1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * IPv4 Forwarding Information Base: semantics.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
12 #include <linux/uaccess.h>
13 #include <linux/bitops.h>
14 #include <linux/types.h>
15 #include <linux/kernel.h>
16 #include <linux/jiffies.h>
18 #include <linux/string.h>
19 #include <linux/socket.h>
20 #include <linux/sockios.h>
21 #include <linux/errno.h>
23 #include <linux/inet.h>
24 #include <linux/inetdevice.h>
25 #include <linux/netdevice.h>
26 #include <linux/if_arp.h>
27 #include <linux/proc_fs.h>
28 #include <linux/skbuff.h>
29 #include <linux/init.h>
30 #include <linux/slab.h>
31 #include <linux/netlink.h>
35 #include <net/protocol.h>
36 #include <net/route.h>
39 #include <net/ip_fib.h>
40 #include <net/ip6_fib.h>
41 #include <net/nexthop.h>
42 #include <net/netlink.h>
44 #include <net/lwtunnel.h>
45 #include <net/fib_notifier.h>
46 #include <net/addrconf.h>
48 #include "fib_lookup.h"
50 static DEFINE_SPINLOCK(fib_info_lock);
51 static struct hlist_head *fib_info_hash;
52 static struct hlist_head *fib_info_laddrhash;
53 static unsigned int fib_info_hash_size;
54 static unsigned int fib_info_cnt;
56 #define DEVINDEX_HASHBITS 8
57 #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS)
58 static struct hlist_head fib_info_devhash[DEVINDEX_HASHSIZE];
60 /* for_nexthops and change_nexthops only used when nexthop object
61 * is not set in a fib_info. The logic within can reference fib_nh.
63 #ifdef CONFIG_IP_ROUTE_MULTIPATH
65 #define for_nexthops(fi) { \
66 int nhsel; const struct fib_nh *nh; \
67 for (nhsel = 0, nh = (fi)->fib_nh; \
68 nhsel < fib_info_num_path((fi)); \
71 #define change_nexthops(fi) { \
72 int nhsel; struct fib_nh *nexthop_nh; \
73 for (nhsel = 0, nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \
74 nhsel < fib_info_num_path((fi)); \
75 nexthop_nh++, nhsel++)
77 #else /* CONFIG_IP_ROUTE_MULTIPATH */
79 /* Hope, that gcc will optimize it to get rid of dummy loop */
81 #define for_nexthops(fi) { \
82 int nhsel; const struct fib_nh *nh = (fi)->fib_nh; \
83 for (nhsel = 0; nhsel < 1; nhsel++)
85 #define change_nexthops(fi) { \
87 struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh); \
88 for (nhsel = 0; nhsel < 1; nhsel++)
90 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
92 #define endfor_nexthops(fi) }
95 const struct fib_prop fib_props[RTN_MAX + 1] = {
98 .scope = RT_SCOPE_NOWHERE,
102 .scope = RT_SCOPE_UNIVERSE,
106 .scope = RT_SCOPE_HOST,
110 .scope = RT_SCOPE_LINK,
114 .scope = RT_SCOPE_LINK,
118 .scope = RT_SCOPE_UNIVERSE,
122 .scope = RT_SCOPE_UNIVERSE,
124 [RTN_UNREACHABLE] = {
125 .error = -EHOSTUNREACH,
126 .scope = RT_SCOPE_UNIVERSE,
130 .scope = RT_SCOPE_UNIVERSE,
134 .scope = RT_SCOPE_UNIVERSE,
138 .scope = RT_SCOPE_NOWHERE,
142 .scope = RT_SCOPE_NOWHERE,
146 static void rt_fibinfo_free(struct rtable __rcu **rtp)
148 struct rtable *rt = rcu_dereference_protected(*rtp, 1);
153 /* Not even needed : RCU_INIT_POINTER(*rtp, NULL);
154 * because we waited an RCU grace period before calling
155 * free_fib_info_rcu()
158 dst_dev_put(&rt->dst);
159 dst_release_immediate(&rt->dst);
162 static void free_nh_exceptions(struct fib_nh_common *nhc)
164 struct fnhe_hash_bucket *hash;
167 hash = rcu_dereference_protected(nhc->nhc_exceptions, 1);
170 for (i = 0; i < FNHE_HASH_SIZE; i++) {
171 struct fib_nh_exception *fnhe;
173 fnhe = rcu_dereference_protected(hash[i].chain, 1);
175 struct fib_nh_exception *next;
177 next = rcu_dereference_protected(fnhe->fnhe_next, 1);
179 rt_fibinfo_free(&fnhe->fnhe_rth_input);
180 rt_fibinfo_free(&fnhe->fnhe_rth_output);
