type __read_mostly = {
479 .type = cpu_to_be16(ETH_P_MPLS_UC),
480 .func = mpls_forward,
483 static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = {
484 [RTA_DST] = { .type = NLA_U32 },
485 [RTA_OIF] = { .type = NLA_U32 },
486 [RTA_TTL_PROPAGATE] = { .type = NLA_U8 },
489 struct mpls_route_config {
494 u8 rc_via[MAX_VIA_ALEN];
498 u32 rc_output_label[MAX_NEW_LABELS];
500 enum mpls_payload_type rc_payload_type;
501 struct nl_info rc_nlinfo;
502 struct rtnexthop *rc_mp;
506 /* all nexthops within a route have the same size based on max
507 * number of labels and max via length for a hop
509 static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels)
511 u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen);
512 struct mpls_route *rt;
515 size = sizeof(*rt) + num_nh * nh_size;
516 if (size > MAX_MPLS_ROUTE_MEM)
517 return ERR_PTR(-EINVAL);
519 rt = kzalloc(size, GFP_KERNEL);
521 return ERR_PTR(-ENOMEM);
524 rt->rt_nhn_alive = num_nh;
525 rt->rt_nh_size = nh_size;
526 rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels);
531 static void mpls_rt_free(struct mpls_route *rt)
534 kfree_rcu(rt, rt_rcu);
537 static void mpls_notify_route(struct net *net, unsigned index,
538 struct mpls_route *old, struct mpls_route *new,
539 const struct nl_info *info)
541 struct nlmsghdr *nlh = info ? info->nlh : NULL;
542 unsigned portid = info ? info->portid : 0;
543 int event = new ? RTM_NEWROUTE : RTM_DELROUTE;
544 struct mpls_route *rt = new ? new : old;
545 unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0;
546 /* Ignore reserved labels for now */
547 if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED))
548 rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags);
551 static void mpls_route_update(struct net *net, unsigned index,
552 struct mpls_route *new,
553 const struct nl_info *info)
555 struct mpls_route __rcu **platform_label;
556 struct mpls_route *rt;
560 platform_label = rtnl_dereference(net->mpls.platform_label);
561 rt = rtnl_dereference(platform_label[index]);
562 rcu_assign_pointer(platform_label[index], new);
564 mpls_notify_route(net, index, rt, new, info);
566 /* If we removed a route free it now */
570 static unsigned find_free_label(struct net *net)
572 struct mpls_route __rcu **platform_label;
573 size_t platform_labels;
576 platform_label = rtnl_dereference(net->mpls.platform_label);
577 platform_labels = net->mpls.platform_labels;
578 for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels;
580 if (!rtnl_dereference(platform_label[index]))
583 return LABEL_NOT_SPECIFIED;
586 #if IS_ENABLED(CONFIG_INET)
587 static struct net_device *inet_fib_lookup_dev(struct net *net,
590 struct net_device *dev;
592 struct in_addr daddr;
594 memcpy(&daddr, addr, sizeof(struct in_addr));
595 rt = ip_route_output(net, daddr.s_addr, 0, 0, 0);
607 static struct net_device *inet_fib_lookup_dev(struct net *net,
610 return ERR_PTR(-EAFNOSUPPORT);
614 #if IS_ENABLED(CONFIG_IPV6)
615 static struct net_device *inet6_fib_lookup_dev(struct net *net,
618 struct net_device *dev;
619 struct dst_entry *dst;
624 return ERR_PTR(-EAFNOSUPPORT);
626 memset(&fl6, 0, sizeof(fl6));
627 memcpy(&fl6.daddr, addr, sizeof(struct in6_addr));
628 err = ipv6_stub->ipv6_dst_lookup(net, NULL, &dst, &fl6);
639 static struct net_device *inet6_fib_lookup_dev(struct net *net,
642 return ERR_PTR(-EAFNOSUPPORT);
646 static struct net_device *find_outdev(struct net *net,
647 struct mpls_route *rt,
648 struct mpls_nh *nh, int oif)
650 struct net_device *dev = NULL;
653 switch (nh->nh_via_table) {
654 case NEIGH_ARP_TABLE:
655 dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh));
658 dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh));
660 case NEIGH_LINK_TABLE:
664 dev = dev_get_by_index(net, oif);
668 return ERR_PTR(-ENODEV);
673 /* The caller is holding rtnl anyways, so release the dev reference */
679 static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
680 struct mpls_nh *nh, int oif)
682 struct net_device *dev = NULL;
685 dev = find_outdev(net, rt, nh, oif);
692 /* Ensure this is a supported device */
694 if (!mpls_dev_get(dev))
697 if ((nh->nh_via_table == NEIGH_LINK_TABLE) &&
698 (dev->addr_len != nh->nh_via_alen))
701 RCU_INIT_POINTER(nh->nh_dev, dev);
703 if (!(dev->flags & IFF_UP)) {
704 nh->nh_flags |= RTNH_F_DEAD;
708 flags = dev_get_flags(dev);
709 if (!(flags & (IFF_RUNNING | IFF_LOWER_UP)))
710 nh->nh_flags |= RTNH_F_LINKDOWN;
719 static int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table,
720 u8 via_addr[], struct netlink_ext_ack *extack)
722 struct rtvia *via = nla_data(nla);
726 if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
727 NL_SET_ERR_MSG_ATTR(extack, nla,
728 "Invalid attribute length for RTA_VIA");
731 alen = nla_len(nla) -
732 offsetof(struct rtvia, rtvia_addr);
733 if (alen > MAX_VIA_ALEN) {
734 NL_SET_ERR_MSG_ATTR(extack, nla,
735 "Invalid address length for RTA_VIA");
739 /* Validate the address family */
740 switch (via->rtvia_family) {
742 *via_table = NEIGH_LINK_TABLE;
745 *via_table = NEIGH_ARP_TABLE;
750 *via_table = NEIGH_ND_TABLE;
755 /* Unsupported address family */
759 memcpy(via_addr, via->rtvia_addr, alen);
767 static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg,
768 struct mpls_route *rt)
770 struct net *net = cfg->rc_nlinfo.nl_net;
771 struct mpls_nh *nh = rt->rt_nh;
778 nh->nh_labels = cfg->rc_output_labels;
779 for (i = 0; i < nh->nh_labels; i++)
780 nh->nh_label[i] = cfg->rc_output_label[i];
782 nh->nh_via_table = cfg->rc_via_table;
783 memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen);
784 nh->nh_via_alen = cfg->rc_via_alen;
786 err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex);
790 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
799 static int mpls_nh_build(struct net *net, struct mpls_route *rt,
800 struct mpls_nh *nh, int oif, struct nlattr *via,
801 struct nlattr *newdst, u8 max_labels,
802 struct netlink_ext_ack *extack)
810 err = nla_get_labels(newdst, max_labels, &nh->nh_labels,
811 nh->nh_label, extack);
817 err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table,
818 __mpls_nh_via(rt, nh), extack);
822 nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC;
825 err = mpls_nh_assign_dev(net, rt, nh, oif);
835 static u8 mpls_count_nexthops(struct rtnexthop *rtnh, int len,
836 u8 cfg_via_alen, u8 *max_via_alen,
845 while (rtnh_ok(rtnh, remaining)) {
846 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
850 attrlen = rtnh_attrlen(rtnh);
851 nla = nla_find(attrs, attrlen, RTA_VIA);
852 if (nla && nla_len(nla) >=
853 offsetof(struct rtvia, rtvia_addr)) {
854 int via_alen = nla_len(nla) -
855 offsetof(struct rtvia, rtvia_addr);
857 if (via_alen <= MAX_VIA_ALEN)
858 *max_via_alen = max_t(u16, *max_via_alen,
862 nla = nla_find(attrs, attrlen, RTA_NEWDST);
864 nla_get_labels(nla, MAX_NEW_LABELS, &n_labels,
868 *max_labels = max_t(u8, *max_labels, n_labels);
870 /* number of nexthops is tracked by a u8.
