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->f6i_list);
110 INIT_LIST_HEAD(&nh->grp_list);
115 static struct nh_group *nexthop_grp_alloc(u16 num_nh)
117 size_t sz = offsetof(struct nexthop, nh_grp)
118 + sizeof(struct nh_group)
119 + sizeof(struct nh_grp_entry) * num_nh;
120 struct nh_group *nhg;
122 nhg = kzalloc(sz, GFP_KERNEL);
124 nhg->num_nh = num_nh;
129 static void nh_base_seq_inc(struct net *net)
131 while (++net->nexthop.seq == 0)
135 /* no reference taken; rcu lock or rtnl must be held */
136 struct nexthop *nexthop_find_by_id(struct net *net, u32 id)
138 struct rb_node **pp, *parent = NULL, *next;
140 pp = &net->nexthop.rb_root.rb_node;
144 next = rcu_dereference_raw(*pp);
149 nh = rb_entry(parent, struct nexthop, rb_node);
152 else if (id > nh->id)
153 pp = &next->rb_right;
159 EXPORT_SYMBOL_GPL(nexthop_find_by_id);
161 /* used for auto id allocation; called with rtnl held */
162 static u32 nh_find_unused_id(struct net *net)
164 u32 id_start = net->nexthop.last_id_allocated;
167 net->nexthop.last_id_allocated++;
168 if (net->nexthop.last_id_allocated == id_start)
171 if (!nexthop_find_by_id(net, net->nexthop.last_id_allocated))
172 return net->nexthop.last_id_allocated;
177 static int nla_put_nh_group(struct sk_buff *skb, struct nh_group *nhg)
179 struct nexthop_grp *p;
180 size_t len = nhg->num_nh * sizeof(*p);
186 group_type = NEXTHOP_GRP_TYPE_MPATH;
188 if (nla_put_u16(skb, NHA_GROUP_TYPE, group_type))
189 goto nla_put_failure;
191 nla = nla_reserve(skb, NHA_GROUP, len);
193 goto nla_put_failure;
196 for (i = 0; i < nhg->num_nh; ++i) {
197 p->id = nhg->nh_entries[i].nh->id;
198 p->weight = nhg->nh_entries[i].weight - 1;
208 static int nh_fill_node(struct sk_buff *skb, struct nexthop *nh,
209 int event, u32 portid, u32 seq, unsigned int nlflags)
211 struct fib6_nh *fib6_nh;
212 struct fib_nh *fib_nh;
213 struct nlmsghdr *nlh;
217 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*nhm), nlflags);
221 nhm = nlmsg_data(nlh);
222 nhm->nh_family = AF_UNSPEC;
223 nhm->nh_flags = nh->nh_flags;
224 nhm->nh_protocol = nh->protocol;
228 if (nla_put_u32(skb, NHA_ID, nh->id))
229 goto nla_put_failure;
232 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
234 if (nla_put_nh_group(skb, nhg))
235 goto nla_put_failure;
239 nhi = rtnl_dereference(nh->nh_info);
240 nhm->nh_family = nhi->family;
241 if (nhi->reject_nh) {
242 if (nla_put_flag(skb, NHA_BLACKHOLE))
243 goto nla_put_failure;
246 const struct net_device *dev;
248 dev = nhi->fib_nhc.nhc_dev;
249 if (dev && nla_put_u32(skb, NHA_OIF, dev->ifindex))
250 goto nla_put_failure;
253 nhm->nh_scope = nhi->fib_nhc.nhc_scope;
254 switch (nhi->family) {
256 fib_nh = &nhi->fib_nh;
257 if (fib_nh->fib_nh_gw_family &&
258 nla_put_u32(skb, NHA_GATEWAY, fib_nh->fib_nh_gw4))
259 goto nla_put_failure;
263 fib6_nh = &nhi->fib6_nh;
264 if (fib6_nh->fib_nh_gw_family &&
265 nla_put_in6_addr(skb, NHA_GATEWAY, &fib6_nh->fib_nh_gw6))
266 goto nla_put_failure;
270 if (nhi->fib_nhc.nhc_lwtstate &&
271 lwtunnel_fill_encap(skb, nhi->fib_nhc.nhc_lwtstate,
272 NHA_ENCAP, NHA_ENCAP_TYPE) < 0)
273 goto nla_put_failure;
283 static size_t nh_nlmsg_size_grp(struct nexthop *nh)
285 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
286 size_t sz = sizeof(struct nexthop_grp) * nhg->num_nh;
288 return nla_total_size(sz) +
289 nla_total_size(2); /* NHA_GROUP_TYPE */
292 static size_t nh_nlmsg_size_single(struct nexthop *nh)
294 struct nh_info *nhi = rtnl_dereference(nh->nh_info);
297 /* covers NHA_BLACKHOLE since NHA_OIF and BLACKHOLE
298 * are mutually exclusive
300 sz = nla_total_size(4); /* NHA_OIF */
302 switch (nhi->family) {
304 if (nhi->fib_nh.fib_nh_gw_family)
305 sz += nla_total_size(4); /* NHA_GATEWAY */
310 if (nhi->fib6_nh.fib_nh_gw_family)
311 sz += nla_total_size(sizeof(const struct in6_addr));
315 if (nhi->fib_nhc.nhc_lwtstate) {
316 sz += lwtunnel_get_encap_size(nhi->fib_nhc.nhc_lwtstate);
