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>
20 #include <net/ip_fib.h>
21 #include <net/netevent.h>
22 #include <net/ip_tunnels.h>
23 #include <net/netns/generic.h>
24 #if IS_ENABLED(CONFIG_IPV6)
27 #include <net/ipv6_stubs.h>
31 /* max memory we will use for mpls_route */
32 #define MAX_MPLS_ROUTE_MEM 4096
34 /* Maximum number of labels to look ahead at when selecting a path of
37 #define MAX_MP_SELECT_LABELS 4
39 #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_rtnl(net->mpls.platform_label);
85 rt = rcu_dereference_rtnl(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 const 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 const struct mpls_nh *nh;
348 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)))
383 /* Read and decode the label */
385 dec = mpls_entry_decode(hdr);
387 rt = mpls_route_input_rcu(net, dec.label);
389 MPLS_INC_STATS(mdev, rx_noroute);
393 nh = mpls_select_multipath(rt, skb);
398 skb_pull(skb, sizeof(*hdr));
399 skb_reset_network_header(skb);
403 if (skb_warn_if_lro(skb))
406 skb_forward_csum(skb);
408 /* Verify ttl is valid */
412 /* Find the output device */
413 out_dev = nh->nh_dev;
414 if (!mpls_output_possible(out_dev))
417 /* Verify the destination can hold the packet */
418 new_header_size = mpls_nh_header_size(nh);
419 mtu = mpls_dev_mtu(out_dev);
420 if (mpls_pkt_too_big(skb, mtu - new_header_size))
423 hh_len = LL_RESERVED_SPACE(out_dev);
424 if (!out_dev->header_ops)
427 /* Ensure there is enough space for the headers in the skb */
428 if (skb_cow(skb, hh_len + new_header_size))
432 skb->protocol = htons(ETH_P_MPLS_UC);
435 if (unlikely(!new_header_size && dec.bos)) {
436 /* Penultimate hop popping */
437 if (!mpls_egress(dev_net(out_dev), rt, skb, dec))
442 skb_push(skb, new_header_size);
443 skb_reset_network_header(skb);
444 /* Push the new labels */
447 for (i = nh->nh_labels - 1; i >= 0; i--) {
448 hdr[i] = mpls_entry_encode(nh->nh_label[i],
454 mpls_stats_inc_outucastpkts(out_dev, skb);
456 /* If via wasn't specified then send out using device address */
457 if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
458 err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,
459 out_dev->dev_addr, skb);
461 err = neigh_xmit(nh->nh_via_table, out_dev,
462 mpls_nh_via(rt, nh), skb);
464 net_dbg_ratelimited("%s: packet transmission failed: %d\n",
469 out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
471 MPLS_INC_STATS(out_mdev, tx_errors);
474 MPLS_INC_STATS(mdev, rx_errors);
480 static struct packet_type mpls_packet_type __read_mostly = {
481 .type = cpu_to_be16(ETH_P_MPLS_UC),
482 .func = mpls_forward,
485 static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = {
486 [RTA_DST] = { .type = NLA_U32 },
487 [RTA_OIF] = { .type = NLA_U32 },
488 [RTA_TTL_PROPAGATE] = { .type = NLA_U8 },
491 struct mpls_route_config {
496 u8 rc_via[MAX_VIA_ALEN];
500 u32 rc_output_label[MAX_NEW_LABELS];
502 enum mpls_payload_type rc_payload_type;
503 struct nl_info rc_nlinfo;
504 struct rtnexthop *rc_mp;
508 /* all nexthops within a route have the same size based on max
509 * number of labels and max via length for a hop
511 static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels)
513 u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen);
514 struct mpls_route *rt;
517 size = sizeof(*rt) + num_nh * nh_size;
518 if (size > MAX_MPLS_ROUTE_MEM)
519 return ERR_PTR(-EINVAL);
521 rt = kzalloc(size, GFP_KERNEL);
523 return ERR_PTR(-ENOMEM);
526 rt->rt_nhn_alive = num_nh;
527 rt->rt_nh_size = nh_size;
528 rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels);
533 static void mpls_rt_free(struct mpls_route *rt)
536 kfree_rcu(rt, rt_rcu);
539 static void mpls_notify_route(struct net *net, unsigned index,
540 struct mpls_route *old, struct mpls_route *new,
541 const struct nl_info *info)
543 struct nlmsghdr *nlh = info ? info->nlh : NULL;
544 unsigned portid = info ? info->portid : 0;
545 int event = new ? RTM_NEWROUTE : RTM_DELROUTE;
546 struct mpls_route *rt = new ? new : old;
547 unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0;
548 /* Ignore reserved labels for now */
549 if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED))
550 rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags);
553 static void mpls_route_update(struct net *net, unsigned index,
554 struct mpls_route *new,
555 const struct nl_info *info)
557 struct mpls_route __rcu **platform_label;
558 struct mpls_route *rt;
562 platform_label = rtnl_dereference(net->mpls.platform_label);
563 rt = rtnl_dereference(platform_label[index]);
564 rcu_assign_pointer(platform_label[index], new);
566 mpls_notify_route(net, index, rt, new, info);
568 /* If we removed a route free it now */
572 static unsigned find_free_label(struct net *net)