190 static void rt_fibinfo_free_cpus(struct rtable __rcu * __percpu *rtp)
197 for_each_possible_cpu(cpu) {
200 rt = rcu_dereference_protected(*per_cpu_ptr(rtp, cpu), 1);
202 dst_dev_put(&rt->dst);
203 dst_release_immediate(&rt->dst);
209 void fib_nh_common_release(struct fib_nh_common *nhc)
212 dev_put(nhc->nhc_dev);
214 lwtstate_put(nhc->nhc_lwtstate);
215 rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
216 rt_fibinfo_free(&nhc->nhc_rth_input);
217 free_nh_exceptions(nhc);
219 EXPORT_SYMBOL_GPL(fib_nh_common_release);
221 void fib_nh_release(struct net *net, struct fib_nh *fib_nh)
223 #ifdef CONFIG_IP_ROUTE_CLASSID
224 if (fib_nh->nh_tclassid)
225 net->ipv4.fib_num_tclassid_users--;
227 fib_nh_common_release(&fib_nh->nh_common);
230 /* Release a nexthop info record */
231 static void free_fib_info_rcu(struct rcu_head *head)
233 struct fib_info *fi = container_of(head, struct fib_info, rcu);
238 change_nexthops(fi) {
239 fib_nh_release(fi->fib_net, nexthop_nh);
240 } endfor_nexthops(fi);
243 ip_fib_metrics_put(fi->fib_metrics);
248 void free_fib_info(struct fib_info *fi)
250 if (fi->fib_dead == 0) {
251 pr_warn("Freeing alive fib_info %p\n", fi);
256 call_rcu(&fi->rcu, free_fib_info_rcu);
258 EXPORT_SYMBOL_GPL(free_fib_info);
260 void fib_release_info(struct fib_info *fi)
262 spin_lock_bh(&fib_info_lock);
263 if (fi && --fi->fib_treeref == 0) {
264 hlist_del(&fi->fib_hash);
266 hlist_del(&fi->fib_lhash);
268 list_del(&fi->nh_list);
270 change_nexthops(fi) {
271 if (!nexthop_nh->fib_nh_dev)
273 hlist_del(&nexthop_nh->nh_hash);
274 } endfor_nexthops(fi)
279 spin_unlock_bh(&fib_info_lock);
282 static inline int nh_comp(struct fib_info *fi, struct fib_info *ofi)
284 const struct fib_nh *onh;
286 if (fi->nh || ofi->nh)
287 return nexthop_cmp(fi->nh, ofi->nh) ? 0 : -1;
289 if (ofi->fib_nhs == 0)
293 onh = fib_info_nh(ofi, nhsel);
295 if (nh->fib_nh_oif != onh->fib_nh_oif ||
296 nh->fib_nh_gw_family != onh->fib_nh_gw_family ||
297 nh->fib_nh_scope != onh->fib_nh_scope ||
298 #ifdef CONFIG_IP_ROUTE_MULTIPATH
299 nh->fib_nh_weight != onh->fib_nh_weight ||
301 #ifdef CONFIG_IP_ROUTE_CLASSID
302 nh->nh_tclassid != onh->nh_tclassid ||
304 lwtunnel_cmp_encap(nh->fib_nh_lws, onh->fib_nh_lws) ||
305 ((nh->fib_nh_flags ^ onh->fib_nh_flags) & ~RTNH_COMPARE_MASK))
308 if (nh->fib_nh_gw_family == AF_INET &&
309 nh->fib_nh_gw4 != onh->fib_nh_gw4)
312 if (nh->fib_nh_gw_family == AF_INET6 &&
313 ipv6_addr_cmp(&nh->fib_nh_gw6, &onh->fib_nh_gw6))
315 } endfor_nexthops(fi);
319 static inline unsigned int fib_devindex_hashfn(unsigned int val)
321 unsigned int mask = DEVINDEX_HASHSIZE - 1;
324 (val >> DEVINDEX_HASHBITS) ^
325 (val >> (DEVINDEX_HASHBITS * 2))) & mask;
328 static unsigned int fib_info_hashfn_1(int init_val, u8 protocol, u8 scope,
329 u32 prefsrc, u32 priority)
331 unsigned int val = init_val;
333 val ^= (protocol << 8) | scope;
340 static unsigned int fib_info_hashfn_result(unsigned int val)
342 unsigned int mask = (fib_info_hash_size - 1);
344 return (val ^ (val >> 7) ^ (val >> 12)) & mask;
347 static inline unsigned int fib_info_hashfn(struct fib_info *fi)
351 val = fib_info_hashfn_1(fi->fib_nhs, fi->fib_protocol,
352 fi->fib_scope, (__force u32)fi->fib_prefsrc,
356 val ^= fib_devindex_hashfn(fi->nh->id);
359 val ^= fib_devindex_hashfn(nh->fib_nh_oif);
360 } endfor_nexthops(fi)
363 return fib_info_hashfn_result(val);
366 /* no metrics, only nexthop id */
367 static struct fib_info *fib_find_info_nh(struct net *net,
368 const struct fib_config *cfg)
370 struct hlist_head *head;
374 hash = fib_info_hashfn_1(fib_devindex_hashfn(cfg->fc_nh_id),