317 sz += nla_total_size(2); /* NHA_ENCAP_TYPE */
323 static size_t nh_nlmsg_size(struct nexthop *nh)
325 size_t sz = nla_total_size(4); /* NHA_ID */
328 sz += nh_nlmsg_size_grp(nh);
330 sz += nh_nlmsg_size_single(nh);
335 static void nexthop_notify(int event, struct nexthop *nh, struct nl_info *info)
337 unsigned int nlflags = info->nlh ? info->nlh->nlmsg_flags : 0;
338 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
342 skb = nlmsg_new(nh_nlmsg_size(nh), gfp_any());
346 err = nh_fill_node(skb, nh, event, info->portid, seq, nlflags);
348 /* -EMSGSIZE implies BUG in nh_nlmsg_size() */
349 WARN_ON(err == -EMSGSIZE);
354 rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_NEXTHOP,
355 info->nlh, gfp_any());
359 rtnl_set_sk_err(info->nl_net, RTNLGRP_NEXTHOP, err);
362 static bool valid_group_nh(struct nexthop *nh, unsigned int npaths,
363 struct netlink_ext_ack *extack)
366 struct nh_group *nhg = rtnl_dereference(nh->nh_grp);
368 /* nested multipath (group within a group) is not
372 NL_SET_ERR_MSG(extack,
373 "Multipath group can not be a nexthop within a group");
377 struct nh_info *nhi = rtnl_dereference(nh->nh_info);
379 if (nhi->reject_nh && npaths > 1) {
380 NL_SET_ERR_MSG(extack,
381 "Blackhole nexthop can not be used in a group with more than 1 path");
389 static int nh_check_attr_group(struct net *net, struct nlattr *tb[],
390 struct netlink_ext_ack *extack)
392 unsigned int len = nla_len(tb[NHA_GROUP]);
393 struct nexthop_grp *nhg;
396 if (len & (sizeof(struct nexthop_grp) - 1)) {
397 NL_SET_ERR_MSG(extack,
398 "Invalid length for nexthop group attribute");
402 /* convert len to number of nexthop ids */
405 nhg = nla_data(tb[NHA_GROUP]);
406 for (i = 0; i < len; ++i) {
407 if (nhg[i].resvd1 || nhg[i].resvd2) {
408 NL_SET_ERR_MSG(extack, "Reserved fields in nexthop_grp must be 0");
411 if (nhg[i].weight > 254) {
412 NL_SET_ERR_MSG(extack, "Invalid value for weight");
415 for (j = i + 1; j < len; ++j) {
416 if (nhg[i].id == nhg[j].id) {
417 NL_SET_ERR_MSG(extack, "Nexthop id can not be used twice in a group");
423 nhg = nla_data(tb[NHA_GROUP]);
424 for (i = 0; i < len; ++i) {
427 nh = nexthop_find_by_id(net, nhg[i].id);
429 NL_SET_ERR_MSG(extack, "Invalid nexthop id");
432 if (!valid_group_nh(nh, len, extack))
435 for (i = NHA_GROUP + 1; i < __NHA_MAX; ++i) {
439 NL_SET_ERR_MSG(extack,
440 "No other attributes can be set in nexthop groups");
447 static bool ipv6_good_nh(const struct fib6_nh *nh)
449 int state = NUD_REACHABLE;
454 n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev, &nh->fib_nh_gw6);
456 state = n->nud_state;
458 rcu_read_unlock_bh();
460 return !!(state & NUD_VALID);
463 static bool ipv4_good_nh(const struct fib_nh *nh)
465 int state = NUD_REACHABLE;
470 n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
471 (__force u32)nh->fib_nh_gw4);
473 state = n->nud_state;
475 rcu_read_unlock_bh();
477 return !!(state & NUD_VALID);
480 struct nexthop *nexthop_select_path(struct nexthop *nh, int hash)
482 struct nexthop *rc = NULL;
483 struct nh_group *nhg;
489 nhg = rcu_dereference(nh->nh_grp);
490 for (i = 0; i < nhg->num_nh; ++i) {
491 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
494 if (hash > atomic_read(&nhge->upper_bound))
497 /* nexthops always check if it is good and does
498 * not rely on a sysctl for this behavior
500 nhi = rcu_dereference(nhge->nh->nh_info);
501 switch (nhi->family) {
503 if (ipv4_good_nh(&nhi->fib_nh))
507 if (ipv6_good_nh(&nhi->fib6_nh))
518 EXPORT_SYMBOL_GPL(nexthop_select_path);
520 int fib6_check_nexthop(struct nexthop *nh, struct fib6_config *cfg,
521 struct netlink_ext_ack *extack)
525 /* fib6_src is unique to a fib6_info and limits the ability to cache
526 * routes in fib6_nh within a nexthop that is potentially shared
527 * across multiple fib entries. If the config wants to use source
528 * routing it can not use nexthop objects. mlxsw also does not allow
529 * fib6_src on routes.