871 * Check for overflow.
877 rtnh = rtnh_next(rtnh, &remaining);
880 /* leftover implies invalid nexthop configuration, discard it */
881 return remaining > 0 ? 0 : nhs;
884 static int mpls_nh_build_multi(struct mpls_route_config *cfg,
885 struct mpls_route *rt, u8 max_labels,
886 struct netlink_ext_ack *extack)
888 struct rtnexthop *rtnh = cfg->rc_mp;
889 struct nlattr *nla_via, *nla_newdst;
890 int remaining = cfg->rc_mp_len;
894 change_nexthops(rt) {
901 if (!rtnh_ok(rtnh, remaining))
904 /* neither weighted multipath nor any flags
907 if (rtnh->rtnh_hops || rtnh->rtnh_flags)
910 attrlen = rtnh_attrlen(rtnh);
912 struct nlattr *attrs = rtnh_attrs(rtnh);
914 nla_via = nla_find(attrs, attrlen, RTA_VIA);
915 nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST);
918 err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh,
919 rtnh->rtnh_ifindex, nla_via, nla_newdst,
924 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
927 rtnh = rtnh_next(rtnh, &remaining);
929 } endfor_nexthops(rt);
939 static bool mpls_label_ok(struct net *net, unsigned int *index,
940 struct netlink_ext_ack *extack)
944 /* Reserved labels may not be set */
945 if (*index < MPLS_LABEL_FIRST_UNRESERVED) {
946 NL_SET_ERR_MSG(extack,
947 "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher");
951 /* The full 20 bit range may not be supported. */
952 if (is_ok && *index >= net->mpls.platform_labels) {
953 NL_SET_ERR_MSG(extack,
954 "Label >= configured maximum in platform_labels");
958 *index = array_index_nospec(*index, net->mpls.platform_labels);
962 static int mpls_route_add(struct mpls_route_config *cfg,
963 struct netlink_ext_ack *extack)
965 struct mpls_route __rcu **platform_label;
966 struct net *net = cfg->rc_nlinfo.nl_net;
967 struct mpls_route *rt, *old;
974 index = cfg->rc_label;
976 /* If a label was not specified during insert pick one */
977 if ((index == LABEL_NOT_SPECIFIED) &&
978 (cfg->rc_nlflags & NLM_F_CREATE)) {
979 index = find_free_label(net);
982 if (!mpls_label_ok(net, &index, extack))
985 /* Append makes no sense with mpls */
987 if (cfg->rc_nlflags & NLM_F_APPEND) {
988 NL_SET_ERR_MSG(extack, "MPLS does not support route append");
993 platform_label = rtnl_dereference(net->mpls.platform_label);
994 old = rtnl_dereference(platform_label[index]);
995 if ((cfg->rc_nlflags & NLM_F_EXCL) && old)
999 if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old)
1003 if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old)
1008 nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len,
1009 cfg->rc_via_alen, &max_via_alen,
1012 max_via_alen = cfg->rc_via_alen;
1013 max_labels = cfg->rc_output_labels;
1018 NL_SET_ERR_MSG(extack, "Route does not contain a nexthop");
1023 rt = mpls_rt_alloc(nhs, max_via_alen, max_labels);
1029 rt->rt_protocol = cfg->rc_protocol;
1030 rt->rt_payload_type = cfg->rc_payload_type;
1031 rt->rt_ttl_propagate = cfg->rc_ttl_propagate;
1034 err = mpls_nh_build_multi(cfg, rt, max_labels, extack);
1036 err = mpls_nh_build_from_cfg(cfg, rt);
1040 mpls_route_update(net, index, rt, &cfg->rc_nlinfo);
1050 static int mpls_route_del(struct mpls_route_config *cfg,
1051 struct netlink_ext_ack *extack)
1053 struct net *net = cfg->rc_nlinfo.nl_net;
1057 index = cfg->rc_label;
1059 if (!mpls_label_ok(net, &index, extack))
1062 mpls_route_update(net, index, NULL, &cfg->rc_nlinfo);
1069 static void mpls_get_stats(struct mpls_dev *mdev,
1070 struct mpls_link_stats *stats)
1072 struct mpls_pcpu_stats *p;
1075 memset(stats, 0, sizeof(*stats));
1077 for_each_possible_cpu(i) {
1078 struct mpls_link_stats local;
1081 p = per_cpu_ptr(mdev->stats, i);
1083 start = u64_stats_fetch_begin(&p->syncp);
1085 } while (u64_stats_fetch_retry(&p->syncp, start));
1087 stats->rx_packets += local.rx_packets;
1088 stats->rx_bytes += local.rx_bytes;
1089 stats->tx_packets += local.tx_packets;
1090 stats->tx_bytes += local.tx_bytes;
1091 stats->rx_errors += local.rx_errors;
1092 stats->tx_errors += local.tx_errors;
1093 stats->rx_dropped += local.rx_dropped;
1094 stats->tx_dropped += local.tx_dropped;
1095 stats->rx_noroute += local.rx_noroute;
1099 static int mpls_fill_stats_af(struct sk_buff *skb,
1100 const struct net_device *dev)
1102 struct mpls_link_stats *stats;
1103 struct mpls_dev *mdev;
1106 mdev = mpls_dev_get(dev);
1110 nla = nla_reserve_64bit(skb, MPLS_STATS_LINK,
1111 sizeof(struct mpls_link_stats),
1116 stats = nla_data(nla);
1117 mpls_get_stats(mdev, stats);
1122 static size_t mpls_get_stats_af_size(const struct net_device *dev)
1124 struct mpls_dev *mdev;
1126 mdev = mpls_dev_get(dev);
1130 return nla_total_size_64bit(sizeof(struct mpls_link_stats));
1133 static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev,
1134 u32 portid, u32 seq, int event,
1135 unsigned int flags, int type)
1137 struct nlmsghdr *nlh;
1138 struct netconfmsg *ncm;
1141 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1146 if (type == NETCONFA_ALL)
1149 ncm = nlmsg_data(nlh);