574 struct mpls_route __rcu **platform_label;
575 size_t platform_labels;
578 platform_label = rtnl_dereference(net->mpls.platform_label);
579 platform_labels = net->mpls.platform_labels;
580 for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels;
582 if (!rtnl_dereference(platform_label[index]))
585 return LABEL_NOT_SPECIFIED;
588 #if IS_ENABLED(CONFIG_INET)
589 static struct net_device *inet_fib_lookup_dev(struct net *net,
592 struct net_device *dev;
594 struct in_addr daddr;
596 memcpy(&daddr, addr, sizeof(struct in_addr));
597 rt = ip_route_output(net, daddr.s_addr, 0, 0, 0, RT_SCOPE_UNIVERSE);
609 static struct net_device *inet_fib_lookup_dev(struct net *net,
612 return ERR_PTR(-EAFNOSUPPORT);
616 #if IS_ENABLED(CONFIG_IPV6)
617 static struct net_device *inet6_fib_lookup_dev(struct net *net,
620 struct net_device *dev;
621 struct dst_entry *dst;
625 return ERR_PTR(-EAFNOSUPPORT);
627 memset(&fl6, 0, sizeof(fl6));
628 memcpy(&fl6.daddr, addr, sizeof(struct in6_addr));
629 dst = ipv6_stub->ipv6_dst_lookup_flow(net, NULL, &fl6, NULL);
631 return ERR_CAST(dst);
640 static struct net_device *inet6_fib_lookup_dev(struct net *net,
643 return ERR_PTR(-EAFNOSUPPORT);
647 static struct net_device *find_outdev(struct net *net,
648 struct mpls_route *rt,
649 struct mpls_nh *nh, int oif)
651 struct net_device *dev = NULL;
654 switch (nh->nh_via_table) {
655 case NEIGH_ARP_TABLE:
656 dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh));
659 dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh));
661 case NEIGH_LINK_TABLE:
665 dev = dev_get_by_index(net, oif);
669 return ERR_PTR(-ENODEV);
674 /* The caller is holding rtnl anyways, so release the dev reference */
680 static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
681 struct mpls_nh *nh, int oif)
683 struct net_device *dev = NULL;
686 dev = find_outdev(net, rt, nh, oif);
693 /* Ensure this is a supported device */
695 if (!mpls_dev_get(dev))
698 if ((nh->nh_via_table == NEIGH_LINK_TABLE) &&
699 (dev->addr_len != nh->nh_via_alen))
704 if (!(dev->flags & IFF_UP)) {
705 nh->nh_flags |= RTNH_F_DEAD;
709 flags = dev_get_flags(dev);
710 if (!(flags & (IFF_RUNNING | IFF_LOWER_UP)))
711 nh->nh_flags |= RTNH_F_LINKDOWN;
720 static int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table,
721 u8 via_addr[], struct netlink_ext_ack *extack)
723 struct rtvia *via = nla_data(nla);
727 if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
728 NL_SET_ERR_MSG_ATTR(extack, nla,
729 "Invalid attribute length for RTA_VIA");
732 alen = nla_len(nla) -
733 offsetof(struct rtvia, rtvia_addr);
734 if (alen > MAX_VIA_ALEN) {
735 NL_SET_ERR_MSG_ATTR(extack, nla,
736 "Invalid address length for RTA_VIA");
740 /* Validate the address family */
741 switch (via->rtvia_family) {
743 *via_table = NEIGH_LINK_TABLE;
746 *via_table = NEIGH_ARP_TABLE;
751 *via_table = NEIGH_ND_TABLE;
756 /* Unsupported address family */
760 memcpy(via_addr, via->rtvia_addr, alen);
768 static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg,
769 struct mpls_route *rt)
771 struct net *net = cfg->rc_nlinfo.nl_net;
772 struct mpls_nh *nh = rt->rt_nh;
779 nh->nh_labels = cfg->rc_output_labels;
780 for (i = 0; i < nh->nh_labels; i++)
781 nh->nh_label[i] = cfg->rc_output_label[i];
783 nh->nh_via_table = cfg->rc_via_table;
784 memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen);
785 nh->nh_via_alen = cfg->rc_via_alen;
787 err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex);
791 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
800 static int mpls_nh_build(struct net *net, struct mpls_route *rt,
801 struct mpls_nh *nh, int oif, struct nlattr *via,
802 struct nlattr *newdst, u8 max_labels,
803 struct netlink_ext_ack *extack)
811 err = nla_get_labels(newdst, max_labels, &nh->nh_labels,
812 nh->nh_label, extack);
818 err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table,
819 __mpls_nh_via(rt, nh), extack);
823 nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC;
826 err = mpls_nh_assign_dev(net, rt, nh, oif);
836 static u8 mpls_count_nexthops(struct rtnexthop *rtnh, int len,
837 u8 cfg_via_alen, u8 *max_via_alen,
846 while (rtnh_ok(rtnh, remaining)) {
847 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
851 attrlen = rtnh_attrlen(rtnh);
852 nla = nla_find(attrs, attrlen, RTA_VIA);
853 if (nla && nla_len(nla) >=
854 offsetof(struct rtvia, rtvia_addr)) {
855 int via_alen = nla_len(nla) -
856 offsetof(struct rtvia, rtvia_addr);
858 if (via_alen <= MAX_VIA_ALEN)
859 *max_via_alen = max_t(u16, *max_via_alen,
863 nla = nla_find(attrs, attrlen, RTA_NEWDST);
865 nla_get_labels(nla, MAX_NEW_LABELS, &n_labels,
869 *max_labels = max_t(u8, *max_labels, n_labels);
871 /* number of nexthops is tracked by a u8.
872 * Check for overflow.