531 if (!ipv6_addr_any(&cfg->fc_src)) {
532 NL_SET_ERR_MSG(extack, "IPv6 routes using source address can not use nexthop objects");
537 struct nh_group *nhg;
539 nhg = rtnl_dereference(nh->nh_grp);
543 nhi = rtnl_dereference(nh->nh_info);
544 if (nhi->family == AF_INET)
550 NL_SET_ERR_MSG(extack, "IPv6 routes can not use an IPv4 nexthop");
553 EXPORT_SYMBOL_GPL(fib6_check_nexthop);
555 static int nexthop_check_scope(struct nexthop *nh, u8 scope,
556 struct netlink_ext_ack *extack)
560 nhi = rtnl_dereference(nh->nh_info);
561 if (scope == RT_SCOPE_HOST && nhi->fib_nhc.nhc_gw_family) {
562 NL_SET_ERR_MSG(extack,
563 "Route with host scope can not have a gateway");
567 if (nhi->fib_nhc.nhc_flags & RTNH_F_ONLINK && scope >= RT_SCOPE_LINK) {
568 NL_SET_ERR_MSG(extack, "Scope mismatch with nexthop");
575 /* Invoked by fib add code to verify nexthop by id is ok with
576 * config for prefix; parts of fib_check_nh not done when nexthop
579 int fib_check_nexthop(struct nexthop *nh, u8 scope,
580 struct netlink_ext_ack *extack)
585 struct nh_group *nhg;
587 if (scope == RT_SCOPE_HOST) {
588 NL_SET_ERR_MSG(extack, "Route with host scope can not have multiple nexthops");
593 nhg = rtnl_dereference(nh->nh_grp);
594 /* all nexthops in a group have the same scope */
595 err = nexthop_check_scope(nhg->nh_entries[0].nh, scope, extack);
597 err = nexthop_check_scope(nh, scope, extack);
603 static void nh_group_rebalance(struct nh_group *nhg)
609 for (i = 0; i < nhg->num_nh; ++i)
610 total += nhg->nh_entries[i].weight;
612 for (i = 0; i < nhg->num_nh; ++i) {
613 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
617 upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31, total) - 1;
618 atomic_set(&nhge->upper_bound, upper_bound);
622 static void remove_nh_grp_entry(struct nh_grp_entry *nhge,
623 struct nh_group *nhg,
624 struct nl_info *nlinfo)
626 struct nexthop *nh = nhge->nh;
627 struct nh_grp_entry *nhges;
633 nhges = nhg->nh_entries;
634 for (i = 0; i < nhg->num_nh; ++i) {
636 nhges[i-1].nh = nhges[i].nh;
637 nhges[i-1].weight = nhges[i].weight;
638 list_del(&nhges[i].nh_list);
639 list_add(&nhges[i-1].nh_list, &nhges[i-1].nh->grp_list);
640 } else if (nhg->nh_entries[i].nh == nh) {
649 nhg->nh_entries[nhg->num_nh].nh = NULL;
651 nh_group_rebalance(nhg);
656 nexthop_notify(RTM_NEWNEXTHOP, nhge->nh_parent, nlinfo);
659 static void remove_nexthop_from_groups(struct net *net, struct nexthop *nh,
660 struct nl_info *nlinfo)
662 struct nh_grp_entry *nhge, *tmp;
664 list_for_each_entry_safe(nhge, tmp, &nh->grp_list, nh_list) {
665 struct nh_group *nhg;
667 list_del(&nhge->nh_list);
668 nhg = rtnl_dereference(nhge->nh_parent->nh_grp);
669 remove_nh_grp_entry(nhge, nhg, nlinfo);
671 /* if this group has no more entries then remove it */
673 remove_nexthop(net, nhge->nh_parent, nlinfo);
677 static void remove_nexthop_group(struct nexthop *nh, struct nl_info *nlinfo)
679 struct nh_group *nhg = rcu_dereference_rtnl(nh->nh_grp);
680 int i, num_nh = nhg->num_nh;
682 for (i = 0; i < num_nh; ++i) {
683 struct nh_grp_entry *nhge = &nhg->nh_entries[i];
685 if (WARN_ON(!nhge->nh))
688 list_del(&nhge->nh_list);
689 nexthop_put(nhge->nh);
695 static void __remove_nexthop_fib(struct net *net, struct nexthop *nh)
697 struct fib6_info *f6i, *tmp;