1150 ncm->ncm_family = AF_MPLS;
1152 if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0)
1153 goto nla_put_failure;
1155 if ((all || type == NETCONFA_INPUT) &&
1156 nla_put_s32(skb, NETCONFA_INPUT,
1157 mdev->input_enabled) < 0)
1158 goto nla_put_failure;
1160 nlmsg_end(skb, nlh);
1164 nlmsg_cancel(skb, nlh);
1168 static int mpls_netconf_msgsize_devconf(int type)
1170 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1171 + nla_total_size(4); /* NETCONFA_IFINDEX */
1174 if (type == NETCONFA_ALL)
1177 if (all || type == NETCONFA_INPUT)
1178 size += nla_total_size(4);
1183 static void mpls_netconf_notify_devconf(struct net *net, int event,
1184 int type, struct mpls_dev *mdev)
1186 struct sk_buff *skb;
1189 skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL);
1193 err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type);
1195 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1196 WARN_ON(err == -EMSGSIZE);
1201 rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL);
1205 rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err);
1208 static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = {
1209 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1212 static int mpls_netconf_valid_get_req(struct sk_buff *skb,
1213 const struct nlmsghdr *nlh,
1215 struct netlink_ext_ack *extack)
1219 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
1220 NL_SET_ERR_MSG_MOD(extack,
1221 "Invalid header for netconf get request");
1225 if (!netlink_strict_get_check(skb))
1226 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
1228 devconf_mpls_policy, extack);
1230 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
1232 devconf_mpls_policy, extack);
1236 for (i = 0; i <= NETCONFA_MAX; i++) {
1241 case NETCONFA_IFINDEX:
1244 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
1252 static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
1253 struct nlmsghdr *nlh,
1254 struct netlink_ext_ack *extack)
1256 struct net *net = sock_net(in_skb->sk);
1257 struct nlattr *tb[NETCONFA_MAX + 1];
1258 struct net_device *dev;
1259 struct mpls_dev *mdev;
1260 struct sk_buff *skb;
1264 err = mpls_netconf_valid_get_req(in_skb, nlh, tb, extack);
1269 if (!tb[NETCONFA_IFINDEX])
1272 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1273 dev = __dev_get_by_index(net, ifindex);
1277 mdev = mpls_dev_get(dev);
1282 skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
1286 err = mpls_netconf_fill_devconf(skb, mdev,
1287 NETLINK_CB(in_skb).portid,
1288 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1291 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1292 WARN_ON(err == -EMSGSIZE);
1296 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1301 static int mpls_netconf_dump_devconf(struct sk_buff *skb,
1302 struct netlink_callback *cb)
1304 const struct nlmsghdr *nlh = cb->nlh;
1305 struct net *net = sock_net(skb->sk);
1306 struct hlist_head *head;
1307 struct net_device *dev;
1308 struct mpls_dev *mdev;
1312 if (cb->strict_check) {
1313 struct netlink_ext_ack *extack = cb->extack;
1314 struct netconfmsg *ncm;
1316 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
1317 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
1321 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
1322 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
1328 s_idx = idx = cb->args[1];
1330 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1332 head = &net->dev_index_head[h];
1334 cb->seq = net->dev_base_seq;
1335 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1338 mdev = mpls_dev_get(dev);
1341 if (mpls_netconf_fill_devconf(skb, mdev,
1342 NETLINK_CB(cb->skb).portid,
1346 NETCONFA_ALL) < 0) {
1350 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1363 #define MPLS_PERDEV_SYSCTL_OFFSET(field) \
1364 (&((struct mpls_dev *)0)->field)
1366 static int mpls_conf_proc(struct ctl_table *ctl, int write,
1367 void __user *buffer,
1368 size_t *lenp, loff_t *ppos)
1370 int oval = *(int *)ctl->data;
1371 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1374 struct mpls_dev *mdev = ctl->extra1;
1375 int i = (int *)ctl->data - (int *)mdev;
1376 struct net *net = ctl->extra2;
1377 int val = *(int *)ctl->data;
1379 if (i == offsetof(struct mpls_dev, input_enabled) &&
1381 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF,
1382 NETCONFA_INPUT, mdev);
1389 static const struct ctl_table mpls_dev_table[] = {
1391 .procname = "input",
1392 .maxlen = sizeof(int),
1394 .proc_handler = mpls_conf_proc,
1395 .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled),
1400 static int mpls_dev_sysctl_register(struct net_device *dev,
1401 struct mpls_dev *mdev)
1403 char path[sizeof("net/mpls/conf/") + IFNAMSIZ];
1404 struct net *net = dev_net(dev);
1405 struct ctl_table *table;
1408 table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL);
1412 /* Table data contains only offsets relative to the base of
1413 * the mdev at this point, so make them absolute.