375 cfg->fc_protocol, cfg->fc_scope,
376 (__force u32)cfg->fc_prefsrc,
378 hash = fib_info_hashfn_result(hash);
379 head = &fib_info_hash[hash];
381 hlist_for_each_entry(fi, head, fib_hash) {
382 if (!net_eq(fi->fib_net, net))
384 if (!fi->nh || fi->nh->id != cfg->fc_nh_id)
386 if (cfg->fc_protocol == fi->fib_protocol &&
387 cfg->fc_scope == fi->fib_scope &&
388 cfg->fc_prefsrc == fi->fib_prefsrc &&
389 cfg->fc_priority == fi->fib_priority &&
390 cfg->fc_type == fi->fib_type &&
391 cfg->fc_table == fi->fib_tb_id &&
392 !((cfg->fc_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK))
399 static struct fib_info *fib_find_info(struct fib_info *nfi)
401 struct hlist_head *head;
405 hash = fib_info_hashfn(nfi);
406 head = &fib_info_hash[hash];
408 hlist_for_each_entry(fi, head, fib_hash) {
409 if (!net_eq(fi->fib_net, nfi->fib_net))
411 if (fi->fib_nhs != nfi->fib_nhs)
413 if (nfi->fib_protocol == fi->fib_protocol &&
414 nfi->fib_scope == fi->fib_scope &&
415 nfi->fib_prefsrc == fi->fib_prefsrc &&
416 nfi->fib_priority == fi->fib_priority &&
417 nfi->fib_type == fi->fib_type &&
418 memcmp(nfi->fib_metrics, fi->fib_metrics,
419 sizeof(u32) * RTAX_MAX) == 0 &&
420 !((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK) &&
421 nh_comp(fi, nfi) == 0)
428 /* Check, that the gateway is already configured.
429 * Used only by redirect accept routine.
431 int ip_fib_check_default(__be32 gw, struct net_device *dev)
433 struct hlist_head *head;
437 spin_lock(&fib_info_lock);
439 hash = fib_devindex_hashfn(dev->ifindex);
440 head = &fib_info_devhash[hash];
441 hlist_for_each_entry(nh, head, nh_hash) {
442 if (nh->fib_nh_dev == dev &&
443 nh->fib_nh_gw4 == gw &&
444 !(nh->fib_nh_flags & RTNH_F_DEAD)) {
445 spin_unlock(&fib_info_lock);
450 spin_unlock(&fib_info_lock);
455 size_t fib_nlmsg_size(struct fib_info *fi)
457 size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
458 + nla_total_size(4) /* RTA_TABLE */
459 + nla_total_size(4) /* RTA_DST */
460 + nla_total_size(4) /* RTA_PRIORITY */
461 + nla_total_size(4) /* RTA_PREFSRC */
462 + nla_total_size(TCP_CA_NAME_MAX); /* RTAX_CC_ALGO */
463 unsigned int nhs = fib_info_num_path(fi);
465 /* space for nested metrics */
466 payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
469 payload += nla_total_size(4); /* RTA_NH_ID */
472 size_t nh_encapsize = 0;
473 /* Also handles the special case nhs == 1 */
475 /* each nexthop is packed in an attribute */
476 size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
479 /* may contain flow and gateway attribute */
480 nhsize += 2 * nla_total_size(4);
482 /* grab encap info */
483 for (i = 0; i < fib_info_num_path(fi); i++) {
484 struct fib_nh_common *nhc = fib_info_nhc(fi, i);
486 if (nhc->nhc_lwtstate) {
488 nh_encapsize += lwtunnel_get_encap_size(
491 nh_encapsize += nla_total_size(2);
495 /* all nexthops are packed in a nested attribute */
496 payload += nla_total_size((nhs * nhsize) + nh_encapsize);
503 void rtmsg_fib(int event, __be32 key, struct fib_alias *fa,
504 int dst_len, u32 tb_id, const struct nl_info *info,
505 unsigned int nlm_flags)
507 struct fib_rt_info fri;
509 u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
512 skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL);
516 fri.fi = fa->fa_info;
519 fri.dst_len = dst_len;
520 fri.tos = fa->fa_tos;
521 fri.type = fa->fa_type;
522 fri.offload = fa->offload;
524 fri.offload_failed = fa->offload_failed;
525 err = fib_dump_info(skb, info->portid, seq, event, &fri, nlm_flags);
527 /* -EMSGSIZE implies BUG in fib_nlmsg_size() */
528 WARN_ON(err == -EMSGSIZE);