698 bool do_flush = false;
701 list_for_each_entry(fi, &nh->fi_list, nh_list) {
702 fi->fib_flags |= RTNH_F_DEAD;
708 /* ip6_del_rt removes the entry from this list hence the _safe */
709 list_for_each_entry_safe(f6i, tmp, &nh->f6i_list, nh_list) {
710 /* __ip6_del_rt does a release, so do a hold here */
712 ipv6_stub->ip6_del_rt(net, f6i);
716 static void __remove_nexthop(struct net *net, struct nexthop *nh,
717 struct nl_info *nlinfo)
719 __remove_nexthop_fib(net, nh);
722 remove_nexthop_group(nh, nlinfo);
726 nhi = rtnl_dereference(nh->nh_info);
727 if (nhi->fib_nhc.nhc_dev)
728 hlist_del(&nhi->dev_hash);
730 remove_nexthop_from_groups(net, nh, nlinfo);
734 static void remove_nexthop(struct net *net, struct nexthop *nh,
735 struct nl_info *nlinfo)
737 /* remove from the tree */
738 rb_erase(&nh->rb_node, &net->nexthop.rb_root);
741 nexthop_notify(RTM_DELNEXTHOP, nh, nlinfo);
743 __remove_nexthop(net, nh, nlinfo);
744 nh_base_seq_inc(net);
749 static int replace_nexthop(struct net *net, struct nexthop *old,
750 struct nexthop *new, struct netlink_ext_ack *extack)
755 /* called with rtnl_lock held */
756 static int insert_nexthop(struct net *net, struct nexthop *new_nh,
757 struct nh_config *cfg, struct netlink_ext_ack *extack)
759 struct rb_node **pp, *parent = NULL, *next;
760 struct rb_root *root = &net->nexthop.rb_root;
761 bool replace = !!(cfg->nlflags & NLM_F_REPLACE);
762 bool create = !!(cfg->nlflags & NLM_F_CREATE);
763 u32 new_id = new_nh->id;
770 next = rtnl_dereference(*pp);
776 nh = rb_entry(parent, struct nexthop, rb_node);
777 if (new_id < nh->id) {
779 } else if (new_id > nh->id) {
780 pp = &next->rb_right;
781 } else if (replace) {
782 rc = replace_nexthop(net, nh, new_nh, extack);
784 new_nh = nh; /* send notification with old nh */
787 /* id already exists and not a replace */
792 if (replace && !create) {
793 NL_SET_ERR_MSG(extack, "Replace specified without create and no entry exists");
798 rb_link_node_rcu(&new_nh->rb_node, parent, pp);
799 rb_insert_color(&new_nh->rb_node, root);
803 nh_base_seq_inc(net);
804 nexthop_notify(RTM_NEWNEXTHOP, new_nh, &cfg->nlinfo);
811 /* remove all nexthops tied to a device being deleted */
812 static void nexthop_flush_dev(struct net_device *dev)
814 unsigned int hash = nh_dev_hashfn(dev->ifindex);
815 struct net *net = dev_net(dev);
816 struct hlist_head *head = &net->nexthop.devhash[hash];
817 struct hlist_node *n;
820 hlist_for_each_entry_safe(nhi, n, head, dev_hash) {
821 if (nhi->fib_nhc.nhc_dev != dev)
824 remove_nexthop(net, nhi->nh_parent, NULL);
828 /* rtnl; called when net namespace is deleted */
829 static void flush_all_nexthops(struct net *net)
831 struct rb_root *root = &net->nexthop.rb_root;
832 struct rb_node *node;
835 while ((node = rb_first(root))) {
836 nh = rb_entry(node, struct nexthop, rb_node);
837 remove_nexthop(net, nh, NULL);
842 static struct nexthop *nexthop_create_group(struct net *net,
843 struct nh_config *cfg)
845 struct nlattr *grps_attr = cfg->nh_grp;
846 struct nexthop_grp *entry = nla_data(grps_attr);
847 struct nh_group *nhg;
851 nh = nexthop_alloc();
853 return ERR_PTR(-ENOMEM);
857 nhg = nexthop_grp_alloc(nla_len(grps_attr) / sizeof(*entry));
860 return ERR_PTR(-ENOMEM);
863 for (i = 0; i < nhg->num_nh; ++i) {