1415 for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) {
1416 table[i].data = (char *)mdev + (uintptr_t)table[i].data;
1417 table[i].extra1 = mdev;
1418 table[i].extra2 = net;
1421 snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name);
1423 mdev->sysctl = register_net_sysctl(net, path, table);
1427 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev);
1436 static void mpls_dev_sysctl_unregister(struct net_device *dev,
1437 struct mpls_dev *mdev)
1439 struct net *net = dev_net(dev);
1440 struct ctl_table *table;
1442 table = mdev->sysctl->ctl_table_arg;
1443 unregister_net_sysctl_table(mdev->sysctl);
1446 mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev);
1449 static struct mpls_dev *mpls_add_dev(struct net_device *dev)
1451 struct mpls_dev *mdev;
1457 mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
1459 return ERR_PTR(err);
1461 mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
1465 for_each_possible_cpu(i) {
1466 struct mpls_pcpu_stats *mpls_stats;
1468 mpls_stats = per_cpu_ptr(mdev->stats, i);
1469 u64_stats_init(&mpls_stats->syncp);
1474 err = mpls_dev_sysctl_register(dev, mdev);
1478 rcu_assign_pointer(dev->mpls_ptr, mdev);
1483 free_percpu(mdev->stats);
1485 return ERR_PTR(err);
1488 static void mpls_dev_destroy_rcu(struct rcu_head *head)
1490 struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
1492 free_percpu(mdev->stats);
1496 static void mpls_ifdown(struct net_device *dev, int event)
1498 struct mpls_route __rcu **platform_label;
1499 struct net *net = dev_net(dev);
1503 platform_label = rtnl_dereference(net->mpls.platform_label);
1504 for (index = 0; index < net->mpls.platform_labels; index++) {
1505 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1512 change_nexthops(rt) {
1513 unsigned int nh_flags = nh->nh_flags;
1515 if (rtnl_dereference(nh->nh_dev) != dev)
1520 case NETDEV_UNREGISTER:
1521 nh_flags |= RTNH_F_DEAD;
1524 nh_flags |= RTNH_F_LINKDOWN;
1527 if (event == NETDEV_UNREGISTER)
1528 RCU_INIT_POINTER(nh->nh_dev, NULL);
1530 if (nh->nh_flags != nh_flags)
1531 WRITE_ONCE(nh->nh_flags, nh_flags);
1533 if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)))
1535 if (!rtnl_dereference(nh->nh_dev))
1537 } endfor_nexthops(rt);
1539 WRITE_ONCE(rt->rt_nhn_alive, alive);
1541 /* if there are no more nexthops, delete the route */
1542 if (event == NETDEV_UNREGISTER && deleted == rt->rt_nhn)
1543 mpls_route_update(net, index, NULL, NULL);
1547 static void mpls_ifup(struct net_device *dev, unsigned int flags)
1549 struct mpls_route __rcu **platform_label;
1550 struct net *net = dev_net(dev);
1554 platform_label = rtnl_dereference(net->mpls.platform_label);
1555 for (index = 0; index < net->mpls.platform_labels; index++) {
1556 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1562 change_nexthops(rt) {
1563 unsigned int nh_flags = nh->nh_flags;
1564 struct net_device *nh_dev =
1565 rtnl_dereference(nh->nh_dev);
1567 if (!(nh_flags & flags)) {
1575 WRITE_ONCE(nh->nh_flags, nh_flags);
1576 } endfor_nexthops(rt);
1578 WRITE_ONCE(rt->rt_nhn_alive, alive);
1582 static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
1585 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1586 struct mpls_dev *mdev;
1589 if (event == NETDEV_REGISTER) {
1591 /* For now just support Ethernet, IPGRE, IP6GRE, SIT and
1594 if (dev->type == ARPHRD_ETHER ||
1595 dev->type == ARPHRD_LOOPBACK ||
1596 dev->type == ARPHRD_IPGRE ||
1597 dev->type == ARPHRD_IP6GRE ||
1598 dev->type == ARPHRD_SIT ||
1599 dev->type == ARPHRD_TUNNEL) {
1600 mdev = mpls_add_dev(dev);
1602 return notifier_from_errno(PTR_ERR(mdev));
1607 mdev = mpls_dev_get(dev);
1613 mpls_ifdown(dev, event);
1616 flags = dev_get_flags(dev);
1617 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1618 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1620 mpls_ifup(dev, RTNH_F_DEAD);
1623 flags = dev_get_flags(dev);
1624 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1625 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1627 mpls_ifdown(dev, event);
1629 case NETDEV_UNREGISTER:
1630 mpls_ifdown(dev, event);
1631 mdev = mpls_dev_get(dev);
1633 mpls_dev_sysctl_unregister(dev, mdev);
1634 RCU_INIT_POINTER(dev->mpls_ptr, NULL);
1635 call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
1638 case NETDEV_CHANGENAME:
1639 mdev = mpls_dev_get(dev);
1643 mpls_dev_sysctl_unregister(dev, mdev);
1644 err = mpls_dev_sysctl_register(dev, mdev);
1646 return notifier_from_errno(err);
1653 static struct notifier_block mpls_dev_notifier = {
1654 .notifier_call = mpls_dev_notify,
1657 static int nla_put_via(struct sk_buff *skb,
1658 u8 table, const void *addr, int alen)
1660 static const int table_to_family[NEIGH_NR_TABLES + 1] = {
1661 AF_INET, AF_INET6, AF_DECnet, AF_PACKET,
1665 int family = AF_UNSPEC;
1667 nla = nla_reserve(skb, RTA_VIA, alen + 2);
1671 if (table <= NEIGH_NR_TABLES)
1672 family = table_to_family[table];
1674 via = nla_data(nla);
1675 via->rtvia_family = family;
1676 memcpy(via->rtvia_addr, addr, alen);
1680 int nla_put_labels(struct sk_buff *skb, int attrtype,
1681 u8 labels, const u32 label[])
1684 struct mpls_shim_hdr *nla_label;
1687 nla = nla_reserve(skb, attrtype, labels*4);
1691 nla_label = nla_data(nla);
1693 for (i = labels - 1; i >= 0; i--) {
1694 nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos);
1700 EXPORT_SYMBOL_GPL(nla_put_labels);
1702 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
1703 u32 label[], struct netlink_ext_ack *extack)
1705 unsigned len = nla_len(nla);
1706 struct mpls_shim_hdr *nla_label;
1711 /* len needs to be an even multiple of 4 (the label size). Number
1712 * of labels is a u8 so check for overflow.