878 rtnh = rtnh_next(rtnh, &remaining);
881 /* leftover implies invalid nexthop configuration, discard it */
882 return remaining > 0 ? 0 : nhs;
885 static int mpls_nh_build_multi(struct mpls_route_config *cfg,
886 struct mpls_route *rt, u8 max_labels,
887 struct netlink_ext_ack *extack)
889 struct rtnexthop *rtnh = cfg->rc_mp;
890 struct nlattr *nla_via, *nla_newdst;
891 int remaining = cfg->rc_mp_len;
895 change_nexthops(rt) {
902 if (!rtnh_ok(rtnh, remaining))
905 /* neither weighted multipath nor any flags
908 if (rtnh->rtnh_hops || rtnh->rtnh_flags)
911 attrlen = rtnh_attrlen(rtnh);
913 struct nlattr *attrs = rtnh_attrs(rtnh);
915 nla_via = nla_find(attrs, attrlen, RTA_VIA);
916 nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST);
919 err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh,
920 rtnh->rtnh_ifindex, nla_via, nla_newdst,
925 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
928 rtnh = rtnh_next(rtnh, &remaining);
930 } endfor_nexthops(rt);
940 static bool mpls_label_ok(struct net *net, unsigned int *index,
941 struct netlink_ext_ack *extack)
945 /* Reserved labels may not be set */
946 if (*index < MPLS_LABEL_FIRST_UNRESERVED) {
947 NL_SET_ERR_MSG(extack,
948 "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher");
952 /* The full 20 bit range may not be supported. */
953 if (is_ok && *index >= net->mpls.platform_labels) {
954 NL_SET_ERR_MSG(extack,
955 "Label >= configured maximum in platform_labels");
959 *index = array_index_nospec(*index, net->mpls.platform_labels);
963 static int mpls_route_add(struct mpls_route_config *cfg,
964 struct netlink_ext_ack *extack)
966 struct mpls_route __rcu **platform_label;
967 struct net *net = cfg->rc_nlinfo.nl_net;
968 struct mpls_route *rt, *old;
975 index = cfg->rc_label;
977 /* If a label was not specified during insert pick one */
978 if ((index == LABEL_NOT_SPECIFIED) &&
979 (cfg->rc_nlflags & NLM_F_CREATE)) {
980 index = find_free_label(net);
983 if (!mpls_label_ok(net, &index, extack))
986 /* Append makes no sense with mpls */
988 if (cfg->rc_nlflags & NLM_F_APPEND) {
989 NL_SET_ERR_MSG(extack, "MPLS does not support route append");
994 platform_label = rtnl_dereference(net->mpls.platform_label);
995 old = rtnl_dereference(platform_label[index]);
996 if ((cfg->rc_nlflags & NLM_F_EXCL) && old)
1000 if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old)
1004 if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old)
1009 nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len,
1010 cfg->rc_via_alen, &max_via_alen,
1013 max_via_alen = cfg->rc_via_alen;
1014 max_labels = cfg->rc_output_labels;
1019 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 READ_ONCE(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);
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);
1306 unsigned long ifindex;
1307 } *ctx = (void *)cb->ctx;
1308 struct net_device *dev;
1309 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 for_each_netdev_dump(net, dev, ctx->ifindex) {
1329 mdev = mpls_dev_get(dev);
1332 err = mpls_netconf_fill_devconf(skb, mdev,
1333 NETLINK_CB(cb->skb).portid,
1346 #define MPLS_PERDEV_SYSCTL_OFFSET(field) \
1347 (&((struct mpls_dev *)0)->field)
1349 static int mpls_conf_proc(const struct ctl_table *ctl, int write,
1350 void *buffer, size_t *lenp, loff_t *ppos)
1352 int oval = *(int *)ctl->data;
1353 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1356 struct mpls_dev *mdev = ctl->extra1;
1357 int i = (int *)ctl->data - (int *)mdev;
1358 struct net *net = ctl->extra2;
1359 int val = *(int *)ctl->data;
1361 if (i == offsetof(struct mpls_dev, input_enabled) &&
1363 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF,
1364 NETCONFA_INPUT, mdev);
1371 static const struct ctl_table mpls_dev_table[] = {
1373 .procname = "input",
1374 .maxlen = sizeof(int),
1376 .proc_handler = mpls_conf_proc,
1377 .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled),
1381 static int mpls_dev_sysctl_register(struct net_device *dev,
1382 struct mpls_dev *mdev)
1384 char path[sizeof("net/mpls/conf/") + IFNAMSIZ];
1385 size_t table_size = ARRAY_SIZE(mpls_dev_table);
1386 struct net *net = dev_net(dev);
1387 struct ctl_table *table;
1390 table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL);
1394 /* Table data contains only offsets relative to the base of
1395 * the mdev at this point, so make them absolute.
1397 for (i = 0; i < table_size; i++) {
1398 table[i].data = (char *)mdev + (uintptr_t)table[i].data;
1399 table[i].extra1 = mdev;
1400 table[i].extra2 = net;
1403 snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name);
1405 mdev->sysctl = register_net_sysctl_sz(net, path, table, table_size);
1409 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev);
1415 mdev->sysctl = NULL;
1419 static void mpls_dev_sysctl_unregister(struct net_device *dev,
1420 struct mpls_dev *mdev)
1422 struct net *net = dev_net(dev);
1423 const struct ctl_table *table;
1428 table = mdev->sysctl->ctl_table_arg;
1429 unregister_net_sysctl_table(mdev->sysctl);
1432 mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev);
1435 static struct mpls_dev *mpls_add_dev(struct net_device *dev)
1437 struct mpls_dev *mdev;
1443 mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
1445 return ERR_PTR(err);
1447 mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
1451 for_each_possible_cpu(i) {
1452 struct mpls_pcpu_stats *mpls_stats;
1454 mpls_stats = per_cpu_ptr(mdev->stats, i);
1455 u64_stats_init(&mpls_stats->syncp);
1460 err = mpls_dev_sysctl_register(dev, mdev);
1464 rcu_assign_pointer(dev->mpls_ptr, mdev);
1469 free_percpu(mdev->stats);
