1 // SPDX-License-Identifier: GPL-2.0
2 /* Generic nexthop implementation
4 * Copyright (c) 2017-19 Cumulus Networks
5 * Copyright (c) 2017-19 David Ahern <dsa@cumulusnetworks.com>
8 #include <linux/nexthop.h>
9 #include <linux/rtnetlink.h>
10 #include <linux/slab.h>
12 #include <net/ipv6_stubs.h>
13 #include <net/lwtunnel.h>
14 #include <net/ndisc.h>
15 #include <net/nexthop.h>
16 #include <net/route.h>
19 static void remove_nexthop(struct net *net, struct nexthop *nh,
20 struct nl_info *nlinfo);
22 #define NH_DEV_HASHBITS 8
23 #define NH_DEV_HASHSIZE (1U << NH_DEV_HASHBITS)
25 static const struct nla_policy rtm_nh_policy[NHA_MAX + 1] = {
26 [NHA_UNSPEC] = { .strict_start_type = NHA_UNSPEC + 1 },
27 [NHA_ID] = { .type = NLA_U32 },
28 [NHA_GROUP] = { .type = NLA_BINARY },
29 [NHA_GROUP_TYPE] = { .type = NLA_U16 },
30 [NHA_BLACKHOLE] = { .type = NLA_FLAG },
31 [NHA_OIF] = { .type = NLA_U32 },
32 [NHA_GATEWAY] = { .type = NLA_BINARY },
33 [NHA_ENCAP_TYPE] = { .type = NLA_U16 },
34 [NHA_ENCAP] = { .type = NLA_NESTED },
35 [NHA_GROUPS] = { .type = NLA_FLAG },
36 [NHA_MASTER] = { .type = NLA_U32 },
39 static unsigned int nh_dev_hashfn(unsigned int val)
41 unsigned int mask = NH_DEV_HASHSIZE - 1;
44 (val >> NH_DEV_HASHBITS) ^
45 (val >> (NH_DEV_HASHBITS * 2))) & mask;
48 static void nexthop_devhash_add(struct net *net, struct nh_info *nhi)
50 struct net_device *dev = nhi->fib_nhc.nhc_dev;
51 struct hlist_head *head;
56 hash = nh_dev_hashfn(dev->ifindex);
57 head = &net->nexthop.devhash[hash];
58 hlist_add_head(&nhi->dev_hash, head);
61 static void nexthop_free_mpath(struct nexthop *nh)
66 nhg = rcu_dereference_raw(nh->nh_grp);
67 for (i = 0; i < nhg->num_nh; ++i)
68 WARN_ON(nhg->nh_entries[i].nh);
73 static void nexthop_free_single(struct nexthop *nh)
77 nhi = rcu_dereference_raw(nh->nh_info);
78 switch (nhi->family) {
80 fib_nh_release(nh->net, &nhi->fib_nh);
83 ipv6_stub->fib6_nh_release(&nhi->fib6_nh);
89 void nexthop_free_rcu(struct rcu_head *head)
91 struct nexthop *nh = container_of(head, struct nexthop, rcu);
94 nexthop_free_mpath(nh);
96 nexthop_free_single(nh);
100 EXPORT_SYMBOL_GPL(nexthop_free_rcu);
102 static struct nexthop *nexthop_alloc(void)
106 nh = kzalloc(sizeof(struct nexthop), GFP_KERNEL);
108 INIT_LIST_HEAD(&nh->fi_list);
109 INIT_LIST_HEAD(&nh->grp_list);
114 static struct nh_group *nexthop_grp_alloc(u16 num_nh)
116 size_t sz = offsetof(struct nexthop, nh_grp)
117 + sizeof(struct nh_group)
118 + sizeof(struct nh_grp_entry) * num_nh;
119 struct nh_group *nhg;
121 nhg = kzalloc(sz, GFP_KERNEL);
123 nhg->num_nh = num_nh;
128 static void nh_base_seq_inc(struct net *net)
130 while (++net->nexthop.seq == 0)
134 /* no reference taken; rcu lock or rtnl must be held */
135 struct nexthop *nexthop_find_by_id(struct net *net, u32 id)
137 struct rb_node **pp, *parent = NULL, *next;
139 pp = &net->nexthop.rb_root.rb_node;
143 next = rcu_dereference_raw(*pp);
148 nh = rb_entry(parent, struct nexthop, rb_node);
151 else if (id > nh->id)
152 pp = &next->rb_right;
158 EXPORT_SYMBOL_GPL(nexthop_find_by_id);
160 /* used for auto id allocation; called with rtnl held */
161 static u32 nh_find_unused_id(struct net *net)
163 u32 id_start = net->nexthop.last_id_allocated;
166 net->nexthop.last_id_allocated++;
167 if (net->nexthop.last_id_allocated == id_start)
170 if (!nexthop_find_by_id(net, net->nexthop.last_id_allocated))
171 return net->nexthop.last_id_allocated;
176 static int nla_put_nh_group(struct sk_buff *skb, struct nh_group *nhg)
178 struct nexthop_grp *p;
179 size_t len = nhg->num_nh * sizeof(*p);