867 nhe = nexthop_find_by_id(net, entry[i].id);
868 if (!nexthop_get(nhe))
871 nhi = rtnl_dereference(nhe->nh_info);
872 if (nhi->family == AF_INET)
875 nhg->nh_entries[i].nh = nhe;
876 nhg->nh_entries[i].weight = entry[i].weight + 1;
877 list_add(&nhg->nh_entries[i].nh_list, &nhe->grp_list);
878 nhg->nh_entries[i].nh_parent = nh;
881 if (cfg->nh_grp_type == NEXTHOP_GRP_TYPE_MPATH) {
883 nh_group_rebalance(nhg);
886 rcu_assign_pointer(nh->nh_grp, nhg);
892 nexthop_put(nhg->nh_entries[i].nh);
897 return ERR_PTR(-ENOENT);
900 static int nh_create_ipv4(struct net *net, struct nexthop *nh,
901 struct nh_info *nhi, struct nh_config *cfg,
902 struct netlink_ext_ack *extack)
904 struct fib_nh *fib_nh = &nhi->fib_nh;
905 struct fib_config fib_cfg = {
906 .fc_oif = cfg->nh_ifindex,
907 .fc_gw4 = cfg->gw.ipv4,
908 .fc_gw_family = cfg->gw.ipv4 ? AF_INET : 0,
909 .fc_flags = cfg->nh_flags,
910 .fc_encap = cfg->nh_encap,
911 .fc_encap_type = cfg->nh_encap_type,
913 u32 tb_id = l3mdev_fib_table(cfg->dev);
916 err = fib_nh_init(net, fib_nh, &fib_cfg, 1, extack);
918 fib_nh_release(net, fib_nh);
922 /* sets nh_dev if successful */
923 err = fib_check_nh(net, fib_nh, tb_id, 0, extack);
925 nh->nh_flags = fib_nh->fib_nh_flags;
926 fib_info_update_nhc_saddr(net, &fib_nh->nh_common,
927 fib_nh->fib_nh_scope);
929 fib_nh_release(net, fib_nh);
935 static int nh_create_ipv6(struct net *net, struct nexthop *nh,
936 struct nh_info *nhi, struct nh_config *cfg,
937 struct netlink_ext_ack *extack)
939 struct fib6_nh *fib6_nh = &nhi->fib6_nh;
940 struct fib6_config fib6_cfg = {
941 .fc_table = l3mdev_fib_table(cfg->dev),
942 .fc_ifindex = cfg->nh_ifindex,
943 .fc_gateway = cfg->gw.ipv6,
944 .fc_flags = cfg->nh_flags,
945 .fc_encap = cfg->nh_encap,
946 .fc_encap_type = cfg->nh_encap_type,
950 if (!ipv6_addr_any(&cfg->gw.ipv6))
951 fib6_cfg.fc_flags |= RTF_GATEWAY;
953 /* sets nh_dev if successful */
954 err = ipv6_stub->fib6_nh_init(net, fib6_nh, &fib6_cfg, GFP_KERNEL,
957 ipv6_stub->fib6_nh_release(fib6_nh);
959 nh->nh_flags = fib6_nh->fib_nh_flags;
964 static struct nexthop *nexthop_create(struct net *net, struct nh_config *cfg,
965 struct netlink_ext_ack *extack)
971 nh = nexthop_alloc();
973 return ERR_PTR(-ENOMEM);
975 nhi = kzalloc(sizeof(*nhi), GFP_KERNEL);
978 return ERR_PTR(-ENOMEM);
981 nh->nh_flags = cfg->nh_flags;
985 nhi->family = cfg->nh_family;
986 nhi->fib_nhc.nhc_scope = RT_SCOPE_LINK;
988 if (cfg->nh_blackhole) {
990 cfg->nh_ifindex = net->loopback_dev->ifindex;
993 switch (cfg->nh_family) {
995 err = nh_create_ipv4(net, nh, nhi, cfg, extack);
998 err = nh_create_ipv6(net, nh, nhi, cfg, extack);
1005 return ERR_PTR(err);
1008 /* add the entry to the device based hash */
1009 nexthop_devhash_add(net, nhi);
1011 rcu_assign_pointer(nh->nh_info, nhi);
1016 /* called with rtnl lock held */
1017 static struct nexthop *nexthop_add(struct net *net, struct nh_config *cfg,
1018 struct netlink_ext_ack *extack)
1023 if (cfg->nlflags & NLM_F_REPLACE && !cfg->nh_id) {
1024 NL_SET_ERR_MSG(extack, "Replace requires nexthop id");
1025 return ERR_PTR(-EINVAL);
1029 cfg->nh_id = nh_find_unused_id(net);
1031 NL_SET_ERR_MSG(extack, "No unused id");
1032 return ERR_PTR(-EINVAL);
1037 nh = nexthop_create_group(net, cfg);
1039 nh = nexthop_create(net, cfg, extack);