1471 return ERR_PTR(err);
1474 static void mpls_dev_destroy_rcu(struct rcu_head *head)
1476 struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
1478 free_percpu(mdev->stats);
1482 static int mpls_ifdown(struct net_device *dev, int event)
1484 struct mpls_route __rcu **platform_label;
1485 struct net *net = dev_net(dev);
1488 platform_label = rtnl_dereference(net->mpls.platform_label);
1489 for (index = 0; index < net->mpls.platform_labels; index++) {
1490 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1491 bool nh_del = false;
1497 if (event == NETDEV_UNREGISTER) {
1501 if (!nh->nh_dev || nh->nh_dev == dev)
1503 if (nh->nh_dev == dev)
1505 } endfor_nexthops(rt);
1507 /* if there are no more nexthops, delete the route */
1508 if (deleted == rt->rt_nhn) {
1509 mpls_route_update(net, index, NULL, NULL);
1514 size_t size = sizeof(*rt) + rt->rt_nhn *
1516 struct mpls_route *orig = rt;
1518 rt = kmemdup(orig, size, GFP_KERNEL);
1524 change_nexthops(rt) {
1525 unsigned int nh_flags = nh->nh_flags;
1527 if (nh->nh_dev != dev)
1532 case NETDEV_UNREGISTER:
1533 nh_flags |= RTNH_F_DEAD;
1536 nh_flags |= RTNH_F_LINKDOWN;
1539 if (event == NETDEV_UNREGISTER)
1542 if (nh->nh_flags != nh_flags)
1543 WRITE_ONCE(nh->nh_flags, nh_flags);
1545 if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)))
1547 } endfor_nexthops(rt);
1549 WRITE_ONCE(rt->rt_nhn_alive, alive);
1552 mpls_route_update(net, index, rt, NULL);
1558 static void mpls_ifup(struct net_device *dev, unsigned int flags)
1560 struct mpls_route __rcu **platform_label;
1561 struct net *net = dev_net(dev);
1565 platform_label = rtnl_dereference(net->mpls.platform_label);
1566 for (index = 0; index < net->mpls.platform_labels; index++) {
1567 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1573 change_nexthops(rt) {
1574 unsigned int nh_flags = nh->nh_flags;
1576 if (!(nh_flags & flags)) {
1580 if (nh->nh_dev != dev)
1584 WRITE_ONCE(nh->nh_flags, nh_flags);
1585 } endfor_nexthops(rt);
1587 WRITE_ONCE(rt->rt_nhn_alive, alive);
1591 static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
1594 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1595 struct mpls_dev *mdev;
1599 if (event == NETDEV_REGISTER) {
1600 mdev = mpls_add_dev(dev);
1602 return notifier_from_errno(PTR_ERR(mdev));
1607 mdev = mpls_dev_get(dev);
1614 err = mpls_ifdown(dev, event);
1616 return notifier_from_errno(err);
1619 flags = dev_get_flags(dev);
1620 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1621 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1623 mpls_ifup(dev, RTNH_F_DEAD);
1626 flags = dev_get_flags(dev);
1627 if (flags & (IFF_RUNNING | IFF_LOWER_UP)) {
1628 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1630 err = mpls_ifdown(dev, event);
1632 return notifier_from_errno(err);
1635 case NETDEV_UNREGISTER:
1636 err = mpls_ifdown(dev, event);
1638 return notifier_from_errno(err);
1639 mdev = mpls_dev_get(dev);
1641 mpls_dev_sysctl_unregister(dev, mdev);
1642 RCU_INIT_POINTER(dev->mpls_ptr, NULL);
1643 call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
1646 case NETDEV_CHANGENAME:
1647 mdev = mpls_dev_get(dev);
1649 mpls_dev_sysctl_unregister(dev, mdev);
1650 err = mpls_dev_sysctl_register(dev, mdev);
1652 return notifier_from_errno(err);
1659 static struct notifier_block mpls_dev_notifier = {
1660 .notifier_call = mpls_dev_notify,
1663 static int nla_put_via(struct sk_buff *skb,
1664 u8 table, const void *addr, int alen)
1666 static const int table_to_family[NEIGH_NR_TABLES + 1] = {
1667 AF_INET, AF_INET6, AF_PACKET,
1671 int family = AF_UNSPEC;
1673 nla = nla_reserve(skb, RTA_VIA, alen + 2);
1677 if (table <= NEIGH_NR_TABLES)
1678 family = table_to_family[table];
1680 via = nla_data(nla);
1681 via->rtvia_family = family;
1682 memcpy(via->rtvia_addr, addr, alen);
1686 int nla_put_labels(struct sk_buff *skb, int attrtype,
1687 u8 labels, const u32 label[])
1690 struct mpls_shim_hdr *nla_label;
1693 nla = nla_reserve(skb, attrtype, labels*4);
1697 nla_label = nla_data(nla);
1699 for (i = labels - 1; i >= 0; i--) {
1700 nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos);
1706 EXPORT_SYMBOL_GPL(nla_put_labels);
1708 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
1709 u32 label[], struct netlink_ext_ack *extack)
1711 unsigned len = nla_len(nla);
1712 struct mpls_shim_hdr *nla_label;
1717 /* len needs to be an even multiple of 4 (the label size). Number
1718 * of labels is a u8 so check for overflow.
1720 if (len & 3 || len / 4 > 255) {
1721 NL_SET_ERR_MSG_ATTR(extack, nla,
1722 "Invalid length for labels attribute");
1726 /* Limit the number of new labels allowed */
1728 if (nla_labels > max_labels) {
1729 NL_SET_ERR_MSG(extack, "Too many labels");
1733 /* when label == NULL, caller wants number of labels */
1737 nla_label = nla_data(nla);
1739 for (i = nla_labels - 1; i >= 0; i--, bos = false) {
1740 struct mpls_entry_decoded dec;
1741 dec = mpls_entry_decode(nla_label + i);
1743 /* Ensure the bottom of stack flag is properly set
1744 * and ttl and tc are both clear.
1747 NL_SET_ERR_MSG_ATTR(extack, nla,
1748 "TTL in label must be 0");
1753 NL_SET_ERR_MSG_ATTR(extack, nla,
1754 "Traffic class in label must be 0");
1758 if (dec.bos != bos) {
1759 NL_SET_BAD_ATTR(extack, nla);
1761 NL_SET_ERR_MSG(extack,
1762 "BOS bit must be set in first label");
1764 NL_SET_ERR_MSG(extack,
1765 "BOS bit can only be set in first label");
1770 switch (dec.label) {
1771 case MPLS_LABEL_IMPLNULL:
1772 /* RFC3032: This is a label that an LSR may
1773 * assign and distribute, but which never
1774 * actually appears in the encapsulation.