185 group_type = NEXTHOP_GRP_TYPE_MPATH;
187 if (nla_put_u16(skb, NHA_GROUP_TYPE, group_type))
188 goto nla_put_failure;
190 nla = nla_reserve(skb, NHA_GROUP, len);
192 goto nla_put_failure;
195 for (i = 0; i < nhg->num_nh; ++i) {
196 p->id = nhg->nh_entries[i].nh->id;
197 p->weight = nhg->nh_entries[i].weight - 1;
207 static int nh_fill_node(struct sk_buff *skb, struct nexthop *nh,
208 int event, u32 portid, u32 seq, unsigned int nlflags)
210 struct fib6_nh *fib6_nh;
211 struct fib_nh *fib_nh;
212 struct nlmsghdr *nlh;
216 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nhm), nlflags);
220 nhm = nlmsg_data(nlh);
221 nhm->nh_family = AF_UNSPEC;
222 nhm->nh_flags = nh->nh_flags;
223 nhm->nh_protocol = nh->protocol;
227 if (nla_put_u32(skb, NHA_ID, nh->id))
228 goto nla_put_failure;
231 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
233 if (nla_put_nh_group(skb, nhg))
234 goto nla_put_failure;
238 nhi = rtnl_dereference(nh->nh_info);
239 nhm->nh_family = nhi->family;
240 if (nhi->reject_nh) {
241 if (nla_put_flag(skb, NHA_BLACKHOLE))
242 goto nla_put_failure;
245 const struct net_device *dev;
247 dev = nhi->fib_nhc.nhc_dev;
248 if (dev && nla_put_u32(skb, NHA_OIF, dev->ifindex))
249 goto nla_put_failure;
252 nhm->nh_scope = nhi->fib_nhc.nhc_scope;
253 switch (nhi->family) {
255 fib_nh = &nhi->fib_nh;
256 if (fib_nh->fib_nh_gw_family &&
257 nla_put_u32(skb, NHA_GATEWAY, fib_nh->fib_nh_gw4))
258 goto nla_put_failure;
262 fib6_nh = &nhi->fib6_nh;
263 if (fib6_nh->fib_nh_gw_family &&
264 nla_put_in6_addr(skb, NHA_GATEWAY, &fib6_nh->fib_nh_gw6))
265 goto nla_put_failure;
269 if (nhi->fib_nhc.nhc_lwtstate &&
270 lwtunnel_fill_encap(skb, nhi->fib_nhc.nhc_lwtstate,
271 NHA_ENCAP, NHA_ENCAP_TYPE) < 0)
272 goto nla_put_failure;
282 static size_t nh_nlmsg_size_grp(struct nexthop *nh)
284 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
285 size_t sz = sizeof(struct nexthop_grp) * nhg->num_nh;
287 return nla_total_size(sz) +
288 nla_total_size(2); /* NHA_GROUP_TYPE */
291 static size_t nh_nlmsg_size_single(struct nexthop *nh)
293 struct nh_info *nhi = rtnl_dereference(nh->nh_info);
296 /* covers NHA_BLACKHOLE since NHA_OIF and BLACKHOLE
297 * are mutually exclusive
299 sz = nla_total_size(4); /* NHA_OIF */
301 switch (nhi->family) {
303 if (nhi->fib_nh.fib_nh_gw_family)
304 sz += nla_total_size(4); /* NHA_GATEWAY */
309 if (nhi->fib6_nh.fib_nh_gw_family)
310 sz += nla_total_size(sizeof(const struct in6_addr));
314 if (nhi->fib_nhc.nhc_lwtstate) {
315 sz += lwtunnel_get_encap_size(nhi->fib_nhc.nhc_lwtstate);
316 sz += nla_total_size(2); /* NHA_ENCAP_TYPE */
322 static size_t nh_nlmsg_size(struct nexthop *nh)
324 size_t sz = nla_total_size(4); /* NHA_ID */
327 sz += nh_nlmsg_size_grp(nh);
329 sz += nh_nlmsg_size_single(nh);
334 static void nexthop_notify(int event, struct nexthop *nh, struct nl_info *info)
336 unsigned int nlflags = info->nlh ? info->nlh->nlmsg_flags : 0;
337 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
341 skb = nlmsg_new(nh_nlmsg_size(nh), gfp_any());
345 err = nh_fill_node(skb, nh, event, info->portid, seq, nlflags);
347 /* -EMSGSIZE implies BUG in nh_nlmsg_size() */
348 WARN_ON(err == -EMSGSIZE);
353 rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_NEXTHOP,
354 info->nlh, gfp_any());
358 rtnl_set_sk_err(info->nl_net, RTNLGRP_NEXTHOP, err);
361 static bool valid_group_nh(struct nexthop *nh, unsigned int npaths,
362 struct netlink_ext_ack *extack)
365 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
367 /* nested multipath (group within a group) is not
371 NL_SET_ERR_MSG(extack,