1714 if (len & 3 || len / 4 > 255) {
1715 NL_SET_ERR_MSG_ATTR(extack, nla,
1716 "Invalid length for labels attribute");
1720 /* Limit the number of new labels allowed */
1722 if (nla_labels > max_labels) {
1723 NL_SET_ERR_MSG(extack, "Too many labels");
1727 /* when label == NULL, caller wants number of labels */
1731 nla_label = nla_data(nla);
1733 for (i = nla_labels - 1; i >= 0; i--, bos = false) {
1734 struct mpls_entry_decoded dec;
1735 dec = mpls_entry_decode(nla_label + i);
1737 /* Ensure the bottom of stack flag is properly set
1738 * and ttl and tc are both clear.
1741 NL_SET_ERR_MSG_ATTR(extack, nla,
1742 "TTL in label must be 0");
1747 NL_SET_ERR_MSG_ATTR(extack, nla,
1748 "Traffic class in label must be 0");
1752 if (dec.bos != bos) {
1753 NL_SET_BAD_ATTR(extack, nla);
1755 NL_SET_ERR_MSG(extack,
1756 "BOS bit must be set in first label");
1758 NL_SET_ERR_MSG(extack,
1759 "BOS bit can only be set in first label");
1764 switch (dec.label) {
1765 case MPLS_LABEL_IMPLNULL:
1766 /* RFC3032: This is a label that an LSR may
1767 * assign and distribute, but which never
1768 * actually appears in the encapsulation.
1770 NL_SET_ERR_MSG_ATTR(extack, nla,
1771 "Implicit NULL Label (3) can not be used in encapsulation");
1775 label[i] = dec.label;
1778 *labels = nla_labels;
1781 EXPORT_SYMBOL_GPL(nla_get_labels);
1783 static int rtm_to_route_config(struct sk_buff *skb,
1784 struct nlmsghdr *nlh,
1785 struct mpls_route_config *cfg,
1786 struct netlink_ext_ack *extack)
1789 struct nlattr *tb[RTA_MAX+1];
1793 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
1794 rtm_mpls_policy, extack);
1799 rtm = nlmsg_data(nlh);
1801 if (rtm->rtm_family != AF_MPLS) {
1802 NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg");
1805 if (rtm->rtm_dst_len != 20) {
1806 NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS");
1809 if (rtm->rtm_src_len != 0) {
1810 NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS");
1813 if (rtm->rtm_tos != 0) {
1814 NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS");
1817 if (rtm->rtm_table != RT_TABLE_MAIN) {
1818 NL_SET_ERR_MSG(extack,
1819 "MPLS only supports the main route table");
1822 /* Any value is acceptable for rtm_protocol */
1824 /* As mpls uses destination specific addresses
1825 * (or source specific address in the case of multicast)
1826 * all addresses have universal scope.
1828 if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) {
1829 NL_SET_ERR_MSG(extack,
1830 "Invalid route scope - MPLS only supports UNIVERSE");
1833 if (rtm->rtm_type != RTN_UNICAST) {
1834 NL_SET_ERR_MSG(extack,
1835 "Invalid route type - MPLS only supports UNICAST");
1838 if (rtm->rtm_flags != 0) {
1839 NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS");
1843 cfg->rc_label = LABEL_NOT_SPECIFIED;
1844 cfg->rc_protocol = rtm->rtm_protocol;
1845 cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC;
1846 cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
1847 cfg->rc_nlflags = nlh->nlmsg_flags;
1848 cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid;
1849 cfg->rc_nlinfo.nlh = nlh;
1850 cfg->rc_nlinfo.nl_net = sock_net(skb->sk);
1852 for (index = 0; index <= RTA_MAX; index++) {
1853 struct nlattr *nla = tb[index];
1859 cfg->rc_ifindex = nla_get_u32(nla);
1862 if (nla_get_labels(nla, MAX_NEW_LABELS,
1863 &cfg->rc_output_labels,
1864 cfg->rc_output_label, extack))
1870 if (nla_get_labels(nla, 1, &label_count,
1871 &cfg->rc_label, extack))
1874 if (!mpls_label_ok(cfg->rc_nlinfo.nl_net,
1875 &cfg->rc_label, extack))
1880 NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute");
1884 if (nla_get_via(nla, &cfg->rc_via_alen,
1885 &cfg->rc_via_table, cfg->rc_via,
1892 cfg->rc_mp = nla_data(nla);
1893 cfg->rc_mp_len = nla_len(nla);
1896 case RTA_TTL_PROPAGATE:
1898 u8 ttl_propagate = nla_get_u8(nla);
1900 if (ttl_propagate > 1) {
1901 NL_SET_ERR_MSG_ATTR(extack, nla,
1902 "RTA_TTL_PROPAGATE can only be 0 or 1");
1905 cfg->rc_ttl_propagate = ttl_propagate ?