532 rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_IPV4_ROUTE,
533 info->nlh, GFP_KERNEL);
537 rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err);
540 static int fib_detect_death(struct fib_info *fi, int order,
541 struct fib_info **last_resort, int *last_idx,
544 const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
546 int state = NUD_NONE;
548 if (likely(nhc->nhc_gw_family == AF_INET))
549 n = neigh_lookup(&arp_tbl, &nhc->nhc_gw.ipv4, nhc->nhc_dev);
550 else if (nhc->nhc_gw_family == AF_INET6)
551 n = neigh_lookup(ipv6_stub->nd_tbl, &nhc->nhc_gw.ipv6,
557 state = n->nud_state;
562 if (state == NUD_REACHABLE)
564 if ((state & NUD_VALID) && order != dflt)
566 if ((state & NUD_VALID) ||
567 (*last_idx < 0 && order > dflt && state != NUD_INCOMPLETE)) {
574 int fib_nh_common_init(struct net *net, struct fib_nh_common *nhc,
575 struct nlattr *encap, u16 encap_type,
576 void *cfg, gfp_t gfp_flags,
577 struct netlink_ext_ack *extack)
581 nhc->nhc_pcpu_rth_output = alloc_percpu_gfp(struct rtable __rcu *,
583 if (!nhc->nhc_pcpu_rth_output)
587 struct lwtunnel_state *lwtstate;
589 if (encap_type == LWTUNNEL_ENCAP_NONE) {
590 NL_SET_ERR_MSG(extack, "LWT encap type not specified");
594 err = lwtunnel_build_state(net, encap_type, encap,
595 nhc->nhc_family, cfg, &lwtstate,
600 nhc->nhc_lwtstate = lwtstate_get(lwtstate);
606 rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
607 nhc->nhc_pcpu_rth_output = NULL;
610 EXPORT_SYMBOL_GPL(fib_nh_common_init);
612 int fib_nh_init(struct net *net, struct fib_nh *nh,
613 struct fib_config *cfg, int nh_weight,
614 struct netlink_ext_ack *extack)
618 nh->fib_nh_family = AF_INET;
620 err = fib_nh_common_init(net, &nh->nh_common, cfg->fc_encap,
621 cfg->fc_encap_type, cfg, GFP_KERNEL, extack);
625 nh->fib_nh_oif = cfg->fc_oif;
626 nh->fib_nh_gw_family = cfg->fc_gw_family;
627 if (cfg->fc_gw_family == AF_INET)
628 nh->fib_nh_gw4 = cfg->fc_gw4;
629 else if (cfg->fc_gw_family == AF_INET6)
630 nh->fib_nh_gw6 = cfg->fc_gw6;
632 nh->fib_nh_flags = cfg->fc_flags;
634 #ifdef CONFIG_IP_ROUTE_CLASSID
635 nh->nh_tclassid = cfg->fc_flow;
637 net->ipv4.fib_num_tclassid_users++;
639 #ifdef CONFIG_IP_ROUTE_MULTIPATH
640 nh->fib_nh_weight = nh_weight;
645 #ifdef CONFIG_IP_ROUTE_MULTIPATH
647 static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining,
648 struct netlink_ext_ack *extack)
652 while (rtnh_ok(rtnh, remaining)) {
654 rtnh = rtnh_next(rtnh, &remaining);
657 /* leftover implies invalid nexthop configuration, discard it */
659 NL_SET_ERR_MSG(extack,
660 "Invalid nexthop configuration - extra data after nexthops");
667 /* only called when fib_nh is integrated into fib_info */
668 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
669 int remaining, struct fib_config *cfg,
670 struct netlink_ext_ack *extack)
672 struct net *net = fi->fib_net;
673 struct fib_config fib_cfg;
677 change_nexthops(fi) {
680 memset(&fib_cfg, 0, sizeof(fib_cfg));
682 if (!rtnh_ok(rtnh, remaining)) {
683 NL_SET_ERR_MSG(extack,
684 "Invalid nexthop configuration - extra data after nexthop");
688 if (rtnh->rtnh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
689 NL_SET_ERR_MSG(extack,
690 "Invalid flags for nexthop - can not contain DEAD or LINKDOWN");
694 fib_cfg.fc_flags = (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags;
695 fib_cfg.fc_oif = rtnh->rtnh_ifindex;
697 attrlen = rtnh_attrlen(rtnh);
699 struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
701 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
702 nlav = nla_find(attrs, attrlen, RTA_VIA);
704 NL_SET_ERR_MSG(extack,
