1 #include <linux/types.h>
2 #include <linux/skbuff.h>
3 #include <linux/socket.h>
4 #include <linux/sysctl.h>
6 #include <linux/module.h>
7 #include <linux/if_arp.h>
8 #include <linux/ipv6.h>
9 #include <linux/mpls.h>
10 #include <linux/netconf.h>
11 #include <linux/nospec.h>
12 #include <linux/vmalloc.h>
13 #include <linux/percpu.h>
18 #include <net/ip_fib.h>
19 #include <net/netevent.h>
20 #include <net/netns/generic.h>
21 #if IS_ENABLED(CONFIG_IPV6)
24 #include <net/addrconf.h>
25 #include <net/nexthop.h>
28 /* max memory we will use for mpls_route */
29 #define MAX_MPLS_ROUTE_MEM 4096
31 /* Maximum number of labels to look ahead at when selecting a path of
34 #define MAX_MP_SELECT_LABELS 4
36 #define MPLS_NEIGH_TABLE_UNSPEC (NEIGH_LINK_TABLE + 1)
40 static int label_limit = (1 << 20) - 1;
41 static int ttl_max = 255;
43 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
44 struct nlmsghdr *nlh, struct net *net, u32 portid,
45 unsigned int nlm_flags);
47 static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
49 struct mpls_route *rt = NULL;
51 if (index < net->mpls.platform_labels) {
52 struct mpls_route __rcu **platform_label =
53 rcu_dereference(net->mpls.platform_label);
54 rt = rcu_dereference(platform_label[index]);
59 bool mpls_output_possible(const struct net_device *dev)
61 return dev && (dev->flags & IFF_UP) && netif_carrier_ok(dev);
63 EXPORT_SYMBOL_GPL(mpls_output_possible);
65 static u8 *__mpls_nh_via(struct mpls_route *rt, struct mpls_nh *nh)
67 return (u8 *)nh + rt->rt_via_offset;
70 static const u8 *mpls_nh_via(const struct mpls_route *rt,
71 const struct mpls_nh *nh)
73 return __mpls_nh_via((struct mpls_route *)rt, (struct mpls_nh *)nh);
76 static unsigned int mpls_nh_header_size(const struct mpls_nh *nh)
78 /* The size of the layer 2.5 labels to be added for this route */
79 return nh->nh_labels * sizeof(struct mpls_shim_hdr);
82 unsigned int mpls_dev_mtu(const struct net_device *dev)
84 /* The amount of data the layer 2 frame can hold */
87 EXPORT_SYMBOL_GPL(mpls_dev_mtu);
89 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
94 if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu))
99 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
101 void mpls_stats_inc_outucastpkts(struct net_device *dev,
102 const struct sk_buff *skb)
104 struct mpls_dev *mdev;
106 if (skb->protocol == htons(ETH_P_MPLS_UC)) {
107 mdev = mpls_dev_get(dev);
109 MPLS_INC_STATS_LEN(mdev, skb->len,
112 } else if (skb->protocol == htons(ETH_P_IP)) {
113 IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
114 #if IS_ENABLED(CONFIG_IPV6)
115 } else if (skb->protocol == htons(ETH_P_IPV6)) {
116 struct inet6_dev *in6dev = __in6_dev_get(dev);
119 IP6_UPD_PO_STATS(dev_net(dev), in6dev,
120 IPSTATS_MIB_OUT, skb->len);
124 EXPORT_SYMBOL_GPL(mpls_stats_inc_outucastpkts);
126 static u32 mpls_multipath_hash(struct mpls_route *rt, struct sk_buff *skb)
128 struct mpls_entry_decoded dec;
129 unsigned int mpls_hdr_len = 0;
130 struct mpls_shim_hdr *hdr;
131 bool eli_seen = false;
135 for (label_index = 0; label_index < MAX_MP_SELECT_LABELS;
137 mpls_hdr_len += sizeof(*hdr);
138 if (!pskb_may_pull(skb, mpls_hdr_len))
141 /* Read and decode the current label */
142 hdr = mpls_hdr(skb) + label_index;
143 dec = mpls_entry_decode(hdr);
145 /* RFC6790 - reserved labels MUST NOT be used as keys
146 * for the load-balancing function
148 if (likely(dec.label >= MPLS_LABEL_FIRST_UNRESERVED)) {
149 hash = jhash_1word(dec.label, hash);
151 /* The entropy label follows the entropy label
152 * indicator, so this means that the entropy
153 * label was just added to the hash - no need to
154 * go any deeper either in the label stack or in the
159 } else if (dec.label == MPLS_LABEL_ENTROPY) {
166 /* found bottom label; does skb have room for a header? */
167 if (pskb_may_pull(skb, mpls_hdr_len + sizeof(struct iphdr))) {
168 const struct iphdr *v4hdr;
170 v4hdr = (const struct iphdr *)(hdr + 1);
171 if (v4hdr->version == 4) {
172 hash = jhash_3words(ntohl(v4hdr->saddr),
174 v4hdr->protocol, hash);
175 } else if (v4hdr->version == 6 &&
176 pskb_may_pull(skb, mpls_hdr_len +
177 sizeof(struct ipv6hdr))) {
178 const struct ipv6hdr *v6hdr;
180 v6hdr = (const struct ipv6hdr *)(hdr + 1);
181 hash = __ipv6_addr_jhash(&v6hdr->saddr, hash);
182 hash = __ipv6_addr_jhash(&v6hdr->daddr, hash);
183 hash = jhash_1word(v6hdr->nexthdr, hash);
193 static struct mpls_nh *mpls_get_nexthop(struct mpls_route *rt, u8 index)
195 return (struct mpls_nh *)((u8 *)rt->rt_nh + index * rt->rt_nh_size);
198 /* number of alive nexthops (rt->rt_nhn_alive) and the flags for
199 * a next hop (nh->nh_flags) are modified by netdev event handlers.
200 * Since those fields can change at any moment, use READ_ONCE to
203 static struct mpls_nh *mpls_select_multipath(struct mpls_route *rt,
211 /* No need to look further into packet if there's only
217 alive = READ_ONCE(rt->rt_nhn_alive);
221 hash = mpls_multipath_hash(rt, skb);
222 nh_index = hash % alive;
223 if (alive == rt->rt_nhn)
226 unsigned int nh_flags = READ_ONCE(nh->nh_flags);
228 if (nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
233 } endfor_nexthops(rt);
236 return mpls_get_nexthop(rt, nh_index);
239 static bool mpls_egress(struct net *net, struct mpls_route *rt,
240 struct sk_buff *skb, struct mpls_entry_decoded dec)
242 enum mpls_payload_type payload_type;
243 bool success = false;
245 /* The IPv4 code below accesses through the IPv4 header
246 * checksum, which is 12 bytes into the packet.
247 * The IPv6 code below accesses through the IPv6 hop limit
248 * which is 8 bytes into the packet.
250 * For all supported cases there should always be at least 12
251 * bytes of packet data present. The IPv4 header is 20 bytes
252 * without options and the IPv6 header is always 40 bytes
255 if (!pskb_may_pull(skb, 12))
258 payload_type = rt->rt_payload_type;
259 if (payload_type == MPT_UNSPEC)
260 payload_type = ip_hdr(skb)->version;
262 switch (payload_type) {
264 struct iphdr *hdr4 = ip_hdr(skb);
266 skb->protocol = htons(ETH_P_IP);
268 /* If propagating TTL, take the decremented TTL from
269 * the incoming MPLS header, otherwise decrement the
270 * TTL, but only if not 0 to avoid underflow.