372 "Multipath group can not be a nexthop within a group");
376 struct nh_info *nhi = rtnl_dereference(nh->nh_info);
378 if (nhi->reject_nh && npaths > 1) {
379 NL_SET_ERR_MSG(extack,
380 "Blackhole nexthop can not be used in a group with more than 1 path");
388 static int nh_check_attr_group(struct net *net, struct nlattr *tb[],
389 struct netlink_ext_ack *extack)
391 unsigned int len = nla_len(tb[NHA_GROUP]);
392 struct nexthop_grp *nhg;
395 if (len & (sizeof(struct nexthop_grp) - 1)) {
396 NL_SET_ERR_MSG(extack,
397 "Invalid length for nexthop group attribute");
401 /* convert len to number of nexthop ids */
404 nhg = nla_data(tb[NHA_GROUP]);
405 for (i = 0; i < len; ++i) {
406 if (nhg[i].resvd1 || nhg[i].resvd2) {
407 NL_SET_ERR_MSG(extack, "Reserved fields in nexthop_grp must be 0");
410 if (nhg[i].weight > 254) {
411 NL_SET_ERR_MSG(extack, "Invalid value for weight");
414 for (j = i + 1; j < len; ++j) {
415 if (nhg[i].id == nhg[j].id) {
416 NL_SET_ERR_MSG(extack, "Nexthop id can not be used twice in a group");
422 nhg = nla_data(tb[NHA_GROUP]);
423 for (i = 0; i < len; ++i) {
426 nh = nexthop_find_by_id(net, nhg[i].id);
428 NL_SET_ERR_MSG(extack, "Invalid nexthop id");
431 if (!valid_group_nh(nh, len, extack))
434 for (i = NHA_GROUP + 1; i < __NHA_MAX; ++i) {
438 NL_SET_ERR_MSG(extack,
439 "No other attributes can be set in nexthop groups");
446 static bool ipv6_good_nh(const struct fib6_nh *nh)
448 int state = NUD_REACHABLE;
453 n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev, &nh->fib_nh_gw6);
455 state = n->nud_state;
457 rcu_read_unlock_bh();
459 return !!(state & NUD_VALID);
462 static bool ipv4_good_nh(const struct fib_nh *nh)
464 int state = NUD_REACHABLE;
469 n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
470 (__force u32)nh->fib_nh_gw4);
472 state = n->nud_state;
474 rcu_read_unlock_bh();
476 return !!(state & NUD_VALID);
479 struct nexthop *nexthop_select_path(struct nexthop *nh, int hash)
481 struct nexthop *rc = NULL;
482 struct nh_group *nhg;
488 nhg = rcu_dereference(nh->nh_grp);
489 for (i = 0; i < nhg->num_nh; ++i) {
490 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
493 if (hash > atomic_read(&nhge->upper_bound))
496 /* nexthops always check if it is good and does
497 * not rely on a sysctl for this behavior
499 nhi = rcu_dereference(nhge->nh->nh_info);
500 switch (nhi->family) {
502 if (ipv4_good_nh(&nhi->fib_nh))
506 if (ipv6_good_nh(&nhi->fib6_nh))
517 EXPORT_SYMBOL_GPL(nexthop_select_path);
519 static int nexthop_check_scope(struct nexthop *nh, u8 scope,
520 struct netlink_ext_ack *extack)
524 nhi = rtnl_dereference(nh->nh_info);
525 if (scope == RT_SCOPE_HOST && nhi->fib_nhc.nhc_gw_family) {
526 NL_SET_ERR_MSG(extack,
527 "Route with host scope can not have a gateway");
531 if (nhi->fib_nhc.nhc_flags & RTNH_F_ONLINK && scope >= RT_SCOPE_LINK) {
532 NL_SET_ERR_MSG(extack, "Scope mismatch with nexthop");
539 /* Invoked by fib add code to verify nexthop by id is ok with
540 * config for prefix; parts of fib_check_nh not done when nexthop
543 int fib_check_nexthop(struct nexthop *nh, u8 scope,
544 struct netlink_ext_ack *extack)
549 struct nh_group *nhg;
551 if (scope == RT_SCOPE_HOST) {
552 NL_SET_ERR_MSG(extack, "Route with host scope can not have multiple nexthops");
557 nhg = rtnl_dereference(nh->nh_grp);
558 /* all nexthops in a group have the same scope */
559 err = nexthop_check_scope(nhg->nh_entries[0].nh, scope, extack);
561 err = nexthop_check_scope(nh, scope, extack);
567 static void nh_group_rebalance(struct nh_group *nhg)
573 for (i = 0; i < nhg->num_nh; ++i)