1776 NL_SET_ERR_MSG_ATTR(extack, nla,
1777 "Implicit NULL Label (3) can not be used in encapsulation");
1781 label[i] = dec.label;
1784 *labels = nla_labels;
1787 EXPORT_SYMBOL_GPL(nla_get_labels);
1789 static int rtm_to_route_config(struct sk_buff *skb,
1790 struct nlmsghdr *nlh,
1791 struct mpls_route_config *cfg,
1792 struct netlink_ext_ack *extack)
1795 struct nlattr *tb[RTA_MAX+1];
1799 err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
1800 rtm_mpls_policy, extack);
1805 rtm = nlmsg_data(nlh);
1807 if (rtm->rtm_family != AF_MPLS) {
1808 NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg");
1811 if (rtm->rtm_dst_len != 20) {
1812 NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS");
1815 if (rtm->rtm_src_len != 0) {
1816 NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS");
1819 if (rtm->rtm_tos != 0) {
1820 NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS");
1823 if (rtm->rtm_table != RT_TABLE_MAIN) {
1824 NL_SET_ERR_MSG(extack,
1825 "MPLS only supports the main route table");
1828 /* Any value is acceptable for rtm_protocol */
1830 /* As mpls uses destination specific addresses
1831 * (or source specific address in the case of multicast)
1832 * all addresses have universal scope.
1834 if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) {
1835 NL_SET_ERR_MSG(extack,
1836 "Invalid route scope - MPLS only supports UNIVERSE");
1839 if (rtm->rtm_type != RTN_UNICAST) {
1840 NL_SET_ERR_MSG(extack,
1841 "Invalid route type - MPLS only supports UNICAST");
1844 if (rtm->rtm_flags != 0) {
1845 NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS");
1849 cfg->rc_label = LABEL_NOT_SPECIFIED;
1850 cfg->rc_protocol = rtm->rtm_protocol;
1851 cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC;
1852 cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
1853 cfg->rc_nlflags = nlh->nlmsg_flags;
1854 cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid;
1855 cfg->rc_nlinfo.nlh = nlh;
1856 cfg->rc_nlinfo.nl_net = sock_net(skb->sk);
1858 for (index = 0; index <= RTA_MAX; index++) {
1859 struct nlattr *nla = tb[index];
1865 cfg->rc_ifindex = nla_get_u32(nla);
1868 if (nla_get_labels(nla, MAX_NEW_LABELS,
1869 &cfg->rc_output_labels,
1870 cfg->rc_output_label, extack))
1876 if (nla_get_labels(nla, 1, &label_count,
1877 &cfg->rc_label, extack))
1880 if (!mpls_label_ok(cfg->rc_nlinfo.nl_net,
1881 &cfg->rc_label, extack))
1886 NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute");
1890 if (nla_get_via(nla, &cfg->rc_via_alen,
1891 &cfg->rc_via_table, cfg->rc_via,
1898 cfg->rc_mp = nla_data(nla);
1899 cfg->rc_mp_len = nla_len(nla);
1902 case RTA_TTL_PROPAGATE:
1904 u8 ttl_propagate = nla_get_u8(nla);
1906 if (ttl_propagate > 1) {
1907 NL_SET_ERR_MSG_ATTR(extack, nla,
1908 "RTA_TTL_PROPAGATE can only be 0 or 1");
1911 cfg->rc_ttl_propagate = ttl_propagate ?
1912 MPLS_TTL_PROP_ENABLED :
1913 MPLS_TTL_PROP_DISABLED;
1917 NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute");
1918 /* Unsupported attribute */
1928 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1929 struct netlink_ext_ack *extack)
1931 struct mpls_route_config *cfg;
1934 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1938 err = rtm_to_route_config(skb, nlh, cfg, extack);
1942 err = mpls_route_del(cfg, extack);
1950 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1951 struct netlink_ext_ack *extack)
1953 struct mpls_route_config *cfg;
1956 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1960 err = rtm_to_route_config(skb, nlh, cfg, extack);
1964 err = mpls_route_add(cfg, extack);
1971 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event,
1972 u32 label, struct mpls_route *rt, int flags)
1974 struct net_device *dev;
1975 struct nlmsghdr *nlh;
1978 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1982 rtm = nlmsg_data(nlh);
1983 rtm->rtm_family = AF_MPLS;
1984 rtm->rtm_dst_len = 20;
1985 rtm->rtm_src_len = 0;
1987 rtm->rtm_table = RT_TABLE_MAIN;
1988 rtm->rtm_protocol = rt->rt_protocol;
1989 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
1990 rtm->rtm_type = RTN_UNICAST;
1993 if (nla_put_labels(skb, RTA_DST, 1, &label))
1994 goto nla_put_failure;
1996 if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) {
1997 bool ttl_propagate =
1998 rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED;
2000 if (nla_put_u8(skb, RTA_TTL_PROPAGATE,
2002 goto nla_put_failure;
2004 if (rt->rt_nhn == 1) {
2005 const struct mpls_nh *nh = rt->rt_nh;
2007 if (nh->nh_labels &&
2008 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2010 goto nla_put_failure;
2011 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2012 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2014 goto nla_put_failure;
2016 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2017 goto nla_put_failure;
2018 if (nh->nh_flags & RTNH_F_LINKDOWN)
2019 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2020 if (nh->nh_flags & RTNH_F_DEAD)
2021 rtm->rtm_flags |= RTNH_F_DEAD;
2023 struct rtnexthop *rtnh;
2028 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
2030 goto nla_put_failure;
2037 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
2039 goto nla_put_failure;
2041 rtnh->rtnh_ifindex = dev->ifindex;
2042 if (nh->nh_flags & RTNH_F_LINKDOWN) {
2043 rtnh->rtnh_flags |= RTNH_F_LINKDOWN;