272 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
273 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
274 net->mpls.ip_ttl_propagate))
277 new_ttl = hdr4->ttl ? hdr4->ttl - 1 : 0;
279 csum_replace2(&hdr4->check,
280 htons(hdr4->ttl << 8),
281 htons(new_ttl << 8));
287 struct ipv6hdr *hdr6 = ipv6_hdr(skb);
288 skb->protocol = htons(ETH_P_IPV6);
290 /* If propagating TTL, take the decremented TTL from
291 * the incoming MPLS header, otherwise decrement the
292 * hop limit, but only if not 0 to avoid underflow.
294 if (rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED ||
295 (rt->rt_ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
296 net->mpls.ip_ttl_propagate))
297 hdr6->hop_limit = dec.ttl;
298 else if (hdr6->hop_limit)
299 hdr6->hop_limit = hdr6->hop_limit - 1;
304 /* Should have decided which protocol it is by now */
311 static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
312 struct packet_type *pt, struct net_device *orig_dev)
314 struct net *net = dev_net(dev);
315 struct mpls_shim_hdr *hdr;
316 struct mpls_route *rt;
318 struct mpls_entry_decoded dec;
319 struct net_device *out_dev;
320 struct mpls_dev *out_mdev;
321 struct mpls_dev *mdev;
323 unsigned int new_header_size;
327 /* Careful this entire function runs inside of an rcu critical section */
329 mdev = mpls_dev_get(dev);
333 MPLS_INC_STATS_LEN(mdev, skb->len, rx_packets,
336 if (!mdev->input_enabled) {
337 MPLS_INC_STATS(mdev, rx_dropped);
341 if (skb->pkt_type != PACKET_HOST)
344 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
347 if (!pskb_may_pull(skb, sizeof(*hdr)))
350 /* Read and decode the label */
352 dec = mpls_entry_decode(hdr);
354 rt = mpls_route_input_rcu(net, dec.label);
356 MPLS_INC_STATS(mdev, rx_noroute);
360 nh = mpls_select_multipath(rt, skb);
365 skb_pull(skb, sizeof(*hdr));
366 skb_reset_network_header(skb);
370 if (skb_warn_if_lro(skb))
373 skb_forward_csum(skb);
375 /* Verify ttl is valid */
380 /* Find the output device */
381 out_dev = rcu_dereference(nh->nh_dev);
382 if (!mpls_output_possible(out_dev))
385 /* Verify the destination can hold the packet */
386 new_header_size = mpls_nh_header_size(nh);
387 mtu = mpls_dev_mtu(out_dev);
388 if (mpls_pkt_too_big(skb, mtu - new_header_size))
391 hh_len = LL_RESERVED_SPACE(out_dev);
392 if (!out_dev->header_ops)
395 /* Ensure there is enough space for the headers in the skb */
396 if (skb_cow(skb, hh_len + new_header_size))
400 skb->protocol = htons(ETH_P_MPLS_UC);
402 if (unlikely(!new_header_size && dec.bos)) {
403 /* Penultimate hop popping */
404 if (!mpls_egress(dev_net(out_dev), rt, skb, dec))
409 skb_push(skb, new_header_size);
410 skb_reset_network_header(skb);
411 /* Push the new labels */
414 for (i = nh->nh_labels - 1; i >= 0; i--) {
415 hdr[i] = mpls_entry_encode(nh->nh_label[i],
421 mpls_stats_inc_outucastpkts(out_dev, skb);
423 /* If via wasn't specified then send out using device address */
424 if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
425 err = neigh_xmit(NEIGH_LINK_TABLE, out_dev,
426 out_dev->dev_addr, skb);
428 err = neigh_xmit(nh->nh_via_table, out_dev,
429 mpls_nh_via(rt, nh), skb);
431 net_dbg_ratelimited("%s: packet transmission failed: %d\n",
436 out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
438 MPLS_INC_STATS(out_mdev, tx_errors);
441 MPLS_INC_STATS(mdev, rx_errors);
447 static struct packet_type mpls_packet_type __read_mostly = {
448 .type = cpu_to_be16(ETH_P_MPLS_UC),
449 .func = mpls_forward,
452 static const struct nla_policy rtm_mpls_policy[RTA_MAX+1] = {
453 [RTA_DST] = { .type = NLA_U32 },
454 [RTA_OIF] = { .type = NLA_U32 },
455 [RTA_TTL_PROPAGATE] = { .type = NLA_U8 },
458 struct mpls_route_config {
463 u8 rc_via[MAX_VIA_ALEN];
467 u32 rc_output_label[MAX_NEW_LABELS];
469 enum mpls_payload_type rc_payload_type;
470 struct nl_info rc_nlinfo;
471 struct rtnexthop *rc_mp;
475 /* all nexthops within a route have the same size based on max
476 * number of labels and max via length for a hop
478 static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels)
480 u8 nh_size = MPLS_NH_SIZE(max_labels, max_alen);
481 struct mpls_route *rt;
484 size = sizeof(*rt) + num_nh * nh_size;
485 if (size > MAX_MPLS_ROUTE_MEM)
486 return ERR_PTR(-EINVAL);
488 rt = kzalloc(size, GFP_KERNEL);
490 return ERR_PTR(-ENOMEM);
493 rt->rt_nhn_alive = num_nh;
494 rt->rt_nh_size = nh_size;
495 rt->rt_via_offset = MPLS_NH_VIA_OFF(max_labels);
500 static void mpls_rt_free(struct mpls_route *rt)
503 kfree_rcu(rt, rt_rcu);
506 static void mpls_notify_route(struct net *net, unsigned index,
507 struct mpls_route *old, struct mpls_route *new,
508 const struct nl_info *info)
510 struct nlmsghdr *nlh = info ? info->nlh : NULL;
511 unsigned portid = info ? info->portid : 0;
512 int event = new ? RTM_NEWROUTE : RTM_DELROUTE;
513 struct mpls_route *rt = new ? new : old;
514 unsigned nlm_flags = (old && new) ? NLM_F_REPLACE : 0;
515 /* Ignore reserved labels for now */
516 if (rt && (index >= MPLS_LABEL_FIRST_UNRESERVED))
517 rtmsg_lfib(event, index, rt, nlh, net, portid, nlm_flags);
520 static void mpls_route_update(struct net *net, unsigned index,
521 struct mpls_route *new,
522 const struct nl_info *info)
524 struct mpls_route __rcu **platform_label;
525 struct mpls_route *rt;
529 platform_label = rtnl_dereference(net->mpls.platform_label);
530 rt = rtnl_dereference(platform_label[index]);
531 rcu_assign_pointer(platform_label[index], new);
533 mpls_notify_route(net, index, rt, new, info);
535 /* If we removed a route free it now */
539 static unsigned find_free_label(struct net *net)
541 struct mpls_route __rcu **platform_label;
542 size_t platform_labels;
545 platform_label = rtnl_dereference(net->mpls.platform_label);
546 platform_labels = net->mpls.platform_labels;
547 for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels;
549 if (!rtnl_dereference(platform_label[index]))
552 return LABEL_NOT_SPECIFIED;
555 #if IS_ENABLED(CONFIG_INET)
556 static struct net_device *inet_fib_lookup_dev(struct net *net,
559 struct net_device *dev;
561 struct in_addr daddr;
563 memcpy(&daddr, addr, sizeof(struct in_addr));
564 rt = ip_route_output(net, daddr.s_addr, 0, 0, 0);
576 static struct net_device *inet_fib_lookup_dev(struct net *net,
579 return ERR_PTR(-EAFNOSUPPORT);
583 #if IS_ENABLED(CONFIG_IPV6)
584 static struct net_device *inet6_fib_lookup_dev(struct net *net,
587 struct net_device *dev;
588 struct dst_entry *dst;