705 "Nexthop configuration can not contain both GATEWAY and VIA");
709 fib_cfg.fc_gw4 = nla_get_in_addr(nla);
711 fib_cfg.fc_gw_family = AF_INET;
713 ret = fib_gw_from_via(&fib_cfg, nlav, extack);
718 nla = nla_find(attrs, attrlen, RTA_FLOW);
720 fib_cfg.fc_flow = nla_get_u32(nla);
722 fib_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
723 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
725 fib_cfg.fc_encap_type = nla_get_u16(nla);
728 ret = fib_nh_init(net, nexthop_nh, &fib_cfg,
729 rtnh->rtnh_hops + 1, extack);
733 rtnh = rtnh_next(rtnh, &remaining);
734 } endfor_nexthops(fi);
737 nh = fib_info_nh(fi, 0);
738 if (cfg->fc_oif && nh->fib_nh_oif != cfg->fc_oif) {
739 NL_SET_ERR_MSG(extack,
740 "Nexthop device index does not match RTA_OIF");
743 if (cfg->fc_gw_family) {
744 if (cfg->fc_gw_family != nh->fib_nh_gw_family ||
745 (cfg->fc_gw_family == AF_INET &&
746 nh->fib_nh_gw4 != cfg->fc_gw4) ||
747 (cfg->fc_gw_family == AF_INET6 &&
748 ipv6_addr_cmp(&nh->fib_nh_gw6, &cfg->fc_gw6))) {
749 NL_SET_ERR_MSG(extack,
750 "Nexthop gateway does not match RTA_GATEWAY or RTA_VIA");
754 #ifdef CONFIG_IP_ROUTE_CLASSID
755 if (cfg->fc_flow && nh->nh_tclassid != cfg->fc_flow) {
756 NL_SET_ERR_MSG(extack,
757 "Nexthop class id does not match RTA_FLOW");
766 /* only called when fib_nh is integrated into fib_info */
767 static void fib_rebalance(struct fib_info *fi)
772 if (fib_info_num_path(fi) < 2)
777 if (nh->fib_nh_flags & RTNH_F_DEAD)
780 if (ip_ignore_linkdown(nh->fib_nh_dev) &&
781 nh->fib_nh_flags & RTNH_F_LINKDOWN)
784 total += nh->fib_nh_weight;
785 } endfor_nexthops(fi);
788 change_nexthops(fi) {
791 if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD) {
793 } else if (ip_ignore_linkdown(nexthop_nh->fib_nh_dev) &&
794 nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN) {
797 w += nexthop_nh->fib_nh_weight;
798 upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31,
802 atomic_set(&nexthop_nh->fib_nh_upper_bound, upper_bound);
803 } endfor_nexthops(fi);
805 #else /* CONFIG_IP_ROUTE_MULTIPATH */
807 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
808 int remaining, struct fib_config *cfg,
809 struct netlink_ext_ack *extack)
811 NL_SET_ERR_MSG(extack, "Multipath support not enabled in kernel");
816 #define fib_rebalance(fi) do { } while (0)
818 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
820 static int fib_encap_match(struct net *net, u16 encap_type,
821 struct nlattr *encap,
822 const struct fib_nh *nh,
823 const struct fib_config *cfg,
824 struct netlink_ext_ack *extack)
826 struct lwtunnel_state *lwtstate;
829 if (encap_type == LWTUNNEL_ENCAP_NONE)
832 ret = lwtunnel_build_state(net, encap_type, encap, AF_INET,
833 cfg, &lwtstate, extack);
835 result = lwtunnel_cmp_encap(lwtstate, nh->fib_nh_lws);
836 lwtstate_free(lwtstate);
842 int fib_nh_match(struct net *net, struct fib_config *cfg, struct fib_info *fi,
843 struct netlink_ext_ack *extack)
845 #ifdef CONFIG_IP_ROUTE_MULTIPATH
846 struct rtnexthop *rtnh;
850 if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority)
854 if (fi->nh && cfg->fc_nh_id == fi->nh->id)
859 if (cfg->fc_oif || cfg->fc_gw_family) {
860 struct fib_nh *nh = fib_info_nh(fi, 0);
863 if (fib_encap_match(net, cfg->fc_encap_type,
864 cfg->fc_encap, nh, cfg, extack))
867 #ifdef CONFIG_IP_ROUTE_CLASSID
869 cfg->fc_flow != nh->nh_tclassid)
872 if ((cfg->fc_oif && cfg->fc_oif != nh->fib_nh_oif) ||
873 (cfg->fc_gw_family &&
874 cfg->fc_gw_family != nh->fib_nh_gw_family))
877 if (cfg->fc_gw_family == AF_INET &&
878 cfg->fc_gw4 != nh->fib_nh_gw4)
881 if (cfg->fc_gw_family == AF_INET6 &&