1044 refcount_set(&nh->refcnt, 1);
1045 nh->id = cfg->nh_id;
1046 nh->protocol = cfg->nh_protocol;
1049 err = insert_nexthop(net, nh, cfg, extack);
1051 __remove_nexthop(net, nh, NULL);
1059 static int rtm_to_nh_config(struct net *net, struct sk_buff *skb,
1060 struct nlmsghdr *nlh, struct nh_config *cfg,
1061 struct netlink_ext_ack *extack)
1063 struct nhmsg *nhm = nlmsg_data(nlh);
1064 struct nlattr *tb[NHA_MAX + 1];
1067 err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1073 if (nhm->resvd || nhm->nh_scope) {
1074 NL_SET_ERR_MSG(extack, "Invalid values in ancillary header");
1077 if (nhm->nh_flags & ~NEXTHOP_VALID_USER_FLAGS) {
1078 NL_SET_ERR_MSG(extack, "Invalid nexthop flags in ancillary header");
1082 switch (nhm->nh_family) {
1091 NL_SET_ERR_MSG(extack, "Invalid address family");
1095 if (tb[NHA_GROUPS] || tb[NHA_MASTER]) {
1096 NL_SET_ERR_MSG(extack, "Invalid attributes in request");
1100 memset(cfg, 0, sizeof(*cfg));
1101 cfg->nlflags = nlh->nlmsg_flags;
1102 cfg->nlinfo.portid = NETLINK_CB(skb).portid;
1103 cfg->nlinfo.nlh = nlh;
1104 cfg->nlinfo.nl_net = net;
1106 cfg->nh_family = nhm->nh_family;
1107 cfg->nh_protocol = nhm->nh_protocol;
1108 cfg->nh_flags = nhm->nh_flags;
1111 cfg->nh_id = nla_get_u32(tb[NHA_ID]);
1113 if (tb[NHA_GROUP]) {
1114 if (nhm->nh_family != AF_UNSPEC) {
1115 NL_SET_ERR_MSG(extack, "Invalid family for group");
1118 cfg->nh_grp = tb[NHA_GROUP];
1120 cfg->nh_grp_type = NEXTHOP_GRP_TYPE_MPATH;
1121 if (tb[NHA_GROUP_TYPE])
1122 cfg->nh_grp_type = nla_get_u16(tb[NHA_GROUP_TYPE]);
1124 if (cfg->nh_grp_type > NEXTHOP_GRP_TYPE_MAX) {
1125 NL_SET_ERR_MSG(extack, "Invalid group type");
1128 err = nh_check_attr_group(net, tb, extack);
1130 /* no other attributes should be set */
1134 if (tb[NHA_BLACKHOLE]) {
1135 if (tb[NHA_GATEWAY] || tb[NHA_OIF] ||
1136 tb[NHA_ENCAP] || tb[NHA_ENCAP_TYPE]) {
1137 NL_SET_ERR_MSG(extack, "Blackhole attribute can not be used with gateway or oif");
1141 cfg->nh_blackhole = 1;
1147 NL_SET_ERR_MSG(extack, "Device attribute required for non-blackhole nexthops");
1151 cfg->nh_ifindex = nla_get_u32(tb[NHA_OIF]);
1152 if (cfg->nh_ifindex)
1153 cfg->dev = __dev_get_by_index(net, cfg->nh_ifindex);
1156 NL_SET_ERR_MSG(extack, "Invalid device index");
1158 } else if (!(cfg->dev->flags & IFF_UP)) {
1159 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1162 } else if (!netif_carrier_ok(cfg->dev)) {
1163 NL_SET_ERR_MSG(extack, "Carrier for nexthop device is down");
1169 if (tb[NHA_GATEWAY]) {
1170 struct nlattr *gwa = tb[NHA_GATEWAY];
1172 switch (cfg->nh_family) {
1174 if (nla_len(gwa) != sizeof(u32)) {
1175 NL_SET_ERR_MSG(extack, "Invalid gateway");
1178 cfg->gw.ipv4 = nla_get_be32(gwa);
1181 if (nla_len(gwa) != sizeof(struct in6_addr)) {
1182 NL_SET_ERR_MSG(extack, "Invalid gateway");
1185 cfg->gw.ipv6 = nla_get_in6_addr(gwa);
1188 NL_SET_ERR_MSG(extack,
1189 "Unknown address family for gateway");
1193 /* device only nexthop (no gateway) */
1194 if (cfg->nh_flags & RTNH_F_ONLINK) {
1195 NL_SET_ERR_MSG(extack,
1196 "ONLINK flag can not be set for nexthop without a gateway");
1201 if (tb[NHA_ENCAP]) {
1202 cfg->nh_encap = tb[NHA_ENCAP];
1204 if (!tb[NHA_ENCAP_TYPE]) {
1205 NL_SET_ERR_MSG(extack, "LWT encapsulation type is missing");
