1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/types.h>
3 #include <linux/skbuff.h>
4 #include <linux/socket.h>
5 #include <linux/sysctl.h>
7 #include <linux/module.h>
8 #include <linux/if_arp.h>
9 #include <linux/ipv6.h>
10 #include <linux/mpls.h>
11 #include <linux/netconf.h>
12 #include <linux/nospec.h>
13 #include <linux/vmalloc.h>
14 #include <linux/percpu.h>
19 #include <net/ip_fib.h>
20 #include <net/netevent.h>
21 #include <net/ip_tunnels.h>
22 #include <net/netns/generic.h>
23 #if IS_ENABLED(CONFIG_IPV6)
26 #include <net/ipv6_stubs.h>
30 /* max memory we will use for mpls_route */
31 #define MAX_MPLS_ROUTE_MEM 4096
33 /* Maximum number of labels to look ahead at when selecting a path of
36 #define MAX_MP_SELECT_LABELS 4
38 #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1)
40 static int label_limit = (1 << 20) - 1;
41 static int ttl_max = 255;
43 #if IS_ENABLED(CONFIG_NET_IP_TUNNEL)
44 static size_t ipgre_mpls_encap_hlen(struct ip_tunnel_encap *e)
46 return sizeof(struct mpls_shim_hdr);
49 static const struct ip_tunnel_encap_ops mpls_iptun_ops = {
50 .encap_hlen = ipgre_mpls_encap_hlen,
53 static int ipgre_tunnel_encap_add_mpls_ops(void)
55 return ip_tunnel_encap_add_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
58 static void ipgre_tunnel_encap_del_mpls_ops(void)
60 ip_tunnel_encap_del_ops(&mpls_iptun_ops, TUNNEL_ENCAP_MPLS);
63 static int ipgre_tunnel_encap_add_mpls_ops(void)
68 static void ipgre_tunnel_encap_del_mpls_ops(void)
73 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
74 struct nlmsghdr *nlh, struct net *net, u32 portid,
75 unsigned int nlm_flags);
77 static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
79 struct mpls_route *rt = NULL;
81 if (index < net->mpls.platform_labels) {
82 struct mpls_route __rcu **platform_label =
83 rcu_dereference(net->mpls.platform_label);
84 rt = rcu_dereference(platform_label[index]);
89 bool mpls_output_possible(const struct net_device *dev)
91 return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev);
93 EXPORT_SYMBOL_GPL(mpls_output_possible);
95 static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh)
97 return (u8 *)nh + rt->rt_via_offset;
100 static const u8 *mpls_nh_via(const struct mpls_route *rt,
101 const struct mpls_nh *nh)
103 return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh);
106 static unsigned int mpls_nh_header_size(const struct mpls_nh *nh)
108 /* The size of the layer 2.5 labels to be added for this route */
109 return nh->nh_labels * sizeof(struct mpls_shim_hdr);
112 unsigned int mpls_dev_mtu(const struct net_device *dev)
114 /* The amount of data the layer 2 frame can hold */
117 EXPORT_SYMBOL_GPL(mpls_dev_mtu);
119 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
124 if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
129 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
131 void mpls_stats_inc_outucastpkts(struct net_device *dev,
132 const struct sk_buff *skb)
134 struct mpls_dev *mdev;
136 if (skb->protocol == htons(ETH_P_MPLS_UC)) {
137 mdev = mpls_dev_get(dev);
139 MPLS_INC_STATS_LEN(mdev, skb->len,
142 } else if (skb->protocol == htons(ETH_P_IP)) {
143 IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
144 #if IS_ENABLED(CONFIG_IPV6)
145 } else if (skb->protocol == htons(ETH_P_IPV6)) {
146 struct inet6_dev *in6dev = __in6_dev_get(dev);
149 IP6_UPD_PO_STATS(dev_net(dev), in6dev,
150 IPSTATS_MIB_OUT, skb->len);
154 EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts);
156 static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb)
158 struct mpls_entry_decoded dec;
159 unsigned int mpls_hdr_len = 0;
160 struct mpls_shim_hdr *hdr;
161 bool eli_seen = false;
165 for (label_index = 0; label_index < MAX_MP_SELECT_LABELS;
167 mpls_hdr_len += sizeof(*hdr);
168 if (!pskb_may_pull(skb, mpls_hdr_len))
171 /* Read and decode the current label */
172 hdr = mpls_hdr(skb) + label_index;
173 dec = mpls_entry_decode(hdr);
175 /* RFC6790 - reserved labels MUST NOT be used as keys
176 * for the load-balancing function
178 if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) {
179 hash = jhash_1word(dec.label, hash);
181 /* The entropy label follows the entropy label
182 * indicator, so this means that the entropy
183 * label was just added to the hash - no need to
184 * go any deeper either in the label stack or in the
189 } else if (dec.label == MPLS_LABEL_ENTROPY) {
196 /* found bottom label; does skb have room for a header? */
197 if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) {
198 const struct iphdr *v4hdr;
200 v4hdr = (const struct iphdr *)(hdr + 1);
201 if (v4hdr->version == 4) {
202 hash = jhash_3words(ntohl(v4hdr->saddr),
204 v4hdr->protocol, hash);
205 } else if (v4hdr->version == 6 &&
206 pskb_may_pull(skb, mpls_hdr_len +
207 sizeof(struct ipv6hdr))) {
208 const struct ipv6hdr *v6hdr;
210 v6hdr = (const struct ipv6hdr *)(hdr + 1);
211 hash = __ipv6_addr_jhash(&v6hdr->saddr, hash);
212 hash = __ipv6_addr_jhash(&v6hdr->daddr, hash);
213 hash = jhash_1word(v6hdr->nexthdr, hash);
223 static struct mpls_nh *mpls_get_nexthop(struct mpls_route *rt, u8 index)
225 return (struct mpls_nh *)((u8 *)rt->rt_nh + index * rt->rt_nh_size);
228 /* number of alive nexthops (rt->rt_nhn_alive) and the flags for
229 * a next hop (nh->nh_flags) are modified by netdev event handlers.
230 * Since those fields can change at any moment, use READ_ONCE to
233 static const struct mpls_nh *mpls_select_multipath(struct mpls_route *rt,
241 /* No need to look further into packet if there's only
247 alive = READ_ONCE(rt->rt_nhn_alive);
251 hash = mpls_multipath_hash(rt, skb);
252 nh_index = hash % alive;
253 if (alive == rt->rt_nhn)
256 unsigned int nh_flags = READ_ONCE(nh->nh_flags);
258 if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
263 } endfor_nexthops(rt);
266 return mpls_get_nexthop(rt, nh_index);
269 static bool mpls_egress(struct net *net, struct mpls_route *rt,
270 struct sk_buff *skb, struct mpls_entry_decoded dec)
272 enum mpls_payload_type payload_type;
273 bool success = false;
275 /* The IPv4 code below accesses through the IPv4 header
276 * checksum, which is 12 bytes into the packet.