882 ipv6_addr_cmp(&cfg->fc_gw6, &nh->fib_nh_gw6))
888 #ifdef CONFIG_IP_ROUTE_MULTIPATH
893 remaining = cfg->fc_mp_len;
898 if (!rtnh_ok(rtnh, remaining))
901 if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->fib_nh_oif)
904 attrlen = rtnh_attrlen(rtnh);
906 struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
908 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
909 nlav = nla_find(attrs, attrlen, RTA_VIA);
911 NL_SET_ERR_MSG(extack,
912 "Nexthop configuration can not contain both GATEWAY and VIA");
917 if (nh->fib_nh_gw_family != AF_INET ||
918 nla_get_in_addr(nla) != nh->fib_nh_gw4)
921 struct fib_config cfg2;
924 err = fib_gw_from_via(&cfg2, nlav, extack);
928 switch (nh->fib_nh_gw_family) {
930 if (cfg2.fc_gw_family != AF_INET ||
931 cfg2.fc_gw4 != nh->fib_nh_gw4)
935 if (cfg2.fc_gw_family != AF_INET6 ||
936 ipv6_addr_cmp(&cfg2.fc_gw6,
943 #ifdef CONFIG_IP_ROUTE_CLASSID
944 nla = nla_find(attrs, attrlen, RTA_FLOW);
945 if (nla && nla_get_u32(nla) != nh->nh_tclassid)
950 rtnh = rtnh_next(rtnh, &remaining);
951 } endfor_nexthops(fi);
956 bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi)
964 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
965 int type = nla_type(nla);
973 if (type == RTAX_CC_ALGO) {
974 char tmp[TCP_CA_NAME_MAX];
977 nla_strscpy(tmp, nla, sizeof(tmp));
978 val = tcp_ca_get_key_by_name(fi->fib_net, tmp, &ecn_ca);
980 if (nla_len(nla) != sizeof(u32))
982 val = nla_get_u32(nla);
985 fi_val = fi->fib_metrics->metrics[type - 1];
986 if (type == RTAX_FEATURES)
987 fi_val &= ~DST_FEATURE_ECN_CA;
996 static int fib_check_nh_v6_gw(struct net *net, struct fib_nh *nh,
997 u32 table, struct netlink_ext_ack *extack)
999 struct fib6_config cfg = {
1001 .fc_flags = nh->fib_nh_flags | RTF_GATEWAY,
1002 .fc_ifindex = nh->fib_nh_oif,
1003 .fc_gateway = nh->fib_nh_gw6,
1005 struct fib6_nh fib6_nh = {};
1008 err = ipv6_stub->fib6_nh_init(net, &fib6_nh, &cfg, GFP_KERNEL, extack);
1010 nh->fib_nh_dev = fib6_nh.fib_nh_dev;
1011 dev_hold(nh->fib_nh_dev);
1012 nh->fib_nh_oif = nh->fib_nh_dev->ifindex;
1013 nh->fib_nh_scope = RT_SCOPE_LINK;
1015 ipv6_stub->fib6_nh_release(&fib6_nh);
1025 * Semantics of nexthop is very messy by historical reasons.
1026 * We have to take into account, that:
1027 * a) gateway can be actually local interface address,
1028 * so that gatewayed route is direct.
1029 * b) gateway must be on-link address, possibly
1030 * described not by an ifaddr, but also by a direct route.
1031 * c) If both gateway and interface are specified, they should not
1033 * d) If we use tunnel routes, gateway could be not on-link.
1035 * Attempt to reconcile all of these (alas, self-contradictory) conditions
1036 * results in pretty ugly and hairy code with obscure logic.
1038 * I chose to generalized it instead, so that the size
1039 * of code does not increase practically, but it becomes
1040 * much more general.
1041 * Every prefix is assigned a "scope" value: "host" is local address,
1042 * "link" is direct route,
1043 * [ ... "site" ... "interior" ... ]
1044 * and "universe" is true gateway route with global meaning.
1046 * Every prefix refers to a set of "nexthop"s (gw, oif),
1047 * where gw must have narrower scope. This recursion stops
1048 * when gw has LOCAL scope or if "nexthop" is declared ONLINK,
1049 * which means that gw is forced to be on link.
1051 * Code is still hairy, but now it is apparently logically
1052 * consistent and very flexible. F.e. as by-product it allows
1053 * to co-exists in peace independent exterior and interior
1054 * routing processes.
1056 * Normally it looks as following.