1906 MPLS_TTL_PROP_ENABLED :
1907 MPLS_TTL_PROP_DISABLED;
1911 NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute");
1912 /* Unsupported attribute */
1922 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1923 struct netlink_ext_ack *extack)
1925 struct mpls_route_config *cfg;
1928 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1932 err = rtm_to_route_config(skb, nlh, cfg, extack);
1936 err = mpls_route_del(cfg, extack);
1944 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1945 struct netlink_ext_ack *extack)
1947 struct mpls_route_config *cfg;
1950 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1954 err = rtm_to_route_config(skb, nlh, cfg, extack);
1958 err = mpls_route_add(cfg, extack);
1965 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event,
1966 u32 label, struct mpls_route *rt, int flags)
1968 struct net_device *dev;
1969 struct nlmsghdr *nlh;
1972 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1976 rtm = nlmsg_data(nlh);
1977 rtm->rtm_family = AF_MPLS;
1978 rtm->rtm_dst_len = 20;
1979 rtm->rtm_src_len = 0;
1981 rtm->rtm_table = RT_TABLE_MAIN;
1982 rtm->rtm_protocol = rt->rt_protocol;
1983 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
1984 rtm->rtm_type = RTN_UNICAST;
1987 if (nla_put_labels(skb, RTA_DST, 1, &label))
1988 goto nla_put_failure;
1990 if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) {
1991 bool ttl_propagate =
1992 rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED;
1994 if (nla_put_u8(skb, RTA_TTL_PROPAGATE,
1996 goto nla_put_failure;
1998 if (rt->rt_nhn == 1) {
1999 const struct mpls_nh *nh = rt->rt_nh;
2001 if (nh->nh_labels &&
2002 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2004 goto nla_put_failure;
2005 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2006 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2008 goto nla_put_failure;
2009 dev = rtnl_dereference(nh->nh_dev);
2010 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2011 goto nla_put_failure;
2012 if (nh->nh_flags & RTNH_F_LINKDOWN)
2013 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2014 if (nh->nh_flags & RTNH_F_DEAD)
2015 rtm->rtm_flags |= RTNH_F_DEAD;
2017 struct rtnexthop *rtnh;
2022 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
2024 goto nla_put_failure;
2027 dev = rtnl_dereference(nh->nh_dev);
2031 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
2033 goto nla_put_failure;
2035 rtnh->rtnh_ifindex = dev->ifindex;
2036 if (nh->nh_flags & RTNH_F_LINKDOWN) {
2037 rtnh->rtnh_flags |= RTNH_F_LINKDOWN;
2040 if (nh->nh_flags & RTNH_F_DEAD) {
2041 rtnh->rtnh_flags |= RTNH_F_DEAD;
2045 if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST,
2048 goto nla_put_failure;
2049 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2050 nla_put_via(skb, nh->nh_via_table,
2051 mpls_nh_via(rt, nh),
2053 goto nla_put_failure;
2055 /* length of rtnetlink header + attributes */
2056 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
2057 } endfor_nexthops(rt);
2059 if (linkdown == rt->rt_nhn)
2060 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2061 if (dead == rt->rt_nhn)
2062 rtm->rtm_flags |= RTNH_F_DEAD;
2064 nla_nest_end(skb, mp);
2067 nlmsg_end(skb, nlh);
2071 nlmsg_cancel(skb, nlh);
2075 #if IS_ENABLED(CONFIG_INET)
2076 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2077 struct fib_dump_filter *filter,
2078 struct netlink_callback *cb)
2080 return ip_valid_fib_dump_req(net, nlh, filter, cb);
2083 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2084 struct fib_dump_filter *filter,
2085 struct netlink_callback *cb)
2087 struct netlink_ext_ack *extack = cb->extack;
2088 struct nlattr *tb[RTA_MAX + 1];
2092 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2093 NL_SET_ERR_MSG_MOD(extack, "Invalid header for FIB dump request");
2097 rtm = nlmsg_data(nlh);
2098 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos ||
2099 rtm->rtm_table || rtm->rtm_scope || rtm->rtm_type ||
2101 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for FIB dump request");
2105 if (rtm->rtm_protocol) {
2106 filter->protocol = rtm->rtm_protocol;
2107 filter->filter_set = 1;
2108 cb->answer_flags = NLM_F_DUMP_FILTERED;
2111 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2112 rtm_mpls_policy, extack);
2116 for (i = 0; i <= RTA_MAX; ++i) {
2120 ifindex = nla_get_u32(tb[i]);
2121 filter->dev = __dev_get_by_index(net, ifindex);
2124 filter->filter_set = 1;
2126 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
2135 static bool mpls_rt_uses_dev(struct mpls_route *rt,
2136 const struct net_device *dev)
2138 struct net_device *nh_dev;
2140 if (rt->rt_nhn == 1) {
2141 struct mpls_nh *nh = rt->rt_nh;
2143 nh_dev = rtnl_dereference(nh->nh_dev);
2148 nh_dev = rtnl_dereference(nh->nh_dev);
2151 } endfor_nexthops(rt);
2157 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
2159 const struct nlmsghdr *nlh = cb->nlh;
2160 struct net *net = sock_net(skb->sk);
2161 struct mpls_route __rcu **platform_label;
2162 struct fib_dump_filter filter = {};
2163 unsigned int flags = NLM_F_MULTI;
2164 size_t platform_labels;
2169 if (cb->strict_check) {
2172 err = mpls_valid_fib_dump_req(net, nlh, &filter, cb);
2176 /* for MPLS, there is only 1 table with fixed type and flags.
2177 * If either are set in the filter then return nothing.