574 total += nhg->nh_entries[i].weight;
576 for (i = 0; i < nhg->num_nh; ++i) {
577 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
581 upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31, total) - 1;
582 atomic_set(&nhge->upper_bound, upper_bound);
586 static void remove_nh_grp_entry(struct nh_grp_entry *nhge,
587 struct nh_group *nhg,
588 struct nl_info *nlinfo)
590 struct nexthop *nh = nhge->nh;
591 struct nh_grp_entry *nhges;
597 nhges = nhg->nh_entries;
598 for (i = 0; i < nhg->num_nh; ++i) {
600 nhges[i-1].nh = nhges[i].nh;
601 nhges[i-1].weight = nhges[i].weight;
602 list_del(&nhges[i].nh_list);
603 list_add(&nhges[i-1].nh_list, &nhges[i-1].nh->grp_list);
604 } else if (nhg->nh_entries[i].nh == nh) {
613 nhg->nh_entries[nhg->num_nh].nh = NULL;
615 nh_group_rebalance(nhg);
620 nexthop_notify(RTM_NEWNEXTHOP, nhge->nh_parent, nlinfo);
623 static void remove_nexthop_from_groups(struct net *net, struct nexthop *nh,
624 struct nl_info *nlinfo)
626 struct nh_grp_entry *nhge, *tmp;
628 list_for_each_entry_safe(nhge, tmp, &nh->grp_list, nh_list) {
629 struct nh_group *nhg;
631 list_del(&nhge->nh_list);
632 nhg = rtnl_dereference(nhge->nh_parent->nh_grp);
633 remove_nh_grp_entry(nhge, nhg, nlinfo);
635 /* if this group has no more entries then remove it */
637 remove_nexthop(net, nhge->nh_parent, nlinfo);
641 static void remove_nexthop_group(struct nexthop *nh, struct nl_info *nlinfo)
643 struct nh_group *nhg = rcu_dereference_rtnl(nh->nh_grp);
644 int i, num_nh = nhg->num_nh;
646 for (i = 0; i < num_nh; ++i) {
647 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
649 if (WARN_ON(!nhge->nh))
652 list_del(&nhge->nh_list);
653 nexthop_put(nhge->nh);
659 static void __remove_nexthop_fib(struct net *net, struct nexthop *nh)
661 bool do_flush = false;
664 list_for_each_entry(fi, &nh->fi_list, nh_list) {
665 fi->fib_flags |= RTNH_F_DEAD;
672 static void __remove_nexthop(struct net *net, struct nexthop *nh,
673 struct nl_info *nlinfo)
675 __remove_nexthop_fib(net, nh);
678 remove_nexthop_group(nh, nlinfo);
682 nhi = rtnl_dereference(nh->nh_info);
683 if (nhi->fib_nhc.nhc_dev)
684 hlist_del(&nhi->dev_hash);
686 remove_nexthop_from_groups(net, nh, nlinfo);
690 static void remove_nexthop(struct net *net, struct nexthop *nh,
691 struct nl_info *nlinfo)
693 /* remove from the tree */
694 rb_erase(&nh->rb_node, &net->nexthop.rb_root);
697 nexthop_notify(RTM_DELNEXTHOP, nh, nlinfo);
699 __remove_nexthop(net, nh, nlinfo);
700 nh_base_seq_inc(net);
705 static int replace_nexthop(struct net *net, struct nexthop *old,
706 struct nexthop *new, struct netlink_ext_ack *extack)
711 /* called with rtnl_lock held */
712 static int insert_nexthop(struct net *net, struct nexthop *new_nh,
713 struct nh_config *cfg, struct netlink_ext_ack *extack)
715 struct rb_node **pp, *parent = NULL, *next;
716 struct rb_root *root = &net->nexthop.rb_root;
717 bool replace = !!(cfg->nlflags & NLM_F_REPLACE);
718 bool create = !!(cfg->nlflags & NLM_F_CREATE);
719 u32 new_id = new_nh->id;
726 next = rtnl_dereference(*pp);
732 nh = rb_entry(parent, struct nexthop, rb_node);
733 if (new_id < nh->id) {
735 } else if (new_id > nh->id) {
736 pp = &next->rb_right;
737 } else if (replace) {
738 rc = replace_nexthop(net, nh, new_nh, extack);
740 new_nh = nh; /* send notification with old nh */
743 /* id already exists and not a replace */
748 if (replace && !create) {
749 NL_SET_ERR_MSG(extack, "Replace specified without create and no entry exists");
754 rb_link_node_rcu(&new_nh->rb_node, parent, pp);
755 rb_insert_color(&new_nh->rb_node, root);