1058 * {universe prefix} -> (gw, oif) [scope link]
1060 * |-> {link prefix} -> (gw, oif) [scope local]
1062 * |-> {local prefix} (terminal node)
1064 static int fib_check_nh_v4_gw(struct net *net, struct fib_nh *nh, u32 table,
1065 u8 scope, struct netlink_ext_ack *extack)
1067 struct net_device *dev;
1068 struct fib_result res;
1071 if (nh->fib_nh_flags & RTNH_F_ONLINK) {
1072 unsigned int addr_type;
1074 if (scope >= RT_SCOPE_LINK) {
1075 NL_SET_ERR_MSG(extack, "Nexthop has invalid scope");
1078 dev = __dev_get_by_index(net, nh->fib_nh_oif);
1080 NL_SET_ERR_MSG(extack, "Nexthop device required for onlink");
1083 if (!(dev->flags & IFF_UP)) {
1084 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1087 addr_type = inet_addr_type_dev_table(net, dev, nh->fib_nh_gw4);
1088 if (addr_type != RTN_UNICAST) {
1089 NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1092 if (!netif_carrier_ok(dev))
1093 nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1094 nh->fib_nh_dev = dev;
1096 nh->fib_nh_scope = RT_SCOPE_LINK;
1101 struct fib_table *tbl = NULL;
1102 struct flowi4 fl4 = {
1103 .daddr = nh->fib_nh_gw4,
1104 .flowi4_scope = scope + 1,
1105 .flowi4_oif = nh->fib_nh_oif,
1106 .flowi4_iif = LOOPBACK_IFINDEX,
1109 /* It is not necessary, but requires a bit of thinking */
1110 if (fl4.flowi4_scope < RT_SCOPE_LINK)
1111 fl4.flowi4_scope = RT_SCOPE_LINK;
1113 if (table && table != RT_TABLE_MAIN)
1114 tbl = fib_get_table(net, table);
1117 err = fib_table_lookup(tbl, &fl4, &res,
1118 FIB_LOOKUP_IGNORE_LINKSTATE |
1121 /* on error or if no table given do full lookup. This
1122 * is needed for example when nexthops are in the local
1123 * table rather than the given table
1126 err = fib_lookup(net, &fl4, &res,
1127 FIB_LOOKUP_IGNORE_LINKSTATE);
1131 NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1137 if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) {
1138 NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1141 nh->fib_nh_scope = res.scope;
1142 nh->fib_nh_oif = FIB_RES_OIF(res);
1143 nh->fib_nh_dev = dev = FIB_RES_DEV(res);
1145 NL_SET_ERR_MSG(extack,
1146 "No egress device for nexthop gateway");
1150 if (!netif_carrier_ok(dev))
1151 nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1152 err = (dev->flags & IFF_UP) ? 0 : -ENETDOWN;
1158 static int fib_check_nh_nongw(struct net *net, struct fib_nh *nh,
1159 struct netlink_ext_ack *extack)
1161 struct in_device *in_dev;
1164 if (nh->fib_nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) {
1165 NL_SET_ERR_MSG(extack,
1166 "Invalid flags for nexthop - PERVASIVE and ONLINK can not be set");
1173 in_dev = inetdev_by_index(net, nh->fib_nh_oif);
1177 if (!(in_dev->dev->flags & IFF_UP)) {
1178 NL_SET_ERR_MSG(extack, "Device for nexthop is not up");
1182 nh->fib_nh_dev = in_dev->dev;
1183 dev_hold(nh->fib_nh_dev);
1184 nh->fib_nh_scope = RT_SCOPE_HOST;
1185 if (!netif_carrier_ok(nh->fib_nh_dev))
1186 nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1193 int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
1194 struct netlink_ext_ack *extack)
1198 if (nh->fib_nh_gw_family == AF_INET)
1199 err = fib_check_nh_v4_gw(net, nh, table, scope, extack);
1200 else if (nh->fib_nh_gw_family == AF_INET6)
1201 err = fib_check_nh_v6_gw(net, nh, table, extack);
1203 err = fib_check_nh_nongw(net, nh, extack);
1208 static inline unsigned int fib_laddr_hashfn(__be32 val)
1210 unsigned int mask = (fib_info_hash_size - 1);
1212 return ((__force u32)val ^
1213 ((__force u32)val >> 7) ^
1214 ((__force u32)val >> 14)) & mask;
1217 static struct hlist_head *fib_info_hash_alloc(int bytes)
1219 if (bytes <= PAGE_SIZE)
1220 return kzalloc(bytes, GFP_KERNEL);
1222 return (struct hlist_head *)
1223 __get_free_pages(GFP_KERNEL | __GFP_ZERO,
1227 static void fib_info_hash_free(struct hlist_head *hash, int bytes)
1232 if (bytes <= PAGE_SIZE)
1235 free_pages((unsigned long) hash, get_order(bytes));
1238 static void fib_info_hash_move(struct hlist_head *new_info_hash,