592 return ERR_PTR(-EAFNOSUPPORT);
594 memset(&fl6, 0, sizeof(fl6));
595 memcpy(&fl6.daddr, addr, sizeof(struct in6_addr));
596 dst = ipv6_stub->ipv6_dst_lookup_flow(net, NULL, &fl6, NULL);
598 return ERR_CAST(dst);
607 static struct net_device *inet6_fib_lookup_dev(struct net *net,
610 return ERR_PTR(-EAFNOSUPPORT);
614 static struct net_device *find_outdev(struct net *net,
615 struct mpls_route *rt,
616 struct mpls_nh *nh, int oif)
618 struct net_device *dev = NULL;
621 switch (nh->nh_via_table) {
622 case NEIGH_ARP_TABLE:
623 dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh));
626 dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh));
628 case NEIGH_LINK_TABLE:
632 dev = dev_get_by_index(net, oif);
636 return ERR_PTR(-ENODEV);
641 /* The caller is holding rtnl anyways, so release the dev reference */
647 static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
648 struct mpls_nh *nh, int oif)
650 struct net_device *dev = NULL;
653 dev = find_outdev(net, rt, nh, oif);
660 /* Ensure this is a supported device */
662 if (!mpls_dev_get(dev))
665 if ((nh->nh_via_table == NEIGH_LINK_TABLE) &&
666 (dev->addr_len != nh->nh_via_alen))
669 RCU_INIT_POINTER(nh->nh_dev, dev);
671 if (!(dev->flags & IFF_UP)) {
672 nh->nh_flags |= RTNH_F_DEAD;
676 flags = dev_get_flags(dev);
677 if (!(flags & (IFF_RUNNING | IFF_LOWER_UP)))
678 nh->nh_flags |= RTNH_F_LINKDOWN;
687 static int nla_get_via(const struct nlattr *nla, u8 *via_alen, u8 *via_table,
688 u8 via_addr[], struct netlink_ext_ack *extack)
690 struct rtvia *via = nla_data(nla);
694 if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
695 NL_SET_ERR_MSG_ATTR(extack, nla,
696 "Invalid attribute length for RTA_VIA");
699 alen = nla_len(nla) -
700 offsetof(struct rtvia, rtvia_addr);
701 if (alen > MAX_VIA_ALEN) {
702 NL_SET_ERR_MSG_ATTR(extack, nla,
703 "Invalid address length for RTA_VIA");
707 /* Validate the address family */
708 switch (via->rtvia_family) {
710 *via_table = NEIGH_LINK_TABLE;
713 *via_table = NEIGH_ARP_TABLE;
718 *via_table = NEIGH_ND_TABLE;
723 /* Unsupported address family */
727 memcpy(via_addr, via->rtvia_addr, alen);
735 static int mpls_nh_build_from_cfg(struct mpls_route_config *cfg,
736 struct mpls_route *rt)
738 struct net *net = cfg->rc_nlinfo.nl_net;
739 struct mpls_nh *nh = rt->rt_nh;
746 nh->nh_labels = cfg->rc_output_labels;
747 for (i = 0; i < nh->nh_labels; i++)
748 nh->nh_label[i] = cfg->rc_output_label[i];
750 nh->nh_via_table = cfg->rc_via_table;
751 memcpy(__mpls_nh_via(rt, nh), cfg->rc_via, cfg->rc_via_alen);
752 nh->nh_via_alen = cfg->rc_via_alen;
754 err = mpls_nh_assign_dev(net, rt, nh, cfg->rc_ifindex);
758 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
767 static int mpls_nh_build(struct net *net, struct mpls_route *rt,
768 struct mpls_nh *nh, int oif, struct nlattr *via,
769 struct nlattr *newdst, u8 max_labels,
770 struct netlink_ext_ack *extack)
778 err = nla_get_labels(newdst, max_labels, &nh->nh_labels,
779 nh->nh_label, extack);
785 err = nla_get_via(via, &nh->nh_via_alen, &nh->nh_via_table,
786 __mpls_nh_via(rt, nh), extack);
790 nh->nh_via_table = MPLS_NEIGH_TABLE_UNSPEC;
793 err = mpls_nh_assign_dev(net, rt, nh, oif);
803 static u8 mpls_count_nexthops(struct rtnexthop *rtnh, int len,
804 u8 cfg_via_alen, u8 *max_via_alen,
813 while (rtnh_ok(rtnh, remaining)) {
814 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
818 attrlen = rtnh_attrlen(rtnh);
819 nla = nla_find(attrs, attrlen, RTA_VIA);
820 if (nla && nla_len(nla) >=
821 offsetof(struct rtvia, rtvia_addr)) {
822 int via_alen = nla_len(nla) -
823 offsetof(struct rtvia, rtvia_addr);
825 if (via_alen <= MAX_VIA_ALEN)
826 *max_via_alen = max_t(u16, *max_via_alen,
830 nla = nla_find(attrs, attrlen, RTA_NEWDST);
832 nla_get_labels(nla, MAX_NEW_LABELS, &n_labels,
836 *max_labels = max_t(u8, *max_labels, n_labels);
838 /* number of nexthops is tracked by a u8.
839 * Check for overflow.
845 rtnh = rtnh_next(rtnh, &remaining);
848 /* leftover implies invalid nexthop configuration, discard it */
849 return remaining > 0 ? 0 : nhs;
852 static int mpls_nh_build_multi(struct mpls_route_config *cfg,
853 struct mpls_route *rt, u8 max_labels,
854 struct netlink_ext_ack *extack)
856 struct rtnexthop *rtnh = cfg->rc_mp;
857 struct nlattr *nla_via, *nla_newdst;
858 int remaining = cfg->rc_mp_len;
862 change_nexthops(rt) {
869 if (!rtnh_ok(rtnh, remaining))
872 /* neither weighted multipath nor any flags
875 if (rtnh->rtnh_hops || rtnh->rtnh_flags)
878 attrlen = rtnh_attrlen(rtnh);
880 struct nlattr *attrs = rtnh_attrs(rtnh);
882 nla_via = nla_find(attrs, attrlen, RTA_VIA);
883 nla_newdst = nla_find(attrs, attrlen, RTA_NEWDST);
886 err = mpls_nh_build(cfg->rc_nlinfo.nl_net, rt, nh,
887 rtnh->rtnh_ifindex, nla_via, nla_newdst,
892 if (nh->nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN))
895 rtnh = rtnh_next(rtnh, &remaining);
897 } endfor_nexthops(rt);
907 static bool mpls_label_ok(struct net *net, unsigned int *index,
908 struct netlink_ext_ack *extack)
912 /* Reserved labels may not be set */
913 if (*index < MPLS_LABEL_FIRST_UNRESERVED) {
914 NL_SET_ERR_MSG(extack,
915 "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher");
919 /* The full 20 bit range may not be supported. */
920 if (is_ok && *index >= net->mpls.platform_labels) {
921 NL_SET_ERR_MSG(extack,
922 "Label >= configured maximum in platform_labels");
926 *index = array_index_nospec(*index, net->mpls.platform_labels);
930 static int mpls_route_add(struct mpls_route_config *cfg,
931 struct netlink_ext_ack *extack)
933 struct mpls_route __rcu **platform_label;
934 struct net *net = cfg->rc_nlinfo.nl_net;
935 struct mpls_route *rt, *old;
942 index = cfg->rc_label;
944 /* If a label was not specified during insert pick one */
945 if ((index == LABEL_NOT_SPECIFIED) &&
946 (cfg->rc_nlflags & NLM_F_CREATE)) {
947 index = find_free_label(net);
950 if (!mpls_label_ok(net, &index, extack))
953 /* Append makes no sense with mpls */
955 if (cfg->rc_nlflags & NLM_F_APPEND) {
956 NL_SET_ERR_MSG(extack, "MPLS does not support route append");
961 platform_label = rtnl_dereference(net->mpls.platform_label);
962 old = rtnl_dereference(platform_label[index]);
963 if ((cfg->rc_nlflags & NLM_F_EXCL) && old)
967 if (!(cfg->rc_nlflags & NLM_F_REPLACE) && old)
971 if (!(cfg->rc_nlflags & NLM_F_CREATE) && !old)
976 nhs = mpls_count_nexthops(cfg->rc_mp, cfg->rc_mp_len,