1209 cfg->nh_encap_type = nla_get_u16(tb[NHA_ENCAP_TYPE]);
1210 err = lwtunnel_valid_encap_type(cfg->nh_encap_type, extack);
1214 } else if (tb[NHA_ENCAP_TYPE]) {
1215 NL_SET_ERR_MSG(extack, "LWT encapsulation attribute is missing");
1226 static int rtm_new_nexthop(struct sk_buff *skb, struct nlmsghdr *nlh,
1227 struct netlink_ext_ack *extack)
1229 struct net *net = sock_net(skb->sk);
1230 struct nh_config cfg;
1234 err = rtm_to_nh_config(net, skb, nlh, &cfg, extack);
1236 nh = nexthop_add(net, &cfg, extack);
1244 static int nh_valid_get_del_req(struct nlmsghdr *nlh, u32 *id,
1245 struct netlink_ext_ack *extack)
1247 struct nhmsg *nhm = nlmsg_data(nlh);
1248 struct nlattr *tb[NHA_MAX + 1];
1251 err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1257 for (i = 0; i < __NHA_MAX; ++i) {
1265 NL_SET_ERR_MSG_ATTR(extack, tb[i],
1266 "Unexpected attribute in request");
1270 if (nhm->nh_protocol || nhm->resvd || nhm->nh_scope || nhm->nh_flags) {
1271 NL_SET_ERR_MSG(extack, "Invalid values in header");
1276 NL_SET_ERR_MSG(extack, "Nexthop id is missing");
1280 *id = nla_get_u32(tb[NHA_ID]);
1282 NL_SET_ERR_MSG(extack, "Invalid nexthop id");
1290 static int rtm_del_nexthop(struct sk_buff *skb, struct nlmsghdr *nlh,
1291 struct netlink_ext_ack *extack)
1293 struct net *net = sock_net(skb->sk);
1294 struct nl_info nlinfo = {
1297 .portid = NETLINK_CB(skb).portid,
1303 err = nh_valid_get_del_req(nlh, &id, extack);
1307 nh = nexthop_find_by_id(net, id);
1311 remove_nexthop(net, nh, &nlinfo);
1317 static int rtm_get_nexthop(struct sk_buff *in_skb, struct nlmsghdr *nlh,
1318 struct netlink_ext_ack *extack)
1320 struct net *net = sock_net(in_skb->sk);
1321 struct sk_buff *skb = NULL;
1326 err = nh_valid_get_del_req(nlh, &id, extack);
1331 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1336 nh = nexthop_find_by_id(net, id);
1340 err = nh_fill_node(skb, nh, RTM_NEWNEXTHOP, NETLINK_CB(in_skb).portid,
1343 WARN_ON(err == -EMSGSIZE);
1347 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1355 static bool nh_dump_filtered(struct nexthop *nh, int dev_idx, int master_idx,
1356 bool group_filter, u8 family)
1358 const struct net_device *dev;
1359 const struct nh_info *nhi;
1361 if (group_filter && !nh->is_group)
1364 if (!dev_idx && !master_idx && !family)
1370 nhi = rtnl_dereference(nh->nh_info);
1371 if (family && nhi->family != family)
1374 dev = nhi->fib_nhc.nhc_dev;
1375 if (dev_idx && (!dev || dev->ifindex != dev_idx))
1379 struct net_device *master;
1384 master = netdev_master_upper_dev_get((struct net_device *)dev);
1385 if (!master || master->ifindex != master_idx)
1392 static int nh_valid_dump_req(const struct nlmsghdr *nlh, int *dev_idx,
1393 int *master_idx, bool *group_filter,
1394 struct netlink_callback *cb)
1396 struct netlink_ext_ack *extack = cb->extack;
1397 struct nlattr *tb[NHA_MAX + 1];
1402 err = nlmsg_parse(nlh, sizeof(*nhm), tb, NHA_MAX, rtm_nh_policy,
1407 for (i = 0; i <= NHA_MAX; ++i) {
1413 idx = nla_get_u32(tb[i]);
1414 if (idx > INT_MAX) {
1415 NL_SET_ERR_MSG(extack, "Invalid device index");
1421 idx = nla_get_u32(tb[i]);
1422 if (idx > INT_MAX) {
1423 NL_SET_ERR_MSG(extack, "Invalid master device index");
1429 *group_filter = true;
1432 NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request");