2179 if ((filter.table_id && filter.table_id != RT_TABLE_MAIN) ||
2180 (filter.rt_type && filter.rt_type != RTN_UNICAST) ||
2185 index = cb->args[0];
2186 if (index < MPLS_LABEL_FIRST_UNRESERVED)
2187 index = MPLS_LABEL_FIRST_UNRESERVED;
2189 platform_label = rtnl_dereference(net->mpls.platform_label);
2190 platform_labels = net->mpls.platform_labels;
2192 if (filter.filter_set)
2193 flags |= NLM_F_DUMP_FILTERED;
2195 for (; index < platform_labels; index++) {
2196 struct mpls_route *rt;
2198 rt = rtnl_dereference(platform_label[index]);
2202 if ((filter.dev && !mpls_rt_uses_dev(rt, filter.dev)) ||
2203 (filter.protocol && rt->rt_protocol != filter.protocol))
2206 if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid,
2207 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2208 index, rt, flags) < 0)
2211 cb->args[0] = index;
2216 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
2219 NLMSG_ALIGN(sizeof(struct rtmsg))
2220 + nla_total_size(4) /* RTA_DST */
2221 + nla_total_size(1); /* RTA_TTL_PROPAGATE */
2223 if (rt->rt_nhn == 1) {
2224 struct mpls_nh *nh = rt->rt_nh;
2227 payload += nla_total_size(4); /* RTA_OIF */
2228 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */
2229 payload += nla_total_size(2 + nh->nh_via_alen);
2230 if (nh->nh_labels) /* RTA_NEWDST */
2231 payload += nla_total_size(nh->nh_labels * 4);
2233 /* each nexthop is packed in an attribute */
2237 if (!rtnl_dereference(nh->nh_dev))
2239 nhsize += nla_total_size(sizeof(struct rtnexthop));
2241 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC)
2242 nhsize += nla_total_size(2 + nh->nh_via_alen);
2244 nhsize += nla_total_size(nh->nh_labels * 4);
2245 } endfor_nexthops(rt);
2246 /* nested attribute */
2247 payload += nla_total_size(nhsize);
2253 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
2254 struct nlmsghdr *nlh, struct net *net, u32 portid,
2255 unsigned int nlm_flags)
2257 struct sk_buff *skb;
2258 u32 seq = nlh ? nlh->nlmsg_seq : 0;
2261 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2265 err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags);
2267 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2268 WARN_ON(err == -EMSGSIZE);
2272 rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL);
2277 rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err);
2280 static int mpls_valid_getroute_req(struct sk_buff *skb,
2281 const struct nlmsghdr *nlh,
2283 struct netlink_ext_ack *extack)
2288 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2289 NL_SET_ERR_MSG_MOD(extack,
2290 "Invalid header for get route request");
2294 if (!netlink_strict_get_check(skb))
2295 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2296 rtm_mpls_policy, extack);
2298 rtm = nlmsg_data(nlh);
2299 if ((rtm->rtm_dst_len && rtm->rtm_dst_len != 20) ||
2300 rtm->rtm_src_len || rtm->rtm_tos || rtm->rtm_table ||
2301 rtm->rtm_protocol || rtm->rtm_scope || rtm->rtm_type) {
2302 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
2305 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
2306 NL_SET_ERR_MSG_MOD(extack,
2307 "Invalid flags for get route request");
2311 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2312 rtm_mpls_policy, extack);
2316 if ((tb[RTA_DST] || tb[RTA_NEWDST]) && !rtm->rtm_dst_len) {
2317 NL_SET_ERR_MSG_MOD(extack, "rtm_dst_len must be 20 for MPLS");
2321 for (i = 0; i <= RTA_MAX; i++) {
2330 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
2338 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
2339 struct netlink_ext_ack *extack)
2341 struct net *net = sock_net(in_skb->sk);
2342 u32 portid = NETLINK_CB(in_skb).portid;
2343 u32 in_label = LABEL_NOT_SPECIFIED;
2344 struct nlattr *tb[RTA_MAX + 1];
2345 u32 labels[MAX_NEW_LABELS];
2346 struct mpls_shim_hdr *hdr;
2347 unsigned int hdr_size = 0;
2348 struct net_device *dev;
2349 struct mpls_route *rt;
2350 struct rtmsg *rtm, *r;
2351 struct nlmsghdr *nlh;
2352 struct sk_buff *skb;
2357 err = mpls_valid_getroute_req(in_skb, in_nlh, tb, extack);
2361 rtm = nlmsg_data(in_nlh);
2366 if (nla_get_labels(tb[RTA_DST], 1, &label_count,
2367 &in_label, extack)) {
2372 if (!mpls_label_ok(net, &in_label, extack)) {
2378 rt = mpls_route_input_rcu(net, in_label);
2384 if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
2385 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2391 err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq,
2392 RTM_NEWROUTE, in_label, rt, 0);
2394 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2395 WARN_ON(err == -EMSGSIZE);
2399 return rtnl_unicast(skb, net, portid);
2402 if (tb[RTA_NEWDST]) {
2403 if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels,
2404 labels, extack) != 0) {
2409 hdr_size = n_labels * sizeof(struct mpls_shim_hdr);
2412 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2418 skb->protocol = htons(ETH_P_MPLS_UC);
2424 if (skb_cow(skb, hdr_size)) {
2429 skb_reserve(skb, hdr_size);
2430 skb_push(skb, hdr_size);
2431 skb_reset_network_header(skb);
2433 /* Push new labels */
2434 hdr = mpls_hdr(skb);
2436 for (i = n_labels - 1; i >= 0; i--) {
2437 hdr[i] = mpls_entry_encode(labels[i],
2443 nh = mpls_select_multipath(rt, skb);
2450 skb_pull(skb, hdr_size);
2451 skb_reset_network_header(skb);
2454 nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq,
2455 RTM_NEWROUTE, sizeof(*r), 0);
2461 r = nlmsg_data(nlh);
2462 r->rtm_family = AF_MPLS;
2463 r->rtm_dst_len = 20;
2465 r->rtm_table = RT_TABLE_MAIN;
2466 r->rtm_type = RTN_UNICAST;
2467 r->rtm_scope = RT_SCOPE_UNIVERSE;
2468 r->rtm_protocol = rt->rt_protocol;
2471 if (nla_put_labels(skb, RTA_DST, 1, &in_label))
2472 goto nla_put_failure;
2474 if (nh->nh_labels &&
2475 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2477 goto nla_put_failure;
2479 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2480 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2482 goto nla_put_failure;
2483 dev = rtnl_dereference(nh->nh_dev);
2484 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2485 goto nla_put_failure;
2487 nlmsg_end(skb, nlh);
2489 err = rtnl_unicast(skb, net, portid);
2494 nlmsg_cancel(skb, nlh);
2501 static int resize_platform_label_table(struct net *net, size_t limit)
2503 size_t size = sizeof(struct mpls_route *) * limit;
2506 struct mpls_route __rcu **labels = NULL, **old;
2507 struct mpls_route *rt0 = NULL, *rt2 = NULL;
2511 labels = kvzalloc(size, GFP_KERNEL);
2516 /* In case the predefined labels need to be populated */