977 cfg->rc_via_alen, &max_via_alen,
980 max_via_alen = cfg->rc_via_alen;
981 max_labels = cfg->rc_output_labels;
986 NL_SET_ERR_MSG(extack, "Route does not contain a nexthop");
991 rt = mpls_rt_alloc(nhs, max_via_alen, max_labels);
997 rt->rt_protocol = cfg->rc_protocol;
998 rt->rt_payload_type = cfg->rc_payload_type;
999 rt->rt_ttl_propagate = cfg->rc_ttl_propagate;
1002 err = mpls_nh_build_multi(cfg, rt, max_labels, extack);
1004 err = mpls_nh_build_from_cfg(cfg, rt);
1008 mpls_route_update(net, index, rt, &cfg->rc_nlinfo);
1018 static int mpls_route_del(struct mpls_route_config *cfg,
1019 struct netlink_ext_ack *extack)
1021 struct net *net = cfg->rc_nlinfo.nl_net;
1025 index = cfg->rc_label;
1027 if (!mpls_label_ok(net, &index, extack))
1030 mpls_route_update(net, index, NULL, &cfg->rc_nlinfo);
1037 static void mpls_get_stats(struct mpls_dev *mdev,
1038 struct mpls_link_stats *stats)
1040 struct mpls_pcpu_stats *p;
1043 memset(stats, 0, sizeof(*stats));
1045 for_each_possible_cpu(i) {
1046 struct mpls_link_stats local;
1049 p = per_cpu_ptr(mdev->stats, i);
1051 start = u64_stats_fetch_begin(&p->syncp);
1053 } while (u64_stats_fetch_retry(&p->syncp, start));
1055 stats->rx_packets += local.rx_packets;
1056 stats->rx_bytes += local.rx_bytes;
1057 stats->tx_packets += local.tx_packets;
1058 stats->tx_bytes += local.tx_bytes;
1059 stats->rx_errors += local.rx_errors;
1060 stats->tx_errors += local.tx_errors;
1061 stats->rx_dropped += local.rx_dropped;
1062 stats->tx_dropped += local.tx_dropped;
1063 stats->rx_noroute += local.rx_noroute;
1067 static int mpls_fill_stats_af(struct sk_buff *skb,
1068 const struct net_device *dev)
1070 struct mpls_link_stats *stats;
1071 struct mpls_dev *mdev;
1074 mdev = mpls_dev_get(dev);
1078 nla = nla_reserve_64bit(skb, MPLS_STATS_LINK,
1079 sizeof(struct mpls_link_stats),
1084 stats = nla_data(nla);
1085 mpls_get_stats(mdev, stats);
1090 static size_t mpls_get_stats_af_size(const struct net_device *dev)
1092 struct mpls_dev *mdev;
1094 mdev = mpls_dev_get(dev);
1098 return nla_total_size_64bit(sizeof(struct mpls_link_stats));
1101 static int mpls_netconf_fill_devconf(struct sk_buff *skb, struct mpls_dev *mdev,
1102 u32 portid, u32 seq, int event,
1103 unsigned int flags, int type)
1105 struct nlmsghdr *nlh;
1106 struct netconfmsg *ncm;
1109 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1114 if (type == NETCONFA_ALL)
1117 ncm = nlmsg_data(nlh);
1118 ncm->ncm_family = AF_MPLS;
1120 if (nla_put_s32(skb, NETCONFA_IFINDEX, mdev->dev->ifindex) < 0)
1121 goto nla_put_failure;
1123 if ((all || type == NETCONFA_INPUT) &&
1124 nla_put_s32(skb, NETCONFA_INPUT,
1125 mdev->input_enabled) < 0)
1126 goto nla_put_failure;
1128 nlmsg_end(skb, nlh);
1132 nlmsg_cancel(skb, nlh);
1136 static int mpls_netconf_msgsize_devconf(int type)
1138 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1139 + nla_total_size(4); /* NETCONFA_IFINDEX */
1142 if (type == NETCONFA_ALL)
1145 if (all || type == NETCONFA_INPUT)
1146 size += nla_total_size(4);
1151 static void mpls_netconf_notify_devconf(struct net *net, int event,
1152 int type, struct mpls_dev *mdev)
1154 struct sk_buff *skb;
1157 skb = nlmsg_new(mpls_netconf_msgsize_devconf(type), GFP_KERNEL);
1161 err = mpls_netconf_fill_devconf(skb, mdev, 0, 0, event, 0, type);
1163 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1164 WARN_ON(err == -EMSGSIZE);
1169 rtnl_notify(skb, net, 0, RTNLGRP_MPLS_NETCONF, NULL, GFP_KERNEL);
1173 rtnl_set_sk_err(net, RTNLGRP_MPLS_NETCONF, err);
1176 static const struct nla_policy devconf_mpls_policy[NETCONFA_MAX + 1] = {
1177 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1180 static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
1181 struct nlmsghdr *nlh,
1182 struct netlink_ext_ack *extack)
1184 struct net *net = sock_net(in_skb->sk);
1185 struct nlattr *tb[NETCONFA_MAX + 1];
1186 struct netconfmsg *ncm;
1187 struct net_device *dev;
1188 struct mpls_dev *mdev;
1189 struct sk_buff *skb;
1193 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1194 devconf_mpls_policy, NULL);
1199 if (!tb[NETCONFA_IFINDEX])
1202 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1203 dev = __dev_get_by_index(net, ifindex);
1207 mdev = mpls_dev_get(dev);
1212 skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
1216 err = mpls_netconf_fill_devconf(skb, mdev,
1217 NETLINK_CB(in_skb).portid,
1218 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1221 /* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
1222 WARN_ON(err == -EMSGSIZE);
1226 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1231 static int mpls_netconf_dump_devconf(struct sk_buff *skb,
1232 struct netlink_callback *cb)
1234 struct net *net = sock_net(skb->sk);
1235 struct hlist_head *head;
1236 struct net_device *dev;
1237 struct mpls_dev *mdev;
1242 s_idx = idx = cb->args[1];
1244 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1246 head = &net->dev_index_head[h];
1248 cb->seq = net->dev_base_seq;
1249 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1252 mdev = mpls_dev_get(dev);
1255 if (mpls_netconf_fill_devconf(skb, mdev,
1256 NETLINK_CB(cb->skb).portid,
1260 NETCONFA_ALL) < 0) {
1264 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1277 #define MPLS_PERDEV_SYSCTL_OFFSET(field) \
1278 (&((struct mpls_dev *)0)->field)
1280 static int mpls_conf_proc(struct ctl_table *ctl, int write,
1281 void __user *buffer,
1282 size_t *lenp, loff_t *ppos)
1284 int oval = *(int *)ctl->data;
1285 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1288 struct mpls_dev *mdev = ctl->extra1;
1289 int i = (int *)ctl->data - (int *)mdev;
1290 struct net *net = ctl->extra2;
1291 int val = *(int *)ctl->data;
1293 if (i == offsetof(struct mpls_dev, input_enabled) &&
1295 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF,
1296 NETCONFA_INPUT, mdev);
1303 static const struct ctl_table mpls_dev_table[] = {
1305 .procname = "input",
1306 .maxlen = sizeof(int),
1308 .proc_handler = mpls_conf_proc,
1309 .data = MPLS_PERDEV_SYSCTL_OFFSET(input_enabled),
1314 static int mpls_dev_sysctl_register(struct net_device *dev,
1315 struct mpls_dev *mdev)
1317 char path[sizeof("net/mpls/conf/") + IFNAMSIZ];
1318 struct net *net = dev_net(dev);
1319 struct ctl_table *table;
1322 table = kmemdup(&mpls_dev_table, sizeof(mpls_dev_table), GFP_KERNEL);
1326 /* Table data contains only offsets relative to the base of
1327 * the mdev at this point, so make them absolute.