1437 nhm = nlmsg_data(nlh);
1438 if (nhm->nh_protocol || nhm->resvd || nhm->nh_scope || nhm->nh_flags) {
1439 NL_SET_ERR_MSG(extack, "Invalid values in header for nexthop dump request");
1447 static int rtm_dump_nexthop(struct sk_buff *skb, struct netlink_callback *cb)
1449 struct nhmsg *nhm = nlmsg_data(cb->nlh);
1450 int dev_filter_idx = 0, master_idx = 0;
1451 struct net *net = sock_net(skb->sk);
1452 struct rb_root *root = &net->nexthop.rb_root;
1453 bool group_filter = false;
1454 struct rb_node *node;
1458 err = nh_valid_dump_req(cb->nlh, &dev_filter_idx, &master_idx,
1463 s_idx = cb->args[0];
1464 for (node = rb_first(root); node; node = rb_next(node)) {
1470 nh = rb_entry(node, struct nexthop, rb_node);
1471 if (nh_dump_filtered(nh, dev_filter_idx, master_idx,
1472 group_filter, nhm->nh_family))
1475 err = nh_fill_node(skb, nh, RTM_NEWNEXTHOP,
1476 NETLINK_CB(cb->skb).portid,
1477 cb->nlh->nlmsg_seq, NLM_F_MULTI);
1479 if (likely(skb->len))
1492 cb->seq = net->nexthop.seq;
1493 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1498 static void nexthop_sync_mtu(struct net_device *dev, u32 orig_mtu)
1500 unsigned int hash = nh_dev_hashfn(dev->ifindex);
1501 struct net *net = dev_net(dev);
1502 struct hlist_head *head = &net->nexthop.devhash[hash];
1503 struct hlist_node *n;
1504 struct nh_info *nhi;
1506 hlist_for_each_entry_safe(nhi, n, head, dev_hash) {
1507 if (nhi->fib_nhc.nhc_dev == dev) {
1508 if (nhi->family == AF_INET)
1509 fib_nhc_update_mtu(&nhi->fib_nhc, dev->mtu,
1516 static int nh_netdev_event(struct notifier_block *this,
1517 unsigned long event, void *ptr)
1519 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1520 struct netdev_notifier_info_ext *info_ext;
1524 case NETDEV_UNREGISTER:
1525 nexthop_flush_dev(dev);
1528 if (!(dev_get_flags(dev) & (IFF_RUNNING | IFF_LOWER_UP)))
1529 nexthop_flush_dev(dev);
1531 case NETDEV_CHANGEMTU:
1533 nexthop_sync_mtu(dev, info_ext->ext.mtu);
1534 rt_cache_flush(dev_net(dev));
1540 static struct notifier_block nh_netdev_notifier = {
1541 .notifier_call = nh_netdev_event,
1544 static void __net_exit nexthop_net_exit(struct net *net)
1547 flush_all_nexthops(net);
1549 kfree(net->nexthop.devhash);
1552 static int __net_init nexthop_net_init(struct net *net)
1554 size_t sz = sizeof(struct hlist_head) * NH_DEV_HASHSIZE;
1556 net->nexthop.rb_root = RB_ROOT;
1557 net->nexthop.devhash = kzalloc(sz, GFP_KERNEL);
1558 if (!net->nexthop.devhash)
1564 static struct pernet_operations nexthop_net_ops = {
1565 .init = nexthop_net_init,
1566 .exit = nexthop_net_exit,
1569 static int __init nexthop_init(void)
1571 register_pernet_subsys(&nexthop_net_ops);
1573 register_netdevice_notifier(&nh_netdev_notifier);
1575 rtnl_register(PF_UNSPEC, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1576 rtnl_register(PF_UNSPEC, RTM_DELNEXTHOP, rtm_del_nexthop, NULL, 0);
1577 rtnl_register(PF_UNSPEC, RTM_GETNEXTHOP, rtm_get_nexthop,
1578 rtm_dump_nexthop, 0);
1580 rtnl_register(PF_INET, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1581 rtnl_register(PF_INET, RTM_GETNEXTHOP, NULL, rtm_dump_nexthop, 0);
1583 rtnl_register(PF_INET6, RTM_NEWNEXTHOP, rtm_new_nexthop, NULL, 0);
1584 rtnl_register(PF_INET6, RTM_GETNEXTHOP, NULL, rtm_dump_nexthop, 0);
1588 subsys_initcall(nexthop_init);