277 * The IPv6 code below accesses through the IPv6 hop limit
278 * which is 8 bytes into the packet.
280 * For all supported cases there should always be at least 12
281 * bytes of packet data present. The IPv4 header is 20 bytes
282 * without options and the IPv6 header is always 40 bytes
285 if (!pskb_may_pull(skb, 12))
288 payload_type = rt->rt_payload_type;
289 if (payload_type == MPT_UNSPEC)
290 payload_type = ip_hdr(skb)->version;
292 switch (payload_type) {
294 struct iphdr *hdr4 = ip_hdr(skb);
296 skb->protocol = htons(ETH_P_IP);
298 /* If propagating TTL, take the decremented TTL from
299 * the incoming MPLS header, otherwise decrement the
300 * TTL, but only if not 0 to avoid underflow.
302 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
303 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
304 net->mpls.ip_ttl_propagate))
307 new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0;
309 csum_replace2(&hdr4->check,
310 htons(hdr4->ttl << 8),
311 htons(new_ttl << 8));
317 struct ipv6hdr *hdr6 = ipv6_hdr(skb);
318 skb->protocol = htons(ETH_P_IPV6);
320 /* If propagating TTL, take the decremented TTL from
321 * the incoming MPLS header, otherwise decrement the
322 * hop limit, but only if not 0 to avoid underflow.
324 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
325 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
326 net->mpls.ip_ttl_propagate))
327 hdr6->hop_limit = dec.ttl;
328 else if (hdr6->hop_limit)
329 hdr6->hop_limit = hdr6->hop_limit - 1;
334 /* Should have decided which protocol it is by now */
341 static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
342 struct packet_type *pt, struct net_device *orig_dev)
344 struct net *net = dev_net(dev);
345 struct mpls_shim_hdr *hdr;
346 const struct mpls_nh *nh;
347 struct mpls_route *rt;
348 struct mpls_entry_decoded dec;
349 struct net_device *out_dev;
350 struct mpls_dev *out_mdev;
351 struct mpls_dev *mdev;
353 unsigned int new_header_size;
357 /* Careful this entire function runs inside of an rcu critical section */
359 mdev = mpls_dev_get(dev);
363 MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets,
366 if (!mdev->input_enabled) {
367 MPLS_INC_STATS(mdev, rx_dropped);
371 if (skb->pkt_type != PACKET_HOST)
374 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
377 if (!pskb_may_pull(skb, sizeof(*hdr)))
382 /* Read and decode the label */
384 dec = mpls_entry_decode(hdr);
386 rt = mpls_route_input_rcu(net, dec.label);
388 MPLS_INC_STATS(mdev, rx_noroute);
392 nh = mpls_select_multipath(rt, skb);
397 skb_pull(skb, sizeof(*hdr));
398 skb_reset_network_header(skb);
402 if (skb_warn_if_lro(skb))
405 skb_forward_csum(skb);
407 /* Verify ttl is valid */
411 /* Find the output device */
412 out_dev = 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);
434 if (unlikely(!new_header_size && dec.bos)) {
435 /* Penultimate hop popping */
436 if (!mpls_egress(dev_net(out_dev), rt, skb, dec))
441 skb_push(skb, new_header_size);
442 skb_reset_network_header(skb);
443 /* Push the new labels */
446 for (i = nh->nh_labels - 1; i >= 0; i--) {
447 hdr[i] = mpls_entry_encode(nh->nh_label[i],
453 mpls_stats_inc_outucastpkts(out_dev, skb);
455 /* If via wasn't specified then send out using device address */
456 if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
457 err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,
458 out_dev->dev_addr, skb);
460 err = neigh_xmit(nh->nh_via_table, out_dev,
461 mpls_nh_via(rt, nh), skb);
463 net_dbg_ratelimited("%s: packet transmission failed: %d\n",
468 out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
470 MPLS_INC_STATS(out_mdev, tx_errors);
473 MPLS_INC_STATS(mdev, rx_errors);
479 static struct packet_type mpls_packet_type __read_mostly = {
480 .type = cpu_to_be16(ETH_P_MPLS_UC),
481 .func = mpls_forward,
484 static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = {
485 [RTA_DST] = { .type = NLA_U32 },
486 [RTA_OIF] = { .type = NLA_U32 },
487 [RTA_TTL_PROPAGATE] = { .type = NLA_U8 },
490 struct mpls_route_config {
495 u8 rc_via[MAX_VIA_ALEN];
499 u32 rc_output_label[MAX_NEW_LABELS];
501 enum mpls_payload_type rc_payload_type;
502 struct nl_info rc_nlinfo;
503 struct rtnexthop *rc_mp;
507 /* all nexthops within a route have the same size based on max
508 * number of labels and max via length for a hop
510 static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels)
512 u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen);
513 struct mpls_route *rt;
516 size = sizeof(*rt) + num_nh * nh_size;
517 if (size > MAX_MPLS_ROUTE_MEM)
518 return ERR_PTR(-EINVAL);
520 rt = kzalloc(size, GFP_KERNEL);
522 return ERR_PTR(-ENOMEM);
525 rt->rt_nhn_alive = num_nh;
526 rt->rt_nh_size = nh_size;
527 rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels);
532 static void mpls_rt_free(struct mpls_route *rt)
535 kfree_rcu(rt, rt_rcu);
538 static void mpls_notify_route(struct net *net, unsigned index,
539 struct mpls_route *old, struct mpls_route *new,
540 const struct nl_info *info)
542 struct nlmsghdr *nlh = info ? info->nlh : NULL;
543 unsigned portid = info ? info->portid : 0;
544 int event = new ? RTM_NEWROUTE : RTM_DELROUTE;
545 struct mpls_route *rt = new ? new : old;
546 unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0;
547 /* Ignore reserved labels for now */
548 if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED))
549 rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags);
552 static void mpls_route_update(struct net *net, unsigned index,
553 struct mpls_route *new,
554 const struct nl_info *info)
556 struct mpls_route __rcu **platform_label;
557 struct mpls_route *rt;
561 platform_label = rtnl_dereference(net->mpls.platform_label);
562 rt = rtnl_dereference(platform_label[index]);
563 rcu_assign_pointer(platform_label[index], new);
565 mpls_notify_route(net, index, rt, new, info);
567 /* If we removed a route free it now */
571 static unsigned find_free_label(struct net *net)