1329 for (i = 0; i < ARRAY_SIZE(mpls_dev_table); i++) {
1330 table[i].data = (char *)mdev + (uintptr_t)table[i].data;
1331 table[i].extra1 = mdev;
1332 table[i].extra2 = net;
1335 snprintf(path, sizeof(path), "net/mpls/conf/%s", dev->name);
1337 mdev->sysctl = register_net_sysctl(net, path, table);
1341 mpls_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, mdev);
1350 static void mpls_dev_sysctl_unregister(struct net_device *dev,
1351 struct mpls_dev *mdev)
1353 struct net *net = dev_net(dev);
1354 struct ctl_table *table;
1356 table = mdev->sysctl->ctl_table_arg;
1357 unregister_net_sysctl_table(mdev->sysctl);
1360 mpls_netconf_notify_devconf(net, RTM_DELNETCONF, 0, mdev);
1363 static struct mpls_dev *mpls_add_dev(struct net_device *dev)
1365 struct mpls_dev *mdev;
1371 mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
1373 return ERR_PTR(err);
1375 mdev->stats = alloc_percpu(struct mpls_pcpu_stats);
1379 for_each_possible_cpu(i) {
1380 struct mpls_pcpu_stats *mpls_stats;
1382 mpls_stats = per_cpu_ptr(mdev->stats, i);
1383 u64_stats_init(&mpls_stats->syncp);
1388 err = mpls_dev_sysctl_register(dev, mdev);
1392 rcu_assign_pointer(dev->mpls_ptr, mdev);
1397 free_percpu(mdev->stats);
1399 return ERR_PTR(err);
1402 static void mpls_dev_destroy_rcu(struct rcu_head *head)
1404 struct mpls_dev *mdev = container_of(head, struct mpls_dev, rcu);
1406 free_percpu(mdev->stats);
1410 static int mpls_ifdown(struct net_device *dev, int event)
1412 struct mpls_route __rcu **platform_label;
1413 struct net *net = dev_net(dev);
1416 platform_label = rtnl_dereference(net->mpls.platform_label);
1417 for (index = 0; index < net->mpls.platform_labels; index++) {
1418 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1419 bool nh_del = false;
1425 if (event == NETDEV_UNREGISTER) {
1429 struct net_device *nh_dev =
1430 rtnl_dereference(nh->nh_dev);
1432 if (!nh_dev || nh_dev == dev)
1436 } endfor_nexthops(rt);
1438 /* if there are no more nexthops, delete the route */
1439 if (deleted == rt->rt_nhn) {
1440 mpls_route_update(net, index, NULL, NULL);
1445 size_t size = sizeof(*rt) + rt->rt_nhn *
1447 struct mpls_route *orig = rt;
1449 rt = kmalloc(size, GFP_KERNEL);
1452 memcpy(rt, orig, size);
1456 change_nexthops(rt) {
1457 unsigned int nh_flags = nh->nh_flags;
1459 if (rtnl_dereference(nh->nh_dev) != dev)
1464 case NETDEV_UNREGISTER:
1465 nh_flags |= RTNH_F_DEAD;
1468 nh_flags |= RTNH_F_LINKDOWN;
1471 if (event == NETDEV_UNREGISTER)
1472 RCU_INIT_POINTER(nh->nh_dev, NULL);
1474 if (nh->nh_flags != nh_flags)
1475 WRITE_ONCE(nh->nh_flags, nh_flags);
1477 if (!(nh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)))
1479 } endfor_nexthops(rt);
1481 WRITE_ONCE(rt->rt_nhn_alive, alive);
1484 mpls_route_update(net, index, rt, NULL);
1490 static void mpls_ifup(struct net_device *dev, unsigned int flags)
1492 struct mpls_route __rcu **platform_label;
1493 struct net *net = dev_net(dev);
1497 platform_label = rtnl_dereference(net->mpls.platform_label);
1498 for (index = 0; index < net->mpls.platform_labels; index++) {
1499 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
1505 change_nexthops(rt) {
1506 unsigned int nh_flags = nh->nh_flags;
1507 struct net_device *nh_dev =
1508 rtnl_dereference(nh->nh_dev);
1510 if (!(nh_flags & flags)) {
1518 WRITE_ONCE(nh->nh_flags, nh_flags);
1519 } endfor_nexthops(rt);
1521 WRITE_ONCE(rt->rt_nhn_alive, alive);
1525 static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
1528 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1529 struct mpls_dev *mdev;
1532 if (event == NETDEV_REGISTER) {
1533 /* For now just support Ethernet, IPGRE, SIT and IPIP devices */
1534 if (dev->type == ARPHRD_ETHER ||
1535 dev->type == ARPHRD_LOOPBACK ||
1536 dev->type == ARPHRD_IPGRE ||
1537 dev->type == ARPHRD_SIT ||
1538 dev->type == ARPHRD_TUNNEL) {
1539 mdev = mpls_add_dev(dev);
1541 return notifier_from_errno(PTR_ERR(mdev));
1546 mdev = mpls_dev_get(dev);
1554 err = mpls_ifdown(dev, event);
1556 return notifier_from_errno(err);
1559 flags = dev_get_flags(dev);
1560 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
1561 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1563 mpls_ifup(dev, RTNH_F_DEAD);
1566 flags = dev_get_flags(dev);
1567 if (flags & (IFF_RUNNING | IFF_LOWER_UP)) {
1568 mpls_ifup(dev, RTNH_F_DEAD | RTNH_F_LINKDOWN);
1570 err = mpls_ifdown(dev, event);
1572 return notifier_from_errno(err);
1575 case NETDEV_UNREGISTER:
1576 err = mpls_ifdown(dev, event);
1578 return notifier_from_errno(err);
1579 mdev = mpls_dev_get(dev);
1581 mpls_dev_sysctl_unregister(dev, mdev);
1582 RCU_INIT_POINTER(dev->mpls_ptr, NULL);
1583 call_rcu(&mdev->rcu, mpls_dev_destroy_rcu);
1586 case NETDEV_CHANGENAME:
1587 mdev = mpls_dev_get(dev);
1589 mpls_dev_sysctl_unregister(dev, mdev);
1590 err = mpls_dev_sysctl_register(dev, mdev);
1592 return notifier_from_errno(err);
1599 static struct notifier_block mpls_dev_notifier = {
1600 .notifier_call = mpls_dev_notify,
1603 static int nla_put_via(struct sk_buff *skb,
1604 u8 table, const void *addr, int alen)
1606 static const int table_to_family[NEIGH_NR_TABLES + 1] = {
1607 AF_INET, AF_INET6, AF_DECnet, AF_PACKET,
1611 int family = AF_UNSPEC;
1613 nla = nla_reserve(skb, RTA_VIA, alen + 2);
1617 if (table <= NEIGH_NR_TABLES)
1618 family = table_to_family[table];
1620 via = nla_data(nla);
1621 via->rtvia_family = family;
1622 memcpy(via->rtvia_addr, addr, alen);
1626 int nla_put_labels(struct sk_buff *skb, int attrtype,
1627 u8 labels, const u32 label[])
1630 struct mpls_shim_hdr *nla_label;
1633 nla = nla_reserve(skb, attrtype, labels*4);
1637 nla_label = nla_data(nla);
1639 for (i = labels - 1; i >= 0; i--) {
1640 nla_label[i] = mpls_entry_encode(label[i], 0, 0, bos);
1646 EXPORT_SYMBOL_GPL(nla_put_labels);
1648 int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
1649 u32 label[], struct netlink_ext_ack *extack)
1651 unsigned len = nla_len(nla);
1652 struct mpls_shim_hdr *nla_label;
1657 /* len needs to be an even multiple of 4 (the label size). Number
1658 * of labels is a u8 so check for overflow.