2046 if (nh->nh_flags & RTNH_F_DEAD) {
2047 rtnh->rtnh_flags |= RTNH_F_DEAD;
2051 if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST,
2054 goto nla_put_failure;
2055 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2056 nla_put_via(skb, nh->nh_via_table,
2057 mpls_nh_via(rt, nh),
2059 goto nla_put_failure;
2061 /* length of rtnetlink header + attributes */
2062 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
2063 } endfor_nexthops(rt);
2065 if (linkdown == rt->rt_nhn)
2066 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2067 if (dead == rt->rt_nhn)
2068 rtm->rtm_flags |= RTNH_F_DEAD;
2070 nla_nest_end(skb, mp);
2073 nlmsg_end(skb, nlh);
2077 nlmsg_cancel(skb, nlh);
2081 #if IS_ENABLED(CONFIG_INET)
2082 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2083 struct fib_dump_filter *filter,
2084 struct netlink_callback *cb)
2086 return ip_valid_fib_dump_req(net, nlh, filter, cb);
2089 static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
2090 struct fib_dump_filter *filter,
2091 struct netlink_callback *cb)
2093 struct netlink_ext_ack *extack = cb->extack;
2094 struct nlattr *tb[RTA_MAX + 1];
2098 rtm = nlmsg_payload(nlh, sizeof(*rtm));
2100 NL_SET_ERR_MSG_MOD(extack, "Invalid header for FIB dump request");
2104 if (rtm->rtm_dst_len || rtm->rtm_src_len || rtm->rtm_tos ||
2105 rtm->rtm_table || rtm->rtm_scope || rtm->rtm_type ||
2107 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for FIB dump request");
2111 if (rtm->rtm_protocol) {
2112 filter->protocol = rtm->rtm_protocol;
2113 filter->filter_set = 1;
2114 cb->answer_flags = NLM_F_DUMP_FILTERED;
2117 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2118 rtm_mpls_policy, extack);
2122 for (i = 0; i <= RTA_MAX; ++i) {
2126 ifindex = nla_get_u32(tb[i]);
2127 filter->dev = __dev_get_by_index(net, ifindex);
2130 filter->filter_set = 1;
2132 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
2141 static bool mpls_rt_uses_dev(struct mpls_route *rt,
2142 const struct net_device *dev)
2144 if (rt->rt_nhn == 1) {
2145 struct mpls_nh *nh = rt->rt_nh;
2147 if (nh->nh_dev == dev)
2151 if (nh->nh_dev == dev)
2153 } endfor_nexthops(rt);
2159 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
2161 const struct nlmsghdr *nlh = cb->nlh;
2162 struct net *net = sock_net(skb->sk);
2163 struct mpls_route __rcu **platform_label;
2164 struct fib_dump_filter filter = {
2167 unsigned int flags = NLM_F_MULTI;
2168 size_t platform_labels;
2173 if (cb->strict_check) {
2176 err = mpls_valid_fib_dump_req(net, nlh, &filter, cb);
2180 /* for MPLS, there is only 1 table with fixed type and flags.
2181 * If either are set in the filter then return nothing.
2183 if ((filter.table_id && filter.table_id != RT_TABLE_MAIN) ||
2184 (filter.rt_type && filter.rt_type != RTN_UNICAST) ||
2189 index = cb->args[0];
2190 if (index < MPLS_LABEL_FIRST_UNRESERVED)
2191 index = MPLS_LABEL_FIRST_UNRESERVED;
2193 platform_label = rtnl_dereference(net->mpls.platform_label);
2194 platform_labels = net->mpls.platform_labels;
2196 if (filter.filter_set)
2197 flags |= NLM_F_DUMP_FILTERED;
2199 for (; index < platform_labels; index++) {
2200 struct mpls_route *rt;
2202 rt = rtnl_dereference(platform_label[index]);
2206 if ((filter.dev && !mpls_rt_uses_dev(rt, filter.dev)) ||
2207 (filter.protocol && rt->rt_protocol != filter.protocol))
2210 if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid,
2211 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2212 index, rt, flags) < 0)
2215 cb->args[0] = index;
2220 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
2223 NLMSG_ALIGN(sizeof(struct rtmsg))
2224 + nla_total_size(4) /* RTA_DST */
2225 + nla_total_size(1); /* RTA_TTL_PROPAGATE */
2227 if (rt->rt_nhn == 1) {
2228 struct mpls_nh *nh = rt->rt_nh;
2231 payload += nla_total_size(4); /* RTA_OIF */
2232 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */
2233 payload += nla_total_size(2 + nh->nh_via_alen);
2234 if (nh->nh_labels) /* RTA_NEWDST */
2235 payload += nla_total_size(nh->nh_labels * 4);
2237 /* each nexthop is packed in an attribute */
2243 nhsize += nla_total_size(sizeof(struct rtnexthop));
2245 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC)
2246 nhsize += nla_total_size(2 + nh->nh_via_alen);
2248 nhsize += nla_total_size(nh->nh_labels * 4);
2249 } endfor_nexthops(rt);
2250 /* nested attribute */
2251 payload += nla_total_size(nhsize);
2257 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
2258 struct nlmsghdr *nlh, struct net *net, u32 portid,
2259 unsigned int nlm_flags)
2261 struct sk_buff *skb;
2262 u32 seq = nlh ? nlh->nlmsg_seq : 0;
2265 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2269 err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags);
2271 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2272 WARN_ON(err == -EMSGSIZE);
2276 rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL);
2280 rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err);
2283 static int mpls_valid_getroute_req(struct sk_buff *skb,
2284 const struct nlmsghdr *nlh,
2286 struct netlink_ext_ack *extack)
2291 rtm = nlmsg_payload(nlh, sizeof(*rtm));
2293 NL_SET_ERR_MSG_MOD(extack,
2294 "Invalid header for get route request");
2298 if (!netlink_strict_get_check(skb))
2299 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
2300 rtm_mpls_policy, extack);