573 struct mpls_route __rcu **platform_label;
574 size_t platform_labels;
577 platform_label = rtnl_dereference(net->mpls.platform_label);
578 platform_labels = net->mpls.platform_labels;
579 for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels;
581 if (!rtnl_dereference(platform_label[index]))
584 return LABEL_NOT_SPECIFIED;
587 #if IS_ENABLED(CONFIG_INET)
588 static struct net_device *inet_fib_lookup_dev(struct net *net,
591 struct net_device *dev;
593 struct in_addr daddr;
595 memcpy(&daddr, addr, sizeof(struct in_addr));
596 rt = ip_route_output(net, daddr.s_addr, 0, 0, 0);
608 static struct net_device *inet_fib_lookup_dev(struct net *net,
611 return ERR_PTR(-EAFNOSUPPORT);
615 #if IS_ENABLED(CONFIG_IPV6)
616 static struct net_device *inet6_fib_lookup_dev(struct net *net,
619 struct net_device *dev;
620 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 dst = ipv6_stub->ipv6_dst_lookup_flow(net, NULL, &fl6, NULL);
630 return ERR_CAST(dst);
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))
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");
1022 rt = mpls_rt_alloc(nhs, max_via_alen, max_labels);
1028 rt->rt_protocol = cfg->rc_protocol;
1029 rt->rt_payload_type = cfg->rc_payload_type;
1030 rt->rt_ttl_propagate = cfg->rc_ttl_propagate;
1033 err = mpls_nh_build_multi(cfg, rt, max_labels, extack);
1035 err = mpls_nh_build_from_cfg(cfg, rt);
1039 mpls_route_update(net, index, rt, &cfg->rc_nlinfo);
1049 static int mpls_route_del(struct mpls_route_config *cfg,
1050 struct netlink_ext_ack *extack)
1052 struct net *net = cfg->rc_nlinfo.nl_net;
1056 index = cfg->rc_label;
1058 if (!mpls_label_ok(net, &index, extack))
1061 mpls_route_update(net, index, NULL, &cfg->rc_nlinfo);
1068 static void mpls_get_stats(struct mpls_dev *mdev,
1069 struct mpls_link_stats *stats)
1071 struct mpls_pcpu_stats *p;
1074 memset(stats, 0, sizeof(*stats));
1076 for_each_possible_cpu(i) {
1077 struct mpls_link_stats local;
1080 p = per_cpu_ptr(mdev->stats, i);
1082 start = u64_stats_fetch_begin_irq(&p->syncp);
1084 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
1086 stats->rx_packets += local.rx_packets;
1087 stats->rx_bytes += local.rx_bytes;
1088 stats->tx_packets += local.tx_packets;
1089 stats->tx_bytes += local.tx_bytes;
1090 stats->rx_errors += local.rx_errors;
1091 stats->tx_errors += local.tx_errors;
1092 stats->rx_dropped += local.rx_dropped;
1093 stats->tx_dropped += local.tx_dropped;
1094 stats->rx_noroute += local.rx_noroute;
1098 static int mpls_fill_stats_af(struct sk_buff *skb,
1099 const struct net_device *dev)
1101 struct mpls_link_stats *stats;
1102 struct mpls_dev *mdev;
1105 mdev = mpls_dev_get(dev);
1109 nla = nla_reserve_64bit(skb, MPLS_STATS_LINK,
1110 sizeof(struct mpls_link_stats),
1115 stats = nla_data(nla);
1116 mpls_get_stats(mdev, stats);
1121 static size_t mpls_get_stats_af_size(const struct net_device *dev)
1123 struct mpls_dev *mdev;
1125 mdev = mpls_dev_get(dev);
1129 return nla_total_size_64bit(sizeof(struct mpls_link_stats));
1132 static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev,
1133 u32 portid, u32 seq, int event,
1134 unsigned int flags, int type)
1136 struct nlmsghdr *nlh;
1137 struct netconfmsg *ncm;
1140 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1145 if (type == NETCONFA_ALL)
1148 ncm = nlmsg_data(nlh);
1149 ncm->ncm_family = AF_MPLS;
1151 if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0)
1152 goto nla_put_failure;
1154 if ((all || type == NETCONFA_INPUT) &&
1155 nla_put_s32(skb, NETCONFA_INPUT,
1156 mdev->input_enabled) < 0)
1157 goto nla_put_failure;
1159 nlmsg_end(skb, nlh);
1163 nlmsg_cancel(skb, nlh);
1167 static int mpls_netconf_msgsize_devconf(int type)
1169 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1170 + nla_total_size(4); /* NETCONFA_IFINDEX */
1173 if (type == NETCONFA_ALL)
1176 if (all || type == NETCONFA_INPUT)
1177 size += nla_total_size(4);
1182 static void mpls_netconf_notify_devconf(struct net *net, int event,
1183 int type, struct mpls_dev *mdev)
1185 struct sk_buff *skb;
1188 skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL);
1192 err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type);
1194 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1195 WARN_ON(err == -EMSGSIZE);
1200 rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL);
1204 rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err);
1207 static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = {
1208 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1211 static int mpls_netconf_valid_get_req(struct sk_buff *skb,
1212 const struct nlmsghdr *nlh,
1214 struct netlink_ext_ack *extack)
1218 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
1219 NL_SET_ERR_MSG_MOD(extack,
1220 "Invalid header for netconf get request");
1224 if (!netlink_strict_get_check(skb))
1225 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
1227 devconf_mpls_policy, extack);
1229 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
1231 devconf_mpls_policy, extack);
1235 for (i = 0; i <= NETCONFA_MAX; i++) {
1240 case NETCONFA_IFINDEX:
1243 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
1251 static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
1252 struct nlmsghdr *nlh,
1253 struct netlink_ext_ack *extack)
1255 struct net *net = sock_net(in_skb->sk);
1256 struct nlattr *tb[NETCONFA_MAX + 1];
1257 struct net_device *dev;
1258 struct mpls_dev *mdev;
1259 struct sk_buff *skb;
1263 err = mpls_netconf_valid_get_req(in_skb, nlh, tb, extack);
1268 if (!tb[NETCONFA_IFINDEX])
1271 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1272 dev = __dev_get_by_index(net, ifindex);
1276 mdev = mpls_dev_get(dev);
1281 skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
1285 err = mpls_netconf_fill_devconf(skb, mdev,
1286 NETLINK_CB(in_skb).portid,
1287 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1290 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1291 WARN_ON(err == -EMSGSIZE);