1660 if (len & 3 || len / 4 > 255) {
1661 NL_SET_ERR_MSG_ATTR(extack, nla,
1662 "Invalid length for labels attribute");
1666 /* Limit the number of new labels allowed */
1668 if (nla_labels > max_labels) {
1669 NL_SET_ERR_MSG(extack, "Too many labels");
1673 /* when label == NULL, caller wants number of labels */
1677 nla_label = nla_data(nla);
1679 for (i = nla_labels - 1; i >= 0; i--, bos = false) {
1680 struct mpls_entry_decoded dec;
1681 dec = mpls_entry_decode(nla_label + i);
1683 /* Ensure the bottom of stack flag is properly set
1684 * and ttl and tc are both clear.
1687 NL_SET_ERR_MSG_ATTR(extack, nla,
1688 "TTL in label must be 0");
1693 NL_SET_ERR_MSG_ATTR(extack, nla,
1694 "Traffic class in label must be 0");
1698 if (dec.bos != bos) {
1699 NL_SET_BAD_ATTR(extack, nla);
1701 NL_SET_ERR_MSG(extack,
1702 "BOS bit must be set in first label");
1704 NL_SET_ERR_MSG(extack,
1705 "BOS bit can only be set in first label");
1710 switch (dec.label) {
1711 case MPLS_LABEL_IMPLNULL:
1712 /* RFC3032: This is a label that an LSR may
1713 * assign and distribute, but which never
1714 * actually appears in the encapsulation.
1716 NL_SET_ERR_MSG_ATTR(extack, nla,
1717 "Implicit NULL Label (3) can not be used in encapsulation");
1721 label[i] = dec.label;
1724 *labels = nla_labels;
1727 EXPORT_SYMBOL_GPL(nla_get_labels);
1729 static int rtm_to_route_config(struct sk_buff *skb,
1730 struct nlmsghdr *nlh,
1731 struct mpls_route_config *cfg,
1732 struct netlink_ext_ack *extack)
1735 struct nlattr *tb[RTA_MAX+1];
1739 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_mpls_policy,
1745 rtm = nlmsg_data(nlh);
1747 if (rtm->rtm_family != AF_MPLS) {
1748 NL_SET_ERR_MSG(extack, "Invalid address family in rtmsg");
1751 if (rtm->rtm_dst_len != 20) {
1752 NL_SET_ERR_MSG(extack, "rtm_dst_len must be 20 for MPLS");
1755 if (rtm->rtm_src_len != 0) {
1756 NL_SET_ERR_MSG(extack, "rtm_src_len must be 0 for MPLS");
1759 if (rtm->rtm_tos != 0) {
1760 NL_SET_ERR_MSG(extack, "rtm_tos must be 0 for MPLS");
1763 if (rtm->rtm_table != RT_TABLE_MAIN) {
1764 NL_SET_ERR_MSG(extack,
1765 "MPLS only supports the main route table");
1768 /* Any value is acceptable for rtm_protocol */
1770 /* As mpls uses destination specific addresses
1771 * (or source specific address in the case of multicast)
1772 * all addresses have universal scope.
1774 if (rtm->rtm_scope != RT_SCOPE_UNIVERSE) {
1775 NL_SET_ERR_MSG(extack,
1776 "Invalid route scope - MPLS only supports UNIVERSE");
1779 if (rtm->rtm_type != RTN_UNICAST) {
1780 NL_SET_ERR_MSG(extack,
1781 "Invalid route type - MPLS only supports UNICAST");
1784 if (rtm->rtm_flags != 0) {
1785 NL_SET_ERR_MSG(extack, "rtm_flags must be 0 for MPLS");
1789 cfg->rc_label = LABEL_NOT_SPECIFIED;
1790 cfg->rc_protocol = rtm->rtm_protocol;
1791 cfg->rc_via_table = MPLS_NEIGH_TABLE_UNSPEC;
1792 cfg->rc_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
1793 cfg->rc_nlflags = nlh->nlmsg_flags;
1794 cfg->rc_nlinfo.portid = NETLINK_CB(skb).portid;
1795 cfg->rc_nlinfo.nlh = nlh;
1796 cfg->rc_nlinfo.nl_net = sock_net(skb->sk);
1798 for (index = 0; index <= RTA_MAX; index++) {
1799 struct nlattr *nla = tb[index];
1805 cfg->rc_ifindex = nla_get_u32(nla);
1808 if (nla_get_labels(nla, MAX_NEW_LABELS,
1809 &cfg->rc_output_labels,
1810 cfg->rc_output_label, extack))
1816 if (nla_get_labels(nla, 1, &label_count,
1817 &cfg->rc_label, extack))
1820 if (!mpls_label_ok(cfg->rc_nlinfo.nl_net,
1821 &cfg->rc_label, extack))
1826 NL_SET_ERR_MSG(extack, "MPLS does not support RTA_GATEWAY attribute");
1830 if (nla_get_via(nla, &cfg->rc_via_alen,
1831 &cfg->rc_via_table, cfg->rc_via,
1838 cfg->rc_mp = nla_data(nla);
1839 cfg->rc_mp_len = nla_len(nla);
1842 case RTA_TTL_PROPAGATE:
1844 u8 ttl_propagate = nla_get_u8(nla);
1846 if (ttl_propagate > 1) {
1847 NL_SET_ERR_MSG_ATTR(extack, nla,
1848 "RTA_TTL_PROPAGATE can only be 0 or 1");
1851 cfg->rc_ttl_propagate = ttl_propagate ?
1852 MPLS_TTL_PROP_ENABLED :
1853 MPLS_TTL_PROP_DISABLED;
1857 NL_SET_ERR_MSG_ATTR(extack, nla, "Unknown attribute");
1858 /* Unsupported attribute */
1868 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1869 struct netlink_ext_ack *extack)
1871 struct mpls_route_config *cfg;
1874 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1878 err = rtm_to_route_config(skb, nlh, cfg, extack);
1882 err = mpls_route_del(cfg, extack);
1890 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
1891 struct netlink_ext_ack *extack)
1893 struct mpls_route_config *cfg;
1896 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1900 err = rtm_to_route_config(skb, nlh, cfg, extack);
1904 err = mpls_route_add(cfg, extack);
1911 static int mpls_dump_route(struct sk_buff *skb, u32 portid, u32 seq, int event,
1912 u32 label, struct mpls_route *rt, int flags)
1914 struct net_device *dev;
1915 struct nlmsghdr *nlh;
1918 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1922 rtm = nlmsg_data(nlh);
1923 rtm->rtm_family = AF_MPLS;
1924 rtm->rtm_dst_len = 20;
1925 rtm->rtm_src_len = 0;
1927 rtm->rtm_table = RT_TABLE_MAIN;
1928 rtm->rtm_protocol = rt->rt_protocol;
1929 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
1930 rtm->rtm_type = RTN_UNICAST;
1933 if (nla_put_labels(skb, RTA_DST, 1, &label))
1934 goto nla_put_failure;
1936 if (rt->rt_ttl_propagate != MPLS_TTL_PROP_DEFAULT) {
1937 bool ttl_propagate =
1938 rt->rt_ttl_propagate == MPLS_TTL_PROP_ENABLED;
1940 if (nla_put_u8(skb, RTA_TTL_PROPAGATE,
1942 goto nla_put_failure;
1944 if (rt->rt_nhn == 1) {
1945 const struct mpls_nh *nh = rt->rt_nh;
1947 if (nh->nh_labels &&
1948 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
1950 goto nla_put_failure;
1951 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
1952 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
1954 goto nla_put_failure;
1955 dev = rtnl_dereference(nh->nh_dev);
1956 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
1957 goto nla_put_failure;
1958 if (nh->nh_flags & RTNH_F_LINKDOWN)
1959 rtm->rtm_flags |= RTNH_F_LINKDOWN;
1960 if (nh->nh_flags & RTNH_F_DEAD)
1961 rtm->rtm_flags |= RTNH_F_DEAD;
1963 struct rtnexthop *rtnh;
1968 mp = nla_nest_start(skb, RTA_MULTIPATH);