759 nh_base_seq_inc(net);
760 nexthop_notify(RTM_NEWNEXTHOP, new_nh, &cfg->nlinfo);
767 /* remove all nexthops tied to a device being deleted */
768 static void nexthop_flush_dev(struct net_device *dev)
770 unsigned int hash = nh_dev_hashfn(dev->ifindex);
771 struct net *net = dev_net(dev);
772 struct hlist_head *head = &net->nexthop.devhash[hash];
773 struct hlist_node *n;
776 hlist_for_each_entry_safe(nhi, n, head, dev_hash) {
777 if (nhi->fib_nhc.nhc_dev != dev)
780 remove_nexthop(net, nhi->nh_parent, NULL);
784 /* rtnl; called when net namespace is deleted */
785 static void flush_all_nexthops(struct net *net)
787 struct rb_root *root = &net->nexthop.rb_root;
788 struct rb_node *node;
791 while ((node = rb_first(root))) {
792 nh = rb_entry(node, struct nexthop, rb_node);
793 remove_nexthop(net, nh, NULL);
798 static struct nexthop *nexthop_create_group(struct net *net,
799 struct nh_config *cfg)
801 struct nlattr *grps_attr = cfg->nh_grp;
802 struct nexthop_grp *entry = nla_data(grps_attr);
803 struct nh_group *nhg;
807 nh = nexthop_alloc();
809 return ERR_PTR(-ENOMEM);
813 nhg = nexthop_grp_alloc(nla_len(grps_attr) / sizeof(*entry));
816 return ERR_PTR(-ENOMEM);
819 for (i = 0; i < nhg->num_nh; ++i) {
823 nhe = nexthop_find_by_id(net, entry[i].id);
824 if (!nexthop_get(nhe))
827 nhi = rtnl_dereference(nhe->nh_info);
828 if (nhi->family == AF_INET)
831 nhg->nh_entries[i].nh = nhe;
832 nhg->nh_entries[i].weight = entry[i].weight + 1;
833 list_add(&nhg->nh_entries[i].nh_list, &nhe->grp_list);
834 nhg->nh_entries[i].nh_parent = nh;
837 if (cfg->nh_grp_type == NEXTHOP_GRP_TYPE_MPATH) {
839 nh_group_rebalance(nhg);
842 rcu_assign_pointer(nh->nh_grp, nhg);
848 nexthop_put(nhg->nh_entries[i].nh);
853 return ERR_PTR(-ENOENT);
856 static int nh_create_ipv4(struct net *net, struct nexthop *nh,
857 struct nh_info *nhi, struct nh_config *cfg,
858 struct netlink_ext_ack *extack)
860 struct fib_nh *fib_nh = &nhi->fib_nh;
861 struct fib_config fib_cfg = {
862 .fc_oif = cfg->nh_ifindex,
863 .fc_gw4 = cfg->gw.ipv4,
864 .fc_gw_family = cfg->gw.ipv4 ? AF_INET : 0,
865 .fc_flags = cfg->nh_flags,
866 .fc_encap = cfg->nh_encap,
867 .fc_encap_type = cfg->nh_encap_type,
869 u32 tb_id = l3mdev_fib_table(cfg->dev);
872 err = fib_nh_init(net, fib_nh, &fib_cfg, 1, extack);
874 fib_nh_release(net, fib_nh);
878 /* sets nh_dev if successful */
879 err = fib_check_nh(net, fib_nh, tb_id, 0, extack);
881 nh->nh_flags = fib_nh->fib_nh_flags;
882 fib_info_update_nhc_saddr(net, &fib_nh->nh_common,
883 fib_nh->fib_nh_scope);
885 fib_nh_release(net, fib_nh);
891 static int nh_create_ipv6(struct net *net, struct nexthop *nh,
892 struct nh_info *nhi, struct nh_config *cfg,
893 struct netlink_ext_ack *extack)
895 struct fib6_nh *fib6_nh = &nhi->fib6_nh;
896 struct fib6_config fib6_cfg = {
897 .fc_table = l3mdev_fib_table(cfg->dev),
898 .fc_ifindex = cfg->nh_ifindex,
899 .fc_gateway = cfg->gw.ipv6,
900 .fc_flags = cfg->nh_flags,
901 .fc_encap = cfg->nh_encap,
902 .fc_encap_type = cfg->nh_encap_type,
906 if (!ipv6_addr_any(&cfg->gw.ipv6))
907 fib6_cfg.fc_flags |= RTF_GATEWAY;
909 /* sets nh_dev if successful */
910 err = ipv6_stub->fib6_nh_init(net, fib6_nh, &fib6_cfg, GFP_KERNEL,
913 ipv6_stub->fib6_nh_release(fib6_nh);
915 nh->nh_flags = fib6_nh->fib_nh_flags;
920 static struct nexthop *nexthop_create(struct net *net, struct nh_config *cfg,
921 struct netlink_ext_ack *extack)
927 nh = nexthop_alloc();
929 return ERR_PTR(-ENOMEM);
931 nhi = kzalloc(sizeof(*nhi), GFP_KERNEL);
934 return ERR_PTR(-ENOMEM);
937 nh->nh_flags = cfg->nh_flags;