1295 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1300 static int mpls_netconf_dump_devconf(struct sk_buff *skb,
1301 struct netlink_callback *cb)
1303 const struct nlmsghdr *nlh = cb->nlh;
1304 struct net *net = sock_net(skb->sk);
1305 struct hlist_head *head;
1306 struct net_device *dev;
1307 struct mpls_dev *mdev;
1311 if (cb->strict_check) {
1312 struct netlink_ext_ack *extack = cb->extack;
1313 struct netconfmsg *ncm;
1315 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
1316 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
1320 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
1321 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
1327 s_idx = idx = cb->args[1];
1329 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1331 head = &net->dev_index_head[h];
1333 cb->seq = net->dev_base_seq;
1334 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1337 mdev = mpls_dev_get(dev);
1340 if (mpls_netconf_fill_devconf(skb, mdev,
1341 NETLINK_CB(cb->skb).portid,
1345 NETCONFA_ALL) < 0) {
1349 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1362 #define MPLS_PERDEV_SYSCTL_OFFSET(field) \
1363 (&((struct mpls_dev *)0)->field)
1365 static int mpls_conf_proc(struct ctl_table *ctl, int write,
1366 void *buffer, size_t *lenp, loff_t *ppos)
1368 int oval = *(int *)ctl->data;
1369 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1372 struct mpls_dev *mdev = ctl->extra1;
1373 int i = (int *)ctl->data - (int *)mdev;
1374 struct net *net = ctl->extra2;
1375 int val = *(int *)ctl->data;
1377 if (i == offsetof(struct mpls_dev, input_enabled) &&
1379 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF,
1380 NETCONFA_INPUT, mdev);
1387 static const struct ctl_table mpls_dev_table[] = {
1389 .procname = "input",
1390 .maxlen = sizeof(int),
1392 .proc_handler = mpls_conf_proc,
1393 .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled),
1398 static int mpls_dev_sysctl_register(struct net_device *dev,
1399 struct mpls_dev *mdev)
1401 char path[sizeof("net/mpls/conf/") + IFNAMSIZ];
1402 struct net *net = dev_net(dev);
1403 struct ctl_table *table;
1406 table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL);
1410 /* Table data contains only offsets relative to the base of
1411 * the mdev at this point, so make them absolute.
1413 for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) {
1414 table[i].data = (char *)mdev + (uintptr_t)table[i].data;
1415 table[i].extra1 = mdev;
1416 table[i].extra2 = net;
1419 snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name);
1421 mdev->sysctl = register_net_sysctl(net, path, table);
1425 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev);
1434 static void mpls_dev_sysctl_unregister(struct net_device *dev,
1435 struct mpls_dev *mdev)
1437 struct net *net = dev_net(dev);
1438 struct ctl_table *table;
1440 table = mdev->sysctl->ctl_table_arg;
1441 unregister_net_sysctl_table(mdev->sysctl);
1444 mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev);
1447 static struct mpls_dev *mpls_add_dev(struct net_device *dev)
1449 struct mpls_dev *mdev;
1455 mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
1457 return ERR_PTR(err);
1459 mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
1463 for_each_possible_cpu(i) {
1464 struct mpls_pcpu_stats *mpls_stats;
1466 mpls_stats = per_cpu_ptr(mdev->stats, i);
1467 u64_stats_init(&mpls_stats->syncp);
1472 err = mpls_dev_sysctl_register(dev, mdev);
1476 rcu_assign_pointer(dev->mpls_ptr, mdev);
1481 free_percpu(mdev->stats);
1483 return ERR_PTR(err);
1486 static void mpls_dev_destroy_rcu(struct rcu_head *head)
1488 struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
1490 free_percpu(mdev->stats);
1494 static int mpls_ifdown(struct net_device *dev, int event)
1496 struct mpls_route __rcu **platform_label;
1497 struct net *net = dev_net(dev);
1500 platform_label = rtnl_dereference(net->mpls.platform_label);
1501 for (index = 0; index < net->mpls.platform_labels; index++) {
1502 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1503 bool nh_del = false;
1509 if (event == NETDEV_UNREGISTER) {
1513 if (!nh->nh_dev || nh->nh_dev == dev)
1515 if (nh->nh_dev == dev)
1517 } endfor_nexthops(rt);
1519 /* if there are no more nexthops, delete the route */
1520 if (deleted == rt->rt_nhn) {
1521 mpls_route_update(net, index, NULL, NULL);
1526 size_t size = sizeof(*rt) + rt->rt_nhn *
1528 struct mpls_route *orig = rt;
1530 rt = kmemdup(orig, size, GFP_KERNEL);
1536 change_nexthops(rt) {
1537 unsigned int nh_flags = nh->nh_flags;
1539 if (nh->nh_dev != dev)
1544 case NETDEV_UNREGISTER:
1545 nh_flags |= RTNH_F_DEAD;
1548 nh_flags |= RTNH_F_LINKDOWN;
1551 if (event == NETDEV_UNREGISTER)
1554 if (nh->nh_flags != nh_flags)
1555 WRITE_ONCE(nh->nh_flags, nh_flags);
1557 if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)))
1559 } endfor_nexthops(rt);
1561 WRITE_ONCE(rt->rt_nhn_alive, alive);
1564 mpls_route_update(net, index, rt, NULL);
1570 static void mpls_ifup(struct net_device *dev, unsigned int flags)
1572 struct mpls_route __rcu **platform_label;
1573 struct net *net = dev_net(dev);
1577 platform_label = rtnl_dereference(net->mpls.platform_label);
1578 for (index = 0; index < net->mpls.platform_labels; index++) {
1579 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1585 change_nexthops(rt) {
1586 unsigned int nh_flags = nh->nh_flags;
1588 if (!(nh_flags & flags)) {
1592 if (nh->nh_dev != dev)
1596 WRITE_ONCE(nh->nh_flags, nh_flags);
1597 } endfor_nexthops(rt);
1599 WRITE_ONCE(rt->rt_nhn_alive, alive);
1603 static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
1606 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1607 struct mpls_dev *mdev;
1611 if (event == NETDEV_REGISTER) {
1612 mdev = mpls_add_dev(dev);
1614 return notifier_from_errno(PTR_ERR(mdev));
1619 mdev = mpls_dev_get(dev);
1626 err = mpls_ifdown(dev, event);
1628 return notifier_from_errno(err);
1631 flags = dev_get_flags(dev);