2302 if ((rtm->rtm_dst_len && rtm->rtm_dst_len != 20) ||
2303 rtm->rtm_src_len || rtm->rtm_tos || rtm->rtm_table ||
2304 rtm->rtm_protocol || rtm->rtm_scope || rtm->rtm_type) {
2305 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
2308 if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
2309 NL_SET_ERR_MSG_MOD(extack,
2310 "Invalid flags for get route request");
2314 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
2315 rtm_mpls_policy, extack);
2319 if ((tb[RTA_DST] || tb[RTA_NEWDST]) && !rtm->rtm_dst_len) {
2320 NL_SET_ERR_MSG_MOD(extack, "rtm_dst_len must be 20 for MPLS");
2324 for (i = 0; i <= RTA_MAX; i++) {
2333 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
2341 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
2342 struct netlink_ext_ack *extack)
2344 struct net *net = sock_net(in_skb->sk);
2345 u32 portid = NETLINK_CB(in_skb).portid;
2346 u32 in_label = LABEL_NOT_SPECIFIED;
2347 struct nlattr *tb[RTA_MAX + 1];
2348 u32 labels[MAX_NEW_LABELS];
2349 struct mpls_shim_hdr *hdr;
2350 unsigned int hdr_size = 0;
2351 const struct mpls_nh *nh;
2352 struct net_device *dev;
2353 struct mpls_route *rt;
2354 struct rtmsg *rtm, *r;
2355 struct nlmsghdr *nlh;
2356 struct sk_buff *skb;
2360 err = mpls_valid_getroute_req(in_skb, in_nlh, tb, extack);
2364 rtm = nlmsg_data(in_nlh);
2369 if (nla_get_labels(tb[RTA_DST], 1, &label_count,
2370 &in_label, extack)) {
2375 if (!mpls_label_ok(net, &in_label, extack)) {
2381 rt = mpls_route_input_rcu(net, in_label);
2387 if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
2388 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2394 err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq,
2395 RTM_NEWROUTE, in_label, rt, 0);
2397 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2398 WARN_ON(err == -EMSGSIZE);
2402 return rtnl_unicast(skb, net, portid);
2405 if (tb[RTA_NEWDST]) {
2406 if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels,
2407 labels, extack) != 0) {
2412 hdr_size = n_labels * sizeof(struct mpls_shim_hdr);
2415 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2421 skb->protocol = htons(ETH_P_MPLS_UC);
2427 if (skb_cow(skb, hdr_size)) {
2432 skb_reserve(skb, hdr_size);
2433 skb_push(skb, hdr_size);
2434 skb_reset_network_header(skb);
2436 /* Push new labels */
2437 hdr = mpls_hdr(skb);
2439 for (i = n_labels - 1; i >= 0; i--) {
2440 hdr[i] = mpls_entry_encode(labels[i],
2446 nh = mpls_select_multipath(rt, skb);
2453 skb_pull(skb, hdr_size);
2454 skb_reset_network_header(skb);
2457 nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq,
2458 RTM_NEWROUTE, sizeof(*r), 0);
2464 r = nlmsg_data(nlh);
2465 r->rtm_family = AF_MPLS;
2466 r->rtm_dst_len = 20;
2468 r->rtm_table = RT_TABLE_MAIN;
2469 r->rtm_type = RTN_UNICAST;
2470 r->rtm_scope = RT_SCOPE_UNIVERSE;
2471 r->rtm_protocol = rt->rt_protocol;
2474 if (nla_put_labels(skb, RTA_DST, 1, &in_label))
2475 goto nla_put_failure;
2477 if (nh->nh_labels &&
2478 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2480 goto nla_put_failure;
2482 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2483 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2485 goto nla_put_failure;
2487 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2488 goto nla_put_failure;
2490 nlmsg_end(skb, nlh);
2492 err = rtnl_unicast(skb, net, portid);
2497 nlmsg_cancel(skb, nlh);
2504 static int resize_platform_label_table(struct net *net, size_t limit)
2506 size_t size = sizeof(struct mpls_route *) * limit;
2509 struct mpls_route __rcu **labels = NULL, **old;
2510 struct mpls_route *rt0 = NULL, *rt2 = NULL;
2514 labels = kvzalloc(size, GFP_KERNEL);
2519 /* In case the predefined labels need to be populated */
2520 if (limit > MPLS_LABEL_IPV4NULL) {
2521 struct net_device *lo = net->loopback_dev;
2522 rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
2525 rt0->rt_nh->nh_dev = lo;
2526 rt0->rt_protocol = RTPROT_KERNEL;
2527 rt0->rt_payload_type = MPT_IPV4;
2528 rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2529 rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2530 rt0->rt_nh->nh_via_alen = lo->addr_len;
2531 memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr,
2534 if (limit > MPLS_LABEL_IPV6NULL) {
2535 struct net_device *lo = net->loopback_dev;
2536 rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
2539 rt2->rt_nh->nh_dev = lo;
2540 rt2->rt_protocol = RTPROT_KERNEL;
2541 rt2->rt_payload_type = MPT_IPV6;
2542 rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2543 rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2544 rt2->rt_nh->nh_via_alen = lo->addr_len;
2545 memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr,
2550 /* Remember the original table */
2551 old = rtnl_dereference(net->mpls.platform_label);
2552 old_limit = net->mpls.platform_labels;
2554 /* Free any labels beyond the new table */
2555 for (index = limit; index < old_limit; index++)
2556 mpls_route_update(net, index, NULL, NULL);
2558 /* Copy over the old labels */
2560 if (old_limit < limit)
2561 cp_size = old_limit * sizeof(struct mpls_route *);
2563 memcpy(labels, old, cp_size);
2565 /* If needed set the predefined labels */
2566 if ((old_limit <= MPLS_LABEL_IPV6NULL) &&
2567 (limit > MPLS_LABEL_IPV6NULL)) {
2568 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2);