1970 goto nla_put_failure;
1973 dev = rtnl_dereference(nh->nh_dev);
1977 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
1979 goto nla_put_failure;
1981 rtnh->rtnh_ifindex = dev->ifindex;
1982 if (nh->nh_flags & RTNH_F_LINKDOWN) {
1983 rtnh->rtnh_flags |= RTNH_F_LINKDOWN;
1986 if (nh->nh_flags & RTNH_F_DEAD) {
1987 rtnh->rtnh_flags |= RTNH_F_DEAD;
1991 if (nh->nh_labels && nla_put_labels(skb, RTA_NEWDST,
1994 goto nla_put_failure;
1995 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
1996 nla_put_via(skb, nh->nh_via_table,
1997 mpls_nh_via(rt, nh),
1999 goto nla_put_failure;
2001 /* length of rtnetlink header + attributes */
2002 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
2003 } endfor_nexthops(rt);
2005 if (linkdown == rt->rt_nhn)
2006 rtm->rtm_flags |= RTNH_F_LINKDOWN;
2007 if (dead == rt->rt_nhn)
2008 rtm->rtm_flags |= RTNH_F_DEAD;
2010 nla_nest_end(skb, mp);
2013 nlmsg_end(skb, nlh);
2017 nlmsg_cancel(skb, nlh);
2021 static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
2023 struct net *net = sock_net(skb->sk);
2024 struct mpls_route __rcu **platform_label;
2025 size_t platform_labels;
2030 index = cb->args[0];
2031 if (index < MPLS_LABEL_FIRST_UNRESERVED)
2032 index = MPLS_LABEL_FIRST_UNRESERVED;
2034 platform_label = rtnl_dereference(net->mpls.platform_label);
2035 platform_labels = net->mpls.platform_labels;
2036 for (; index < platform_labels; index++) {
2037 struct mpls_route *rt;
2038 rt = rtnl_dereference(platform_label[index]);
2042 if (mpls_dump_route(skb, NETLINK_CB(cb->skb).portid,
2043 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2044 index, rt, NLM_F_MULTI) < 0)
2047 cb->args[0] = index;
2052 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
2055 NLMSG_ALIGN(sizeof(struct rtmsg))
2056 + nla_total_size(4) /* RTA_DST */
2057 + nla_total_size(1); /* RTA_TTL_PROPAGATE */
2059 if (rt->rt_nhn == 1) {
2060 struct mpls_nh *nh = rt->rt_nh;
2063 payload += nla_total_size(4); /* RTA_OIF */
2064 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC) /* RTA_VIA */
2065 payload += nla_total_size(2 + nh->nh_via_alen);
2066 if (nh->nh_labels) /* RTA_NEWDST */
2067 payload += nla_total_size(nh->nh_labels * 4);
2069 /* each nexthop is packed in an attribute */
2073 if (!rtnl_dereference(nh->nh_dev))
2075 nhsize += nla_total_size(sizeof(struct rtnexthop));
2077 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC)
2078 nhsize += nla_total_size(2 + nh->nh_via_alen);
2080 nhsize += nla_total_size(nh->nh_labels * 4);
2081 } endfor_nexthops(rt);
2082 /* nested attribute */
2083 payload += nla_total_size(nhsize);
2089 static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
2090 struct nlmsghdr *nlh, struct net *net, u32 portid,
2091 unsigned int nlm_flags)
2093 struct sk_buff *skb;
2094 u32 seq = nlh ? nlh->nlmsg_seq : 0;
2097 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2101 err = mpls_dump_route(skb, portid, seq, event, label, rt, nlm_flags);
2103 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2104 WARN_ON(err == -EMSGSIZE);
2108 rtnl_notify(skb, net, portid, RTNLGRP_MPLS_ROUTE, nlh, GFP_KERNEL);
2113 rtnl_set_sk_err(net, RTNLGRP_MPLS_ROUTE, err);
2116 static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
2117 struct netlink_ext_ack *extack)
2119 struct net *net = sock_net(in_skb->sk);
2120 u32 portid = NETLINK_CB(in_skb).portid;
2121 u32 in_label = LABEL_NOT_SPECIFIED;
2122 struct nlattr *tb[RTA_MAX + 1];
2123 u32 labels[MAX_NEW_LABELS];
2124 struct mpls_shim_hdr *hdr;
2125 unsigned int hdr_size = 0;
2126 struct net_device *dev;
2127 struct mpls_route *rt;
2128 struct rtmsg *rtm, *r;
2129 struct nlmsghdr *nlh;
2130 struct sk_buff *skb;
2135 err = nlmsg_parse(in_nlh, sizeof(*rtm), tb, RTA_MAX,
2136 rtm_mpls_policy, extack);
2140 rtm = nlmsg_data(in_nlh);
2145 if (nla_get_labels(tb[RTA_DST], 1, &label_count,
2146 &in_label, extack)) {
2151 if (!mpls_label_ok(net, &in_label, extack)) {
2157 rt = mpls_route_input_rcu(net, in_label);
2163 if (rtm->rtm_flags & RTM_F_FIB_MATCH) {
2164 skb = nlmsg_new(lfib_nlmsg_size(rt), GFP_KERNEL);
2170 err = mpls_dump_route(skb, portid, in_nlh->nlmsg_seq,
2171 RTM_NEWROUTE, in_label, rt, 0);
2173 /* -EMSGSIZE implies BUG in lfib_nlmsg_size */
2174 WARN_ON(err == -EMSGSIZE);
2178 return rtnl_unicast(skb, net, portid);
2181 if (tb[RTA_NEWDST]) {
2182 if (nla_get_labels(tb[RTA_NEWDST], MAX_NEW_LABELS, &n_labels,
2183 labels, extack) != 0) {
2188 hdr_size = n_labels * sizeof(struct mpls_shim_hdr);
2191 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2197 skb->protocol = htons(ETH_P_MPLS_UC);
2203 if (skb_cow(skb, hdr_size)) {
2208 skb_reserve(skb, hdr_size);
2209 skb_push(skb, hdr_size);
2210 skb_reset_network_header(skb);
2212 /* Push new labels */
2213 hdr = mpls_hdr(skb);
2215 for (i = n_labels - 1; i >= 0; i--) {
2216 hdr[i] = mpls_entry_encode(labels[i],
2222 nh = mpls_select_multipath(rt, skb);
2229 skb_pull(skb, hdr_size);
2230 skb_reset_network_header(skb);
2233 nlh = nlmsg_put(skb, portid, in_nlh->nlmsg_seq,
2234 RTM_NEWROUTE, sizeof(*r), 0);
2240 r = nlmsg_data(nlh);
2241 r->rtm_family = AF_MPLS;
2242 r->rtm_dst_len = 20;
2244 r->rtm_table = RT_TABLE_MAIN;
2245 r->rtm_type = RTN_UNICAST;
2246 r->rtm_scope = RT_SCOPE_UNIVERSE;
2247 r->rtm_protocol = rt->rt_protocol;
2250 if (nla_put_labels(skb, RTA_DST, 1, &in_label))
2251 goto nla_put_failure;
2253 if (nh->nh_labels &&
2254 nla_put_labels(skb, RTA_NEWDST, nh->nh_labels,
2256 goto nla_put_failure;
2258 if (nh->nh_via_table != MPLS_NEIGH_TABLE_UNSPEC &&
2259 nla_put_via(skb, nh->nh_via_table, mpls_nh_via(rt, nh),
2261 goto nla_put_failure;
2262 dev = rtnl_dereference(nh->nh_dev);
2263 if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex))
2264 goto nla_put_failure;
2266 nlmsg_end(skb, nlh);
2268 err = rtnl_unicast(skb, net, portid);
2273 nlmsg_cancel(skb, nlh);
2280 static int resize_platform_label_table(struct net *net, size_t limit)
2282 size_t size = sizeof(struct mpls_route *) * limit;
2285 struct mpls_route __rcu **labels = NULL, **old;
2286 struct mpls_route *rt0 = NULL, *rt2 = NULL;
2290 labels = kvzalloc(size, GFP_KERNEL);
2295 /* In case the predefined labels need to be populated */