1632 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1633 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1635 mpls_ifup(dev, RTNH_F_DEAD);
1638 flags = dev_get_flags(dev);
1639 if (flags & (IFF_RUNNING | IFF_LOWER_UP)) {
1640 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1642 err = mpls_ifdown(dev, event);
1644 return notifier_from_errno(err);
1647 case NETDEV_UNREGISTER:
1648 err = mpls_ifdown(dev, event);
1650 return notifier_from_errno(err);
1651 mdev = mpls_dev_get(dev);
1653 mpls_dev_sysctl_unregister(dev, mdev);
1654 RCU_INIT_POINTER(dev->mpls_ptr, NULL);
1655 call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
1658 case NETDEV_CHANGENAME:
1659 mdev = mpls_dev_get(dev);
1661 mpls_dev_sysctl_unregister(dev, mdev);
1662 err = mpls_dev_sysctl_register(dev, mdev);
1664 return notifier_from_errno(err);
1671 static struct notifier_block mpls_dev_notifier = {
1672 .notifier_call = mpls_dev_notify,
1675 static int nla_put_via(struct sk_buff *skb,
1676 u8 table, const void *addr, int alen)
1678 static const int table_to_family[NEIGH_NR_TABLES + 1] = {
1679 AF_INET, AF_INET6, AF_DECnet, AF_PACKET,
1683 int family = AF_UNSPEC;
1685 nla = nla_reserve(skb, RTA_VIA, alen + 2);
1689 if (table <= NEIGH_NR_TABLES)
1690 family = table_to_family[table];
1692 via = nla_data(nla);
1693 via->rtvia_family = family;
1694 memcpy(via->rtvia_addr, addr, alen);
1698 int nla_put_labels(struct sk_buff *skb, int attrtype,
1699 u8 labels, const u32 label[])
1702 struct mpls_shim_hdr *nla_label;
1705 nla = nla_reserve(skb, attrtype, labels*4);
1709 nla_label = nla_data(nla);
1711 for (i = labels - 1; i >= 0; i--) {
1712 nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos);
1718 EXPORT_SYMBOL_GPL(nla_put_labels);
1720 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
1721 u32 label[], struct netlink_ext_ack *extack)
1723 unsigned len = nla_len(nla);
1724 struct mpls_shim_hdr *nla_label;
1729 /* len needs to be an even multiple of 4 (the label size). Number
1730 * of labels is a u8 so check for overflow.
1732 if (len & 3 || len / 4 > 255) {
1733 NL_SET_ERR_MSG_ATTR(extack, nla,
1734 "Invalid length for labels attribute");
1738 /* Limit the number of new labels allowed */
1740 if (nla_labels > max_labels) {
1741 NL_SET_ERR_MSG(extack, "Too many labels");
1745 /* when label == NULL, caller wants number of labels */
1749 nla_label = nla_data(nla);
1751 for (i = nla_labels - 1; i >= 0; i--, bos = false) {
1752 struct mpls_entry_decoded dec;
1753 dec = mpls_entry_decode(nla_label + i);
1755 /* Ensure the bottom of stack flag is properly set
1756 * and ttl and tc are both clear.
1759 NL_SET_ERR_MSG_ATTR(extack, nla,
1760 "TTL in label must be 0");
1765 NL_SET_ERR_MSG_ATTR(extack, nla,
1766 "Traffic class in label must be 0");
1770 if (dec.bos != bos) {
1771 NL_SET_BAD_ATTR(extack, nla);
1773 NL_SET_ERR_MSG(extack,
1774 "BOS bit must be set in first label");
1776 NL_SET_ERR_MSG(extack,
1777 "BOS bit can only be set in first label");
1782 switch (dec.label) {
1783 case MPLS_LABEL_IMPLNULL:
1784 /* RFC3032: This is a label that an LSR may
1785 * assign and distribute, but which never
1786 * actually appears in the encapsulation.
1788 NL_SET_ERR_MSG_ATTR(extack, nla,
1789 "Implicit NULL Label (3) can not be used in encapsulation");
1793 label[i] = dec.label;
1796 *labels = nla_labels;
1799 EXPORT_SYMBOL_GPL(nla_get_labels);
1801 static int rtm_to_route_config(struct sk_buff *skb,
1802 struct nlmsghdr *nlh,
1803 struct mpls_route_config *cfg,
1804 struct netlink_ext_ack *extack)
1807 struct nlattr *tb[RTA_MAX+1];
1811 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
1812 rtm_mpls_policy, extack);
1817 rtm = nlmsg_data(nlh);
1819 if (rtm->rtm_family != AF_MPLS) {
1820 NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg");
1823 if (rtm->rtm_dst_len != 20) {
1824 NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS");
1827 if (rtm->rtm_src_len != 0) {
1828 NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS");
1831 if (rtm->rtm_tos != 0) {
1832 NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS");
1835 if (rtm->rtm_table != RT_TABLE_MAIN) {
1836 NL_SET_ERR_MSG(extack,
1837 "MPLS only supports the main route table");
1840 /* Any value is acceptable for rtm_protocol */
1842 /* As mpls uses destination specific addresses
1843 * (or source specific address in the case of multicast)
1844 * all addresses have universal scope.
1846 if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) {
1847 NL_SET_ERR_MSG(extack,
1848 "Invalid route scope - MPLS only supports UNIVERSE");
1851 if (rtm->rtm_type != RTN_UNICAST) {
1852 NL_SET_ERR_MSG(extack,
1853 "Invalid route type - MPLS only supports UNICAST");
1856 if (rtm->rtm_flags != 0) {
1857 NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS");
1861 cfg->rc_label = LABEL_NOT_SPECIFIED;
1862 cfg->rc_protocol = rtm->rtm_protocol;
1863 cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC;
1864 cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
1865 cfg->rc_nlflags = nlh->nlmsg_flags;
1866 cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid;
1867 cfg->rc_nlinfo.nlh = nlh;
1868 cfg->rc_nlinfo.nl_net = sock_net(skb->sk);
1870 for (index = 0; index <= RTA_MAX; index++) {
1871 struct nlattr *nla = tb[index];
1877 cfg->rc_ifindex = nla_get_u32(nla);
1880 if (nla_get_labels(nla, MAX_NEW_LABELS,
1881 &cfg->rc_output_labels,
1882 cfg->rc_output_label, extack))
1888 if (nla_get_labels(nla, 1, &label_count,
1889 &cfg->rc_label, extack))
1892 if (!mpls_label_ok(cfg->rc_nlinfo.nl_net,
1893 &cfg->rc_label, extack))
1898 NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute");
1902 if (nla_get_via(nla, &cfg->rc_via_alen,
1903 &cfg->rc_via_table, cfg->rc_via,
1910 cfg->rc_mp = nla_data(nla);
1911 cfg->rc_mp_len = nla_len(nla);
1914 case RTA_TTL_PROPAGATE:
1916 u8 ttl_propagate = nla_get_u8(nla);
1918 if (ttl_propagate > 1) {
1919 NL_SET_ERR_MSG_ATTR(extack, nla,
1920 "RTA_TTL_PROPAGATE can only be 0 or 1");
1923 cfg->rc_ttl_propagate = ttl_propagate ?