2517 if (limit > MPLS_LABEL_IPV4NULL) {
2518 struct net_device *lo = net->loopback_dev;
2519 rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
2522 RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo);
2523 rt0->rt_protocol = RTPROT_KERNEL;
2524 rt0->rt_payload_type = MPT_IPV4;
2525 rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2526 rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2527 rt0->rt_nh->nh_via_alen = lo->addr_len;
2528 memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr,
2531 if (limit > MPLS_LABEL_IPV6NULL) {
2532 struct net_device *lo = net->loopback_dev;
2533 rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
2536 RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo);
2537 rt2->rt_protocol = RTPROT_KERNEL;
2538 rt2->rt_payload_type = MPT_IPV6;
2539 rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2540 rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2541 rt2->rt_nh->nh_via_alen = lo->addr_len;
2542 memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr,
2547 /* Remember the original table */
2548 old = rtnl_dereference(net->mpls.platform_label);
2549 old_limit = net->mpls.platform_labels;
2551 /* Free any labels beyond the new table */
2552 for (index = limit; index < old_limit; index++)
2553 mpls_route_update(net, index, NULL, NULL);
2555 /* Copy over the old labels */
2557 if (old_limit < limit)
2558 cp_size = old_limit * sizeof(struct mpls_route *);
2560 memcpy(labels, old, cp_size);
2562 /* If needed set the predefined labels */
2563 if ((old_limit <= MPLS_LABEL_IPV6NULL) &&
2564 (limit > MPLS_LABEL_IPV6NULL)) {
2565 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2);
2569 if ((old_limit <= MPLS_LABEL_IPV4NULL) &&
2570 (limit > MPLS_LABEL_IPV4NULL)) {
2571 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0);
2575 /* Update the global pointers */
2576 net->mpls.platform_labels = limit;
2577 rcu_assign_pointer(net->mpls.platform_label, labels);
2598 static int mpls_platform_labels(struct ctl_table *table, int write,
2599 void __user *buffer, size_t *lenp, loff_t *ppos)
2601 struct net *net = table->data;
2602 int platform_labels = net->mpls.platform_labels;
2604 struct ctl_table tmp = {
2605 .procname = table->procname,
2606 .data = &platform_labels,
2607 .maxlen = sizeof(int),
2608 .mode = table->mode,
2610 .extra2 = &label_limit,
2613 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2615 if (write && ret == 0)
2616 ret = resize_platform_label_table(net, platform_labels);
2621 #define MPLS_NS_SYSCTL_OFFSET(field) \
2622 (&((struct net *)0)->field)
2624 static const struct ctl_table mpls_table[] = {
2626 .procname = "platform_labels",
2628 .maxlen = sizeof(int),
2630 .proc_handler = mpls_platform_labels,
2633 .procname = "ip_ttl_propagate",
2634 .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate),
2635 .maxlen = sizeof(int),
2637 .proc_handler = proc_dointvec_minmax,
2642 .procname = "default_ttl",
2643 .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl),
2644 .maxlen = sizeof(int),
2646 .proc_handler = proc_dointvec_minmax,
2653 static int mpls_net_init(struct net *net)
2655 struct ctl_table *table;
2658 net->mpls.platform_labels = 0;
2659 net->mpls.platform_label = NULL;
2660 net->mpls.ip_ttl_propagate = 1;
2661 net->mpls.default_ttl = 255;
2663 table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL);
2667 /* Table data contains only offsets relative to the base of
2668 * the mdev at this point, so make them absolute.
2670 for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++)
2671 table[i].data = (char *)net + (uintptr_t)table[i].data;
2673 net->mpls.ctl = register_net_sysctl(net, "net/mpls", table);
2674 if (net->mpls.ctl == NULL) {
2682 static void mpls_net_exit(struct net *net)
2684 struct mpls_route __rcu **platform_label;
2685 size_t platform_labels;
2686 struct ctl_table *table;
2689 table = net->mpls.ctl->ctl_table_arg;
2690 unregister_net_sysctl_table(net->mpls.ctl);
2693 /* An rcu grace period has passed since there was a device in
2694 * the network namespace (and thus the last in flight packet)
2695 * left this network namespace. This is because
2696 * unregister_netdevice_many and netdev_run_todo has completed
2697 * for each network device that was in this network namespace.
2699 * As such no additional rcu synchronization is necessary when
2700 * freeing the platform_label table.
2703 platform_label = rtnl_dereference(net->mpls.platform_label);
2704 platform_labels = net->mpls.platform_labels;
2705 for (index = 0; index < platform_labels; index++) {
2706 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
2707 RCU_INIT_POINTER(platform_label[index], NULL);
2708 mpls_notify_route(net, index, rt, NULL, NULL);
2713 kvfree(platform_label);
2716 static struct pernet_operations mpls_net_ops = {
2717 .init = mpls_net_init,
2718 .exit = mpls_net_exit,
2721 static struct rtnl_af_ops mpls_af_ops __read_mostly = {
2723 .fill_stats_af = mpls_fill_stats_af,
2724 .get_stats_af_size = mpls_get_stats_af_size,
2727 static int __init mpls_init(void)
2731 BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4);
2733 err = register_pernet_subsys(&mpls_net_ops);
2737 err = register_netdevice_notifier(&mpls_dev_notifier);
2739 goto out_unregister_pernet;
2741 dev_add_pack(&mpls_packet_type);
2743 rtnl_af_register(&mpls_af_ops);
2745 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_NEWROUTE,
2746 mpls_rtm_newroute, NULL, 0);
2747 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_DELROUTE,
2748 mpls_rtm_delroute, NULL, 0);
2749 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETROUTE,
2750 mpls_getroute, mpls_dump_routes, 0);
2751 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
2752 mpls_netconf_get_devconf,
2753 mpls_netconf_dump_devconf, 0);
2754 err = ipgre_tunnel_encap_add_mpls_ops();
2756 pr_err("Can't add mpls over gre tunnel ops\n");
2762 out_unregister_pernet:
2763 unregister_pernet_subsys(&mpls_net_ops);
2766 module_init(mpls_init);
2768 static void __exit mpls_exit(void)
2770 rtnl_unregister_all(PF_MPLS);
2771 rtnl_af_unregister(&mpls_af_ops);
2772 dev_remove_pack(&mpls_packet_type);
2773 unregister_netdevice_notifier(&mpls_dev_notifier);
2774 unregister_pernet_subsys(&mpls_net_ops);
2775 ipgre_tunnel_encap_del_mpls_ops();
2777 module_exit(mpls_exit);
2779 MODULE_DESCRIPTION("MultiProtocol Label Switching");
2780 MODULE_LICENSE("GPL v2");
2781 MODULE_ALIAS_NETPROTO(PF_MPLS);