941 nhi->family = cfg->nh_family;
942 nhi->fib_nhc.nhc_scope = RT_SCOPE_LINK;
944 if (cfg->nh_blackhole) {
946 cfg->nh_ifindex = net->loopback_dev->ifindex;
949 switch (cfg->nh_family) {
951 err = nh_create_ipv4(net, nh, nhi, cfg, extack);
954 err = nh_create_ipv6(net, nh, nhi, cfg, extack);
964 /* add the entry to the device based hash */
965 nexthop_devhash_add(net, nhi);
967 rcu_assign_pointer(nh->nh_info, nhi);
972 /* called with rtnl lock held */
973 static struct nexthop *nexthop_add(struct net *net, struct nh_config *cfg,
974 struct netlink_ext_ack *extack)
979 if (cfg->nlflags & NLM_F_REPLACE && !cfg->nh_id) {
980 NL_SET_ERR_MSG(extack, "Replace requires nexthop id");
981 return ERR_PTR(-EINVAL);
985 cfg->nh_id = nh_find_unused_id(net);
987 NL_SET_ERR_MSG(extack, "No unused id");
988 return ERR_PTR(-EINVAL);
993 nh = nexthop_create_group(net, cfg);
995 nh = nexthop_create(net, cfg, extack);
1000 refcount_set(&nh->refcnt, 1);
1001 nh->id = cfg->nh_id;
1002 nh->protocol = cfg->nh_protocol;
1005 err = insert_nexthop(net, nh, cfg, extack);
1007 __remove_nexthop(net, nh, NULL);
1015 static int rtm_to_nh_config(struct net *net, struct sk_buff *skb,
1016 struct nlmsghdr *nlh, struct nh_config *cfg,
1017 struct netlink_ext_ack *extack)
1019 struct nhmsg *nhm = nlmsg_data(nlh);
1020 struct nlattr *tb[NHA_MAX + 1];
1023 err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1029 if (nhm->resvd || nhm->nh_scope) {
1030 NL_SET_ERR_MSG(extack, "Invalid values in ancillary header");
1033 if (nhm->nh_flags & ~NEXTHOP_VALID_USER_FLAGS) {
1034 NL_SET_ERR_MSG(extack, "Invalid nexthop flags in ancillary header");
1038 switch (nhm->nh_family) {
1047 NL_SET_ERR_MSG(extack, "Invalid address family");
1051 if (tb[NHA_GROUPS] || tb[NHA_MASTER]) {
1052 NL_SET_ERR_MSG(extack, "Invalid attributes in request");
1056 memset(cfg, 0, sizeof(*cfg));
1057 cfg->nlflags = nlh->nlmsg_flags;
1058 cfg->nlinfo.portid = NETLINK_CB(skb).portid;
1059 cfg->nlinfo.nlh = nlh;
1060 cfg->nlinfo.nl_net = net;
1062 cfg->nh_family = nhm->nh_family;
1063 cfg->nh_protocol = nhm->nh_protocol;
1064 cfg->nh_flags = nhm->nh_flags;
1067 cfg->nh_id = nla_get_u32(tb[NHA_ID]);
1069 if (tb[NHA_GROUP]) {
1070 if (nhm->nh_family != AF_UNSPEC) {
1071 NL_SET_ERR_MSG(extack, "Invalid family for group");
1074 cfg->nh_grp = tb[NHA_GROUP];
1076 cfg->nh_grp_type = NEXTHOP_GRP_TYPE_MPATH;
1077 if (tb[NHA_GROUP_TYPE])
1078 cfg->nh_grp_type = nla_get_u16(tb[NHA_GROUP_TYPE]);
1080 if (cfg->nh_grp_type > NEXTHOP_GRP_TYPE_MAX) {
1081 NL_SET_ERR_MSG(extack, "Invalid group type");
1084 err = nh_check_attr_group(net, tb, extack);
1086 /* no other attributes should be set */
1090 if (tb[NHA_BLACKHOLE]) {
1091 if (tb[NHA_GATEWAY] || tb[NHA_OIF] ||
1092 tb[NHA_ENCAP] || tb[NHA_ENCAP_TYPE]) {
1093 NL_SET_ERR_MSG(extack, "Blackhole attribute can not be used with gateway or oif");
1097 cfg->nh_blackhole = 1;
1103 NL_SET_ERR_MSG(extack, "Device attribute required for non-blackhole nexthops");
1107 cfg->nh_ifindex = nla_get_u32(tb[NHA_OIF]);
1108 if (cfg->nh_ifindex)
1109 cfg->dev = __dev_get_by_index(net, cfg->nh_ifindex);
1112 NL_SET_ERR_MSG(extack, "Invalid device index");
1114 } else if (!(cfg->dev->flags & IFF_UP)) {
1115 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1118 } else if (!netif_carrier_ok(cfg->dev)) {
1119 NL_SET_ERR_MSG(extack, "Carrier for nexthop device is down");
1125 if (tb[NHA_GATEWAY]) {
1126 struct nlattr *gwa = tb[NHA_GATEWAY];
1128 switch (cfg->nh_family) {
1130 if (nla_len(gwa) != sizeof(u32)) {
1131 NL_SET_ERR_MSG(extack, "Invalid gateway");