1924 MPLS_TTL_PROP_ENABLED :
1925 MPLS_TTL_PROP_DISABLED;
1929 NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute");
1930 /* Unsupported attribute */
1940 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1941 struct netlink_ext_ack *extack)
1943 struct mpls_route_config *cfg;
1946 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1950 err = rtm_to_route_config(skb, nlh, cfg, extack);
1954 err = mpls_route_del(cfg, extack);
1962 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1963 struct netlink_ext_ack *extack)
1965 struct mpls_route_config *cfg;
1968 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1972 err = rtm_to_route_config(skb, nlh, cfg, extack);
1976 err = mpls_route_add(cfg, extack);
1983 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event,
1984 u32 label, struct mpls_route *rt, int flags)
1986 struct net_device *dev;
1987 struct nlmsghdr *nlh;
1990 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1994 rtm = nlmsg_data(nlh);
1995 rtm->rtm_family = AF_MPLS;
1996 rtm->rtm_dst_len = 20;
1997 rtm->rtm_src_len = 0;
1999 rtm->rtm_table = RT_TABLE_MAIN;
2000 rtm->rtm_protocol = rt->rt_protocol;
2001 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2002 rtm->rtm_type = RTN_UNICAST;
2005 if (nla_put_labels(skb, RTA_DST, 1, &label))
2006 goto nla_put_failure;
2008 if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) {
2009 bool ttl_propagate =
2010 rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED;
2012 if (nla_put_u8(skb, RTA_TTL_PROPAGATE,
2014 goto nla_put_failure;
2016 if (rt->rt_nhn == 1) {
2017 const struct mpls_nh *nh = rt->rt_nh;
2019 if (nh->nh_labels &&
2020 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2022 goto nla_put_failure;
2023 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2024 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2026 goto nla_put_failure;
2028 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2029 goto nla_put_failure;
2030 if (nh->nh_flags & RTNH_F_LINKDOWN)
2031 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2032 if (nh->nh_flags & RTNH_F_DEAD)
2033 rtm->rtm_flags |= RTNH_F_DEAD;
2035 struct rtnexthop *rtnh;
2040 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
2042 goto nla_put_failure;
2049 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
2051 goto nla_put_failure;
2053 rtnh->rtnh_ifindex = dev->ifindex;
2054 if (nh->nh_flags & RTNH_F_LINKDOWN) {
2055 rtnh->rtnh_flags |= RTNH_F_LINKDOWN;
2058 if (nh->nh_flags & RTNH_F_DEAD) {
2059 rtnh->rtnh_flags |= RTNH_F_DEAD;
2063 if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST,
2066 goto nla_put_failure;
2067 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2068 nla_put_via(skb, nh->nh_via_table,
2069 mpls_nh_via(rt, nh),
2071 goto nla_put_failure;
2073 /* length of rtnetlink header + attributes */
2074 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
2075 } endfor_nexthops(rt);
2077 if (linkdown == rt->rt_nhn)
2078 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2079 if (dead == rt->rt_nhn)
2080 rtm->rtm_flags |= RTNH_F_DEAD;
2082 nla_nest_end(skb, mp);
2085 nlmsg_end(skb, nlh);
2089 nlmsg_cancel(skb, nlh);
2093 #if IS_ENABLED(CONFIG_INET)
2094 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2095 struct fib_dump_filter *filter,
2096 struct netlink_callback *cb)
2098 return ip_valid_fib_dump_req(net, nlh, filter, cb);
2101 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2102 struct fib_dump_filter *filter,
2103 struct netlink_callback *cb)
2105 struct netlink_ext_ack *extack = cb->extack;
2106 struct nlattr *tb[RTA_MAX + 1];
2110 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2111 NL_SET_ERR_MSG_MOD(extack, "Invalid header for FIB dump request");
2115 rtm = nlmsg_data(nlh);
2116 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos ||
2117 rtm->rtm_table || rtm->rtm_scope || rtm->rtm_type ||
2119 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for FIB dump request");
2123 if (rtm->rtm_protocol) {
2124 filter->protocol = rtm->rtm_protocol;
2125 filter->filter_set = 1;
2126 cb->answer_flags = NLM_F_DUMP_FILTERED;
2129 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2130 rtm_mpls_policy, extack);
2134 for (i = 0; i <= RTA_MAX; ++i) {
2138 ifindex = nla_get_u32(tb[i]);
2139 filter->dev = __dev_get_by_index(net, ifindex);
2142 filter->filter_set = 1;
2144 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
2153 static bool mpls_rt_uses_dev(struct mpls_route *rt,
2154 const struct net_device *dev)
2156 if (rt->rt_nhn == 1) {
2157 struct mpls_nh *nh = rt->rt_nh;
2159 if (nh->nh_dev == dev)
2163 if (nh->nh_dev == dev)
2165 } endfor_nexthops(rt);
2171 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
2173 const struct nlmsghdr *nlh = cb->nlh;
2174 struct net *net = sock_net(skb->sk);
2175 struct mpls_route __rcu **platform_label;
2176 struct fib_dump_filter filter = {};
2177 unsigned int flags = NLM_F_MULTI;
2178 size_t platform_labels;
2183 if (cb->strict_check) {
2186 err = mpls_valid_fib_dump_req(net, nlh, &filter, cb);
2190 /* for MPLS, there is only 1 table with fixed type and flags.
2191 * If either are set in the filter then return nothing.