2572 if ((old_limit <= MPLS_LABEL_IPV4NULL) &&
2573 (limit > MPLS_LABEL_IPV4NULL)) {
2574 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0);
2578 /* Update the global pointers */
2579 net->mpls.platform_labels = limit;
2580 rcu_assign_pointer(net->mpls.platform_label, labels);
2601 static int mpls_platform_labels(const struct ctl_table *table, int write,
2602 void *buffer, size_t *lenp, loff_t *ppos)
2604 struct net *net = table->data;
2605 int platform_labels = net->mpls.platform_labels;
2607 struct ctl_table tmp = {
2608 .procname = table->procname,
2609 .data = &platform_labels,
2610 .maxlen = sizeof(int),
2611 .mode = table->mode,
2612 .extra1 = SYSCTL_ZERO,
2613 .extra2 = &label_limit,
2616 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2618 if (write && ret == 0)
2619 ret = resize_platform_label_table(net, platform_labels);
2624 #define MPLS_NS_SYSCTL_OFFSET(field) \
2625 (&((struct net *)0)->field)
2627 static const struct ctl_table mpls_table[] = {
2629 .procname = "platform_labels",
2631 .maxlen = sizeof(int),
2633 .proc_handler = mpls_platform_labels,
2636 .procname = "ip_ttl_propagate",
2637 .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate),
2638 .maxlen = sizeof(int),
2640 .proc_handler = proc_dointvec_minmax,
2641 .extra1 = SYSCTL_ZERO,
2642 .extra2 = SYSCTL_ONE,
2645 .procname = "default_ttl",
2646 .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl),
2647 .maxlen = sizeof(int),
2649 .proc_handler = proc_dointvec_minmax,
2650 .extra1 = SYSCTL_ONE,
2655 static int mpls_net_init(struct net *net)
2657 size_t table_size = ARRAY_SIZE(mpls_table);
2658 struct ctl_table *table;
2661 net->mpls.platform_labels = 0;
2662 net->mpls.platform_label = NULL;
2663 net->mpls.ip_ttl_propagate = 1;
2664 net->mpls.default_ttl = 255;
2666 table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL);
2670 /* Table data contains only offsets relative to the base of
2671 * the mdev at this point, so make them absolute.
2673 for (i = 0; i < table_size; i++)
2674 table[i].data = (char *)net + (uintptr_t)table[i].data;
2676 net->mpls.ctl = register_net_sysctl_sz(net, "net/mpls", table,
2678 if (net->mpls.ctl == NULL) {
2686 static void mpls_net_exit(struct net *net)
2688 struct mpls_route __rcu **platform_label;
2689 size_t platform_labels;
2690 const struct ctl_table *table;
2693 table = net->mpls.ctl->ctl_table_arg;
2694 unregister_net_sysctl_table(net->mpls.ctl);
2697 /* An rcu grace period has passed since there was a device in
2698 * the network namespace (and thus the last in flight packet)
2699 * left this network namespace. This is because
2700 * unregister_netdevice_many and netdev_run_todo has completed
2701 * for each network device that was in this network namespace.
2703 * As such no additional rcu synchronization is necessary when
2704 * freeing the platform_label table.
2707 platform_label = rtnl_dereference(net->mpls.platform_label);
2708 platform_labels = net->mpls.platform_labels;
2709 for (index = 0; index < platform_labels; index++) {
2710 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
2711 RCU_INIT_POINTER(platform_label[index], NULL);
2712 mpls_notify_route(net, index, rt, NULL, NULL);
2717 kvfree(platform_label);
2720 static struct pernet_operations mpls_net_ops = {
2721 .init = mpls_net_init,
2722 .exit = mpls_net_exit,
2725 static struct rtnl_af_ops mpls_af_ops __read_mostly = {
2727 .fill_stats_af = mpls_fill_stats_af,
2728 .get_stats_af_size = mpls_get_stats_af_size,
2731 static const struct rtnl_msg_handler mpls_rtnl_msg_handlers[] __initdata_or_module = {
2732 {THIS_MODULE, PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, 0},
2733 {THIS_MODULE, PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, 0},
2734 {THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes, 0},
2735 {THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
2736 mpls_netconf_get_devconf, mpls_netconf_dump_devconf,
2737 RTNL_FLAG_DUMP_UNLOCKED},
2740 static int __init mpls_init(void)
2744 BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4);
2746 err = register_pernet_subsys(&mpls_net_ops);
2750 err = register_netdevice_notifier(&mpls_dev_notifier);
2752 goto out_unregister_pernet;
2754 dev_add_pack(&mpls_packet_type);
2756 err = rtnl_af_register(&mpls_af_ops);
2758 goto out_unregister_dev_type;
2760 err = rtnl_register_many(mpls_rtnl_msg_handlers);
2762 goto out_unregister_rtnl_af;
2764 err = ipgre_tunnel_encap_add_mpls_ops();
2766 pr_err("Can't add mpls over gre tunnel ops\n");
2767 goto out_unregister_rtnl;
2774 out_unregister_rtnl:
2775 rtnl_unregister_many(mpls_rtnl_msg_handlers);
2776 out_unregister_rtnl_af:
2777 rtnl_af_unregister(&mpls_af_ops);
2778 out_unregister_dev_type:
2779 dev_remove_pack(&mpls_packet_type);
2780 out_unregister_pernet:
2781 unregister_pernet_subsys(&mpls_net_ops);
2784 module_init(mpls_init);
2786 static void __exit mpls_exit(void)
2788 rtnl_unregister_all(PF_MPLS);
2789 rtnl_af_unregister(&mpls_af_ops);
2790 dev_remove_pack(&mpls_packet_type);
2791 unregister_netdevice_notifier(&mpls_dev_notifier);
2792 unregister_pernet_subsys(&mpls_net_ops);
2793 ipgre_tunnel_encap_del_mpls_ops();
2795 module_exit(mpls_exit);
2797 MODULE_DESCRIPTION("MultiProtocol Label Switching");
2798 MODULE_LICENSE("GPL v2");
2799 MODULE_ALIAS_NETPROTO(PF_MPLS);