1134 cfg->gw.ipv4 = nla_get_be32(gwa);
1137 if (nla_len(gwa) != sizeof(struct in6_addr)) {
1138 NL_SET_ERR_MSG(extack, "Invalid gateway");
1141 cfg->gw.ipv6 = nla_get_in6_addr(gwa);
1144 NL_SET_ERR_MSG(extack,
1145 "Unknown address family for gateway");
1149 /* device only nexthop (no gateway) */
1150 if (cfg->nh_flags & RTNH_F_ONLINK) {
1151 NL_SET_ERR_MSG(extack,
1152 "ONLINK flag can not be set for nexthop without a gateway");
1157 if (tb[NHA_ENCAP]) {
1158 cfg->nh_encap = tb[NHA_ENCAP];
1160 if (!tb[NHA_ENCAP_TYPE]) {
1161 NL_SET_ERR_MSG(extack, "LWT encapsulation type is missing");
1165 cfg->nh_encap_type = nla_get_u16(tb[NHA_ENCAP_TYPE]);
1166 err = lwtunnel_valid_encap_type(cfg->nh_encap_type, extack);
1170 } else if (tb[NHA_ENCAP_TYPE]) {
1171 NL_SET_ERR_MSG(extack, "LWT encapsulation attribute is missing");
1182 static int rtm_new_nexthop(struct sk_buff *skb, struct nlmsghdr *nlh,
1183 struct netlink_ext_ack *extack)
1185 struct net *net = sock_net(skb->sk);
1186 struct nh_config cfg;
1190 err = rtm_to_nh_config(net, skb, nlh, &cfg, extack);
1192 nh = nexthop_add(net, &cfg, extack);
1200 static int nh_valid_get_del_req(struct nlmsghdr *nlh, u32 *id,
1201 struct netlink_ext_ack *extack)
1203 struct nhmsg *nhm = nlmsg_data(nlh);
1204 struct nlattr *tb[NHA_MAX + 1];
1207 err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1213 for (i = 0; i < __NHA_MAX; ++i) {
1221 NL_SET_ERR_MSG_ATTR(extack, tb[i],
1222 "Unexpected attribute in request");
1226 if (nhm->nh_protocol || nhm->resvd || nhm->nh_scope || nhm->nh_flags) {
1227 NL_SET_ERR_MSG(extack, "Invalid values in header");
1232 NL_SET_ERR_MSG(extack, "Nexthop id is missing");
1236 *id = nla_get_u32(tb[NHA_ID]);
1238 NL_SET_ERR_MSG(extack, "Invalid nexthop id");
1246 static int rtm_del_nexthop(struct sk_buff *skb, struct nlmsghdr *nlh,
1247 struct netlink_ext_ack *extack)
1249 struct net *net = sock_net(skb->sk);
1250 struct nl_info nlinfo = {
1253 .portid = NETLINK_CB(skb).portid,
1259 err = nh_valid_get_del_req(nlh, &id, extack);
1263 nh = nexthop_find_by_id(net, id);
1267 remove_nexthop(net, nh, &nlinfo);
1273 static int rtm_get_nexthop(struct sk_buff *in_skb, struct nlmsghdr *nlh,
1274 struct netlink_ext_ack *extack)
1276 struct net *net = sock_net(in_skb->sk);
1277 struct sk_buff *skb = NULL;
1282 err = nh_valid_get_del_req(nlh, &id, extack);
1287 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1292 nh = nexthop_find_by_id(net, id);
1296 err = nh_fill_node(skb, nh, RTM_NEWNEXTHOP, NETLINK_CB(in_skb).portid,
1299 WARN_ON(err == -EMSGSIZE);
1303 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1311 static bool nh_dump_filtered(struct nexthop *nh, int dev_idx, int master_idx,
1312 bool group_filter, u8 family)
1314 const struct net_device *dev;
1315 const struct nh_info *nhi;
1317 if (group_filter && !nh->is_group)
1320 if (!dev_idx && !master_idx && !family)
1326 nhi = rtnl_dereference(nh->nh_info);
1327 if (family && nhi->family != family)
1330 dev = nhi->fib_nhc.nhc_dev;
1331 if (dev_idx && (!dev || dev->ifindex != dev_idx))
1335 struct net_device *master;
1340 master = netdev_master_upper_dev_get((struct net_device *)dev);
1341 if (!master || master->ifindex != master_idx)
1348 static int nh_valid_dump_req(const struct nlmsghdr *nlh, int *dev_idx,
1349 int *master_idx, bool *group_filter,
1350 struct netlink_callback *cb)
1352 struct netlink_ext_ack *extack = cb->extack;
1353 struct nlattr *tb[NHA_MAX + 1];
1358 err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1363 for (i = 0; i <= NHA_MAX; ++i) {
1369 idx = nla_get_u32(tb[i]);