2193 if ((filter.table_id && filter.table_id != RT_TABLE_MAIN) ||
2194 (filter.rt_type && filter.rt_type != RTN_UNICAST) ||
2199 index = cb->args[0];
2200 if (index < MPLS_LABEL_FIRST_UNRESERVED)
2201 index = MPLS_LABEL_FIRST_UNRESERVED;
2203 platform_label = rtnl_dereference(net->mpls.platform_label);
2204 platform_labels = net->mpls.platform_labels;
2206 if (filter.filter_set)
2207 flags |= NLM_F_DUMP_FILTERED;
2209 for (; index < platform_labels; index++) {
2210 struct mpls_route *rt;
2212 rt = rtnl_dereference(platform_label[index]);
2216 if ((filter.dev && !mpls_rt_uses_dev(rt, filter.dev)) ||
2217 (filter.protocol && rt->rt_protocol != filter.protocol))
2220 if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid,
2221 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2222 index, rt, flags) < 0)
2225 cb->args[0] = index;
2230 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
2233 NLMSG_ALIGN(sizeof(struct rtmsg))
2234 + nla_total_size(4) /* RTA_DST */
2235 + nla_total_size(1); /* RTA_TTL_PROPAGATE */
2237 if (rt->rt_nhn == 1) {
2238 struct mpls_nh *nh = rt->rt_nh;
2241 payload += nla_total_size(4); /* RTA_OIF */
2242 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */
2243 payload += nla_total_size(2 + nh->nh_via_alen);
2244 if (nh->nh_labels) /* RTA_NEWDST */
2245 payload += nla_total_size(nh->nh_labels * 4);
2247 /* each nexthop is packed in an attribute */
2253 nhsize += nla_total_size(sizeof(struct rtnexthop));
2255 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC)
2256 nhsize += nla_total_size(2 + nh->nh_via_alen);
2258 nhsize += nla_total_size(nh->nh_labels * 4);
2259 } endfor_nexthops(rt);
2260 /* nested attribute */
2261 payload += nla_total_size(nhsize);
2267 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
2268 struct nlmsghdr *nlh, struct net *net, u32 portid,
2269 unsigned int nlm_flags)
2271 struct sk_buff *skb;
2272 u32 seq = nlh ? nlh->nlmsg_seq : 0;
2275 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2279 err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags);
2281 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2282 WARN_ON(err == -EMSGSIZE);
2286 rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL);
2291 rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err);
2294 static int mpls_valid_getroute_req(struct sk_buff *skb,
2295 const struct nlmsghdr *nlh,
2297 struct netlink_ext_ack *extack)
2302 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
2303 NL_SET_ERR_MSG_MOD(extack,
2304 "Invalid header for get route request");
2308 if (!netlink_strict_get_check(skb))
2309 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2310 rtm_mpls_policy, extack);
2312 rtm = nlmsg_data(nlh);
2313 if ((rtm->rtm_dst_len && rtm->rtm_dst_len != 20) ||
2314 rtm->rtm_src_len || rtm->rtm_tos || rtm->rtm_table ||
2315 rtm->rtm_protocol || rtm->rtm_scope || rtm->rtm_type) {
2316 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
2319 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
2320 NL_SET_ERR_MSG_MOD(extack,
2321 "Invalid flags for get route request");
2325 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2326 rtm_mpls_policy, extack);
2330 if ((tb[RTA_DST] || tb[RTA_NEWDST]) && !rtm->rtm_dst_len) {
2331 NL_SET_ERR_MSG_MOD(extack, "rtm_dst_len must be 20 for MPLS");
2335 for (i = 0; i <= RTA_MAX; i++) {
2344 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
2352 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
2353 struct netlink_ext_ack *extack)
2355 struct net *net = sock_net(in_skb->sk);
2356 u32 portid = NETLINK_CB(in_skb).portid;
2357 u32 in_label = LABEL_NOT_SPECIFIED;
2358 struct nlattr *tb[RTA_MAX + 1];
2359 u32 labels[MAX_NEW_LABELS];
2360 struct mpls_shim_hdr *hdr;
2361 unsigned int hdr_size = 0;
2362 const struct mpls_nh *nh;
2363 struct net_device *dev;
2364 struct mpls_route *rt;
2365 struct rtmsg *rtm, *r;
2366 struct nlmsghdr *nlh;
2367 struct sk_buff *skb;
2371 err = mpls_valid_getroute_req(in_skb, in_nlh, tb, extack);
2375 rtm = nlmsg_data(in_nlh);
2380 if (nla_get_labels(tb[RTA_DST], 1, &label_count,
2381 &in_label, extack)) {
2386 if (!mpls_label_ok(net, &in_label, extack)) {
2392 rt = mpls_route_input_rcu(net, in_label);
2398 if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
2399 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2405 err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq,
2406 RTM_NEWROUTE, in_label, rt, 0);
2408 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2409 WARN_ON(err == -EMSGSIZE);
2413 return rtnl_unicast(skb, net, portid);
2416 if (tb[RTA_NEWDST]) {
2417 if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels,
2418 labels, extack) != 0) {
2423 hdr_size = n_labels * sizeof(struct mpls_shim_hdr);
2426 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2432 skb->protocol = htons(ETH_P_MPLS_UC);
2438 if (skb_cow(skb, hdr_size)) {
2443 skb_reserve(skb, hdr_size);
2444 skb_push(skb, hdr_size);
2445 skb_reset_network_header(skb);
2447 /* Push new labels */
2448 hdr = mpls_hdr(skb);
2450 for (i = n_labels - 1; i >= 0; i--) {
2451 hdr[i] = mpls_entry_encode(labels[i],
2457 nh = mpls_select_multipath(rt, skb);
2464 skb_pull(skb, hdr_size);
2465 skb_reset_network_header(skb);
2468 nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq,
2469 RTM_NEWROUTE, sizeof(*r), 0);
2475 r = nlmsg_data(nlh);
2476 r->rtm_family = AF_MPLS;
2477 r->rtm_dst_len = 20;
2479 r->rtm_table = RT_TABLE_MAIN;
2480 r->rtm_type = RTN_UNICAST;
2481 r->rtm_scope = RT_SCOPE_UNIVERSE;
2482 r->rtm_protocol = rt->rt_protocol;
2485 if (nla_put_labels(skb, RTA_DST, 1, &in_label))
2486 goto nla_put_failure;
2488 if (nh->nh_labels &&
2489 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2491 goto nla_put_failure;
2493 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2494 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2496 goto nla_put_failure;
2498 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2499 goto nla_put_failure;
2501 nlmsg_end(skb, nlh);
2503 err = rtnl_unicast(skb, net, portid);
2508 nlmsg_cancel(skb, nlh);
2515 static int resize_platform_label_table(struct net *net, size_t limit)
2517 size_t size = sizeof(struct mpls_route *) * limit;
2520 struct mpls_route __rcu **labels = NULL, **old;
2521 struct mpls_route *rt0 = NULL, *rt2 = NULL;
2525 labels = kvzalloc(size, GFP_KERNEL);
2530 /* In case the predefined labels need to be populated */
2531 if (limit > MPLS_LABEL_IPV4NULL) {