2296 if (limit > MPLS_LABEL_IPV4NULL) {
2297 struct net_device *lo = net->loopback_dev;
2298 rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
2301 RCU_INIT_POINTER(rt0->rt_nh->nh_dev, lo);
2302 rt0->rt_protocol = RTPROT_KERNEL;
2303 rt0->rt_payload_type = MPT_IPV4;
2304 rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2305 rt0->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2306 rt0->rt_nh->nh_via_alen = lo->addr_len;
2307 memcpy(__mpls_nh_via(rt0, rt0->rt_nh), lo->dev_addr,
2310 if (limit > MPLS_LABEL_IPV6NULL) {
2311 struct net_device *lo = net->loopback_dev;
2312 rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
2315 RCU_INIT_POINTER(rt2->rt_nh->nh_dev, lo);
2316 rt2->rt_protocol = RTPROT_KERNEL;
2317 rt2->rt_payload_type = MPT_IPV6;
2318 rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
2319 rt2->rt_nh->nh_via_table = NEIGH_LINK_TABLE;
2320 rt2->rt_nh->nh_via_alen = lo->addr_len;
2321 memcpy(__mpls_nh_via(rt2, rt2->rt_nh), lo->dev_addr,
2326 /* Remember the original table */
2327 old = rtnl_dereference(net->mpls.platform_label);
2328 old_limit = net->mpls.platform_labels;
2330 /* Free any labels beyond the new table */
2331 for (index = limit; index < old_limit; index++)
2332 mpls_route_update(net, index, NULL, NULL);
2334 /* Copy over the old labels */
2336 if (old_limit < limit)
2337 cp_size = old_limit * sizeof(struct mpls_route *);
2339 memcpy(labels, old, cp_size);
2341 /* If needed set the predefined labels */
2342 if ((old_limit <= MPLS_LABEL_IPV6NULL) &&
2343 (limit > MPLS_LABEL_IPV6NULL)) {
2344 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV6NULL], rt2);
2348 if ((old_limit <= MPLS_LABEL_IPV4NULL) &&
2349 (limit > MPLS_LABEL_IPV4NULL)) {
2350 RCU_INIT_POINTER(labels[MPLS_LABEL_IPV4NULL], rt0);
2354 /* Update the global pointers */
2355 net->mpls.platform_labels = limit;
2356 rcu_assign_pointer(net->mpls.platform_label, labels);
2377 static int mpls_platform_labels(struct ctl_table *table, int write,
2378 void __user *buffer, size_t *lenp, loff_t *ppos)
2380 struct net *net = table->data;
2381 int platform_labels = net->mpls.platform_labels;
2383 struct ctl_table tmp = {
2384 .procname = table->procname,
2385 .data = &platform_labels,
2386 .maxlen = sizeof(int),
2387 .mode = table->mode,
2389 .extra2 = &label_limit,
2392 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
2394 if (write && ret == 0)
2395 ret = resize_platform_label_table(net, platform_labels);
2400 #define MPLS_NS_SYSCTL_OFFSET(field) \
2401 (&((struct net *)0)->field)
2403 static const struct ctl_table mpls_table[] = {
2405 .procname = "platform_labels",
2407 .maxlen = sizeof(int),
2409 .proc_handler = mpls_platform_labels,
2412 .procname = "ip_ttl_propagate",
2413 .data = MPLS_NS_SYSCTL_OFFSET(mpls.ip_ttl_propagate),
2414 .maxlen = sizeof(int),
2416 .proc_handler = proc_dointvec_minmax,
2421 .procname = "default_ttl",
2422 .data = MPLS_NS_SYSCTL_OFFSET(mpls.default_ttl),
2423 .maxlen = sizeof(int),
2425 .proc_handler = proc_dointvec_minmax,
2432 static int mpls_net_init(struct net *net)
2434 struct ctl_table *table;
2437 net->mpls.platform_labels = 0;
2438 net->mpls.platform_label = NULL;
2439 net->mpls.ip_ttl_propagate = 1;
2440 net->mpls.default_ttl = 255;
2442 table = kmemdup(mpls_table, sizeof(mpls_table), GFP_KERNEL);
2446 /* Table data contains only offsets relative to the base of
2447 * the mdev at this point, so make them absolute.
2449 for (i = 0; i < ARRAY_SIZE(mpls_table) - 1; i++)
2450 table[i].data = (char *)net + (uintptr_t)table[i].data;
2452 net->mpls.ctl = register_net_sysctl(net, "net/mpls", table);
2453 if (net->mpls.ctl == NULL) {
2461 static void mpls_net_exit(struct net *net)
2463 struct mpls_route __rcu **platform_label;
2464 size_t platform_labels;
2465 struct ctl_table *table;
2468 table = net->mpls.ctl->ctl_table_arg;
2469 unregister_net_sysctl_table(net->mpls.ctl);
2472 /* An rcu grace period has passed since there was a device in
2473 * the network namespace (and thus the last in flight packet)
2474 * left this network namespace. This is because
2475 * unregister_netdevice_many and netdev_run_todo has completed
2476 * for each network device that was in this network namespace.
2478 * As such no additional rcu synchronization is necessary when
2479 * freeing the platform_label table.
2482 platform_label = rtnl_dereference(net->mpls.platform_label);
2483 platform_labels = net->mpls.platform_labels;
2484 for (index = 0; index < platform_labels; index++) {
2485 struct mpls_route *rt = rtnl_dereference(platform_label[index]);
2486 RCU_INIT_POINTER(platform_label[index], NULL);
2487 mpls_notify_route(net, index, rt, NULL, NULL);
2492 kvfree(platform_label);
2495 static struct pernet_operations mpls_net_ops = {
2496 .init = mpls_net_init,
2497 .exit = mpls_net_exit,
2500 static struct rtnl_af_ops mpls_af_ops __read_mostly = {
2502 .fill_stats_af = mpls_fill_stats_af,
2503 .get_stats_af_size = mpls_get_stats_af_size,
2506 static int __init mpls_init(void)
2510 BUILD_BUG_ON(sizeof(struct mpls_shim_hdr) != 4);
2512 err = register_pernet_subsys(&mpls_net_ops);
2516 err = register_netdevice_notifier(&mpls_dev_notifier);
2518 goto out_unregister_pernet;
2520 dev_add_pack(&mpls_packet_type);
2522 rtnl_af_register(&mpls_af_ops);
2524 rtnl_register(PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, 0);
2525 rtnl_register(PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, 0);
2526 rtnl_register(PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes,
2528 rtnl_register(PF_MPLS, RTM_GETNETCONF, mpls_netconf_get_devconf,
2529 mpls_netconf_dump_devconf, 0);
2534 out_unregister_pernet:
2535 unregister_pernet_subsys(&mpls_net_ops);
2538 module_init(mpls_init);
2540 static void __exit mpls_exit(void)
2542 rtnl_unregister_all(PF_MPLS);
2543 rtnl_af_unregister(&mpls_af_ops);
2544 dev_remove_pack(&mpls_packet_type);
2545 unregister_netdevice_notifier(&mpls_dev_notifier);
2546 unregister_pernet_subsys(&mpls_net_ops);
2548 module_exit(mpls_exit);
2550 MODULE_DESCRIPTION("MultiProtocol Label Switching");
2551 MODULE_LICENSE("GPL v2");
2552 MODULE_ALIAS_NETPROTO(PF_MPLS);