1370 if (idx > INT_MAX) {
1371 NL_SET_ERR_MSG(extack, "Invalid device index");
1377 idx = nla_get_u32(tb[i]);
1378 if (idx > INT_MAX) {
1379 NL_SET_ERR_MSG(extack, "Invalid master device index");
1385 *group_filter = true;
1388 NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request");
1393 nhm = nlmsg_data(nlh);
1394 if (nhm->nh_protocol || nhm->resvd || nhm->nh_scope || nhm->nh_flags) {
1395 NL_SET_ERR_MSG(extack, "Invalid values in header for nexthop dump request");
1403 static int rtm_dump_nexthop(struct sk_buff *skb, struct netlink_callback *cb)
1405 struct nhmsg *nhm = nlmsg_data(cb->nlh);
1406 int dev_filter_idx = 0, master_idx = 0;
1407 struct net *net = sock_net(skb->sk);
1408 struct rb_root *root = &net->nexthop.rb_root;
1409 bool group_filter = false;
1410 struct rb_node *node;
1414 err = nh_valid_dump_req(cb->nlh, &dev_filter_idx, &master_idx,
1419 s_idx = cb->args[0];
1420 for (node = rb_first(root); node; node = rb_next(node)) {
1426 nh = rb_entry(node, struct nexthop, rb_node);
1427 if (nh_dump_filtered(nh, dev_filter_idx, master_idx,
1428 group_filter, nhm->nh_family))
1431 err = nh_fill_node(skb, nh, RTM_NEWNEXTHOP,
1432 NETLINK_CB(cb->skb).portid,
1433 cb->nlh->nlmsg_seq, NLM_F_MULTI);
1435 if (likely(skb->len))
1448 cb->seq = net->nexthop.seq;
1449 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1454 static void nexthop_sync_mtu(struct net_device *dev, u32 orig_mtu)
1456 unsigned int hash = nh_dev_hashfn(dev->ifindex);
1457 struct net *net = dev_net(dev);
1458 struct hlist_head *head = &net->nexthop.devhash[hash];
1459 struct hlist_node *n;
1460 struct nh_info *nhi;
1462 hlist_for_each_entry_safe(nhi, n, head, dev_hash) {
1463 if (nhi->fib_nhc.nhc_dev == dev) {
1464 if (nhi->family == AF_INET)
1465 fib_nhc_update_mtu(&nhi->fib_nhc, dev->mtu,
1472 static int nh_netdev_event(struct notifier_block *this,
1473 unsigned long event, void *ptr)
1475 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1476 struct netdev_notifier_info_ext *info_ext;
1480 case NETDEV_UNREGISTER:
1481 nexthop_flush_dev(dev);
1484 if (!(dev_get_flags(dev) & (IFF_RUNNING | IFF_LOWER_UP)))
1485 nexthop_flush_dev(dev);
1487 case NETDEV_CHANGEMTU:
1489 nexthop_sync_mtu(dev, info_ext->ext.mtu);
1490 rt_cache_flush(dev_net(dev));
1496 static struct notifier_block nh_netdev_notifier = {
1497 .notifier_call = nh_netdev_event,
1500 static void __net_exit nexthop_net_exit(struct net *net)
1503 flush_all_nexthops(net);
1505 kfree(net->nexthop.devhash);
1508 static int __net_init nexthop_net_init(struct net *net)
1510 size_t sz = sizeof(struct hlist_head) * NH_DEV_HASHSIZE;
1512 net->nexthop.rb_root = RB_ROOT;
1513 net->nexthop.devhash = kzalloc(sz, GFP_KERNEL);
1514 if (!net->nexthop.devhash)
1520 static struct pernet_operations nexthop_net_ops = {
1521 .init = nexthop_net_init,
1522 .exit = nexthop_net_exit,
1525 static int __init nexthop_init(void)
1527 register_pernet_subsys(&nexthop_net_ops);
1529 register_netdevice_notifier(&nh_netdev_notifier);
1531 rtnl_register(PF_UNSPEC, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1532 rtnl_register(PF_UNSPEC, RTM_DELNEXTHOP, rtm_del_nexthop, NULL, 0);
1533 rtnl_register(PF_UNSPEC, RTM_GETNEXTHOP, rtm_get_nexthop,
1534 rtm_dump_nexthop, 0);
1536 rtnl_register(PF_INET, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1537 rtnl_register(PF_INET, RTM_GETNEXTHOP, NULL, rtm_dump_nexthop, 0);
1539 rtnl_register(PF_INET6, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1540 rtnl_register(PF_INET6, RTM_GETNEXTHOP, NULL, rtm_dump_nexthop, 0);
1544 subsys_initcall(nexthop_init);