2532 struct net_device *lo = net->loopback_dev;
2533 rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
2536 rt0->rt_nh->nh_dev = lo;
2537 rt0->rt_protocol = RTPROT_KERNEL;
2538 rt0->rt_payload_type = MPT_IPV4;
2539 rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2540 rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2541 rt0->rt_nh->nh_via_alen = lo->addr_len;
2542 memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr,
2545 if (limit > MPLS_LABEL_IPV6NULL) {
2546 struct net_device *lo = net->loopback_dev;
2547 rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
2550 rt2->rt_nh->nh_dev = lo;
2551 rt2->rt_protocol = RTPROT_KERNEL;
2552 rt2->rt_payload_type = MPT_IPV6;
2553 rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2554 rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2555 rt2->rt_nh->nh_via_alen = lo->addr_len;
2556 memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr,
2561 /* Remember the original table */
2562 old = rtnl_dereference(net->mpls.platform_label);
2563 old_limit = net->mpls.platform_labels;
2565 /* Free any labels beyond the new table */
2566 for (index = limit; index < old_limit; index++)
2567 mpls_route_update(net, index, NULL, NULL);
2569 /* Copy over the old labels */
2571 if (old_limit < limit)
2572 cp_size = old_limit * sizeof(struct mpls_route *);
2574 memcpy(labels, old, cp_size);
2576 /* If needed set the predefined labels */
2577 if ((old_limit <= MPLS_LABEL_IPV6NULL) &&
2578 (limit > MPLS_LABEL_IPV6NULL)) {
2579 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2);
2583 if ((old_limit <= MPLS_LABEL_IPV4NULL) &&
2584 (limit > MPLS_LABEL_IPV4NULL)) {
2585 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0);
2589 /* Update the global pointers */
2590 net->mpls.platform_labels = limit;
2591 rcu_assign_pointer(net->mpls.platform_label, labels);
2612 static int mpls_platform_labels(struct ctl_table *table, int write,
2613 void *buffer, size_t *lenp, loff_t *ppos)
2615 struct net *net = table->data;
2616 int platform_labels = net->mpls.platform_labels;
2618 struct ctl_table tmp = {
2619 .procname = table->procname,
2620 .data = &platform_labels,
2621 .maxlen = sizeof(int),
2622 .mode = table->mode,
2623 .extra1 = SYSCTL_ZERO,
2624 .extra2 = &label_limit,
2627 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2629 if (write && ret == 0)
2630 ret = resize_platform_label_table(net, platform_labels);
2635 #define MPLS_NS_SYSCTL_OFFSET(field) \
2636 (&((struct net *)0)->field)
2638 static const struct ctl_table mpls_table[] = {
2640 .procname = "platform_labels",
2642 .maxlen = sizeof(int),
2644 .proc_handler = mpls_platform_labels,
2647 .procname = "ip_ttl_propagate",
2648 .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate),
2649 .maxlen = sizeof(int),
2651 .proc_handler = proc_dointvec_minmax,
2652 .extra1 = SYSCTL_ZERO,
2653 .extra2 = SYSCTL_ONE,
2656 .procname = "default_ttl",
2657 .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl),
2658 .maxlen = sizeof(int),
2660 .proc_handler = proc_dointvec_minmax,
2661 .extra1 = SYSCTL_ONE,
2667 static int mpls_net_init(struct net *net)
2669 struct ctl_table *table;
2672 net->mpls.platform_labels = 0;
2673 net->mpls.platform_label = NULL;
2674 net->mpls.ip_ttl_propagate = 1;
2675 net->mpls.default_ttl = 255;
2677 table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL);
2681 /* Table data contains only offsets relative to the base of
2682 * the mdev at this point, so make them absolute.
2684 for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++)
2685 table[i].data = (char *)net + (uintptr_t)table[i].data;
2687 net->mpls.ctl = register_net_sysctl(net, "net/mpls", table);
2688 if (net->mpls.ctl == NULL) {
2696 static void mpls_net_exit(struct net *net)
2698 struct mpls_route __rcu **platform_label;
2699 size_t platform_labels;
2700 struct ctl_table *table;
2703 table = net->mpls.ctl->ctl_table_arg;
2704 unregister_net_sysctl_table(net->mpls.ctl);
2707 /* An rcu grace period has passed since there was a device in
2708 * the network namespace (and thus the last in flight packet)
2709 * left this network namespace. This is because
2710 * unregister_netdevice_many and netdev_run_todo has completed
2711 * for each network device that was in this network namespace.
2713 * As such no additional rcu synchronization is necessary when
2714 * freeing the platform_label table.
2717 platform_label = rtnl_dereference(net->mpls.platform_label);
2718 platform_labels = net->mpls.platform_labels;
2719 for (index = 0; index < platform_labels; index++) {
2720 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
2721 RCU_INIT_POINTER(platform_label[index], NULL);
2722 mpls_notify_route(net, index, rt, NULL, NULL);
2727 kvfree(platform_label);
2730 static struct pernet_operations mpls_net_ops = {
2731 .init = mpls_net_init,
2732 .exit = mpls_net_exit,
2735 static struct rtnl_af_ops mpls_af_ops __read_mostly = {
2737 .fill_stats_af = mpls_fill_stats_af,
2738 .get_stats_af_size = mpls_get_stats_af_size,
2741 static int __init mpls_init(void)
2745 BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4);
2747 err = register_pernet_subsys(&mpls_net_ops);
2751 err = register_netdevice_notifier(&mpls_dev_notifier);
2753 goto out_unregister_pernet;
2755 dev_add_pack(&mpls_packet_type);
2757 rtnl_af_register(&mpls_af_ops);
2759 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_NEWROUTE,
2760 mpls_rtm_newroute, NULL, 0);
2761 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_DELROUTE,
2762 mpls_rtm_delroute, NULL, 0);
2763 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETROUTE,
2764 mpls_getroute, mpls_dump_routes, 0);
2765 rtnl_register_module(THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
2766 mpls_netconf_get_devconf,
2767 mpls_netconf_dump_devconf, 0);
2768 err = ipgre_tunnel_encap_add_mpls_ops();
2770 pr_err("Can't add mpls over gre tunnel ops\n");
2776 out_unregister_pernet:
2777 unregister_pernet_subsys(&mpls_net_ops);
2780 module_init(mpls_init);
2782 static void __exit mpls_exit(void)
2784 rtnl_unregister_all(PF_MPLS);
2785 rtnl_af_unregister(&mpls_af_ops);
2786 dev_remove_pack(&mpls_packet_type);
2787 unregister_netdevice_notifier(&mpls_dev_notifier);
2788 unregister_pernet_subsys(&mpls_net_ops);
2789 ipgre_tunnel_encap_del_mpls_ops();
2791 module_exit(mpls_exit);
2793 MODULE_DESCRIPTION("MultiProtocol Label Switching");
2794 MODULE_LICENSE("GPL v2");
2795 MODULE_ALIAS_NETPROTO(PF_MPLS);