1239 struct hlist_head *new_laddrhash,
1240 unsigned int new_size)
1242 struct hlist_head *old_info_hash, *old_laddrhash;
1243 unsigned int old_size = fib_info_hash_size;
1244 unsigned int i, bytes;
1246 spin_lock_bh(&fib_info_lock);
1247 old_info_hash = fib_info_hash;
1248 old_laddrhash = fib_info_laddrhash;
1249 fib_info_hash_size = new_size;
1251 for (i = 0; i < old_size; i++) {
1252 struct hlist_head *head = &fib_info_hash[i];
1253 struct hlist_node *n;
1254 struct fib_info *fi;
1256 hlist_for_each_entry_safe(fi, n, head, fib_hash) {
1257 struct hlist_head *dest;
1258 unsigned int new_hash;
1260 new_hash = fib_info_hashfn(fi);
1261 dest = &new_info_hash[new_hash];
1262 hlist_add_head(&fi->fib_hash, dest);
1265 fib_info_hash = new_info_hash;
1267 for (i = 0; i < old_size; i++) {
1268 struct hlist_head *lhead = &fib_info_laddrhash[i];
1269 struct hlist_node *n;
1270 struct fib_info *fi;
1272 hlist_for_each_entry_safe(fi, n, lhead, fib_lhash) {
1273 struct hlist_head *ldest;
1274 unsigned int new_hash;
1276 new_hash = fib_laddr_hashfn(fi->fib_prefsrc);
1277 ldest = &new_laddrhash[new_hash];
1278 hlist_add_head(&fi->fib_lhash, ldest);
1281 fib_info_laddrhash = new_laddrhash;
1283 spin_unlock_bh(&fib_info_lock);
1285 bytes = old_size * sizeof(struct hlist_head *);
1286 fib_info_hash_free(old_info_hash, bytes);
1287 fib_info_hash_free(old_laddrhash, bytes);
1290 __be32 fib_info_update_nhc_saddr(struct net *net, struct fib_nh_common *nhc,
1291 unsigned char scope)
1295 if (nhc->nhc_family != AF_INET)
1296 return inet_select_addr(nhc->nhc_dev, 0, scope);
1298 nh = container_of(nhc, struct fib_nh, nh_common);
1299 nh->nh_saddr = inet_select_addr(nh->fib_nh_dev, nh->fib_nh_gw4, scope);
1300 nh->nh_saddr_genid = atomic_read(&net->ipv4.dev_addr_genid);
1302 return nh->nh_saddr;
1305 __be32 fib_result_prefsrc(struct net *net, struct fib_result *res)
1307 struct fib_nh_common *nhc = res->nhc;
1309 if (res->fi->fib_prefsrc)
1310 return res->fi->fib_prefsrc;
1312 if (nhc->nhc_family == AF_INET) {
1315 nh = container_of(nhc, struct fib_nh, nh_common);
1316 if (nh->nh_saddr_genid == atomic_read(&net->ipv4.dev_addr_genid))
1317 return nh->nh_saddr;
1320 return fib_info_update_nhc_saddr(net, nhc, res->fi->fib_scope);
1323 static bool fib_valid_prefsrc(struct fib_config *cfg, __be32 fib_prefsrc)
1325 if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst ||
1326 fib_prefsrc != cfg->fc_dst) {
1327 u32 tb_id = cfg->fc_table;
1330 if (tb_id == RT_TABLE_MAIN)
1331 tb_id = RT_TABLE_LOCAL;
1333 rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1334 fib_prefsrc, tb_id);
1336 if (rc != RTN_LOCAL && tb_id != RT_TABLE_LOCAL) {
1337 rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1338 fib_prefsrc, RT_TABLE_LOCAL);
1341 if (rc != RTN_LOCAL)
1347 struct fib_info *fib_create_info(struct fib_config *cfg,
1348 struct netlink_ext_ack *extack)
1351 struct fib_info *fi = NULL;
1352 struct nexthop *nh = NULL;
1353 struct fib_info *ofi;
1355 struct net *net = cfg->fc_nlinfo.nl_net;
1357 if (cfg->fc_type > RTN_MAX)
1360 /* Fast check to catch the most weird cases */
1361 if (fib_props[cfg->fc_type].scope > cfg->fc_scope) {
1362 NL_SET_ERR_MSG(extack, "Invalid scope");
1366 if (cfg->fc_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
1367 NL_SET_ERR_MSG(extack,
1368 "Invalid rtm_flags - can not contain DEAD or LINKDOWN");
1372 if (cfg->fc_nh_id) {
1374 fi = fib_find_info_nh(net, cfg);
1381 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
1383 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
1389 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1391 nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len, extack);
1398 if (fib_info_cnt >= fib_info_hash_size) {
1399 unsigned int new_size = fib_info_hash_size << 1;
1400 struct hlist_head *new_info_hash;
1401 struct hlist_head *new_laddrhash;
1406 bytes = new_size * sizeof(struct hlist_head *);
1407 new_info_hash = fib_info_hash_alloc(bytes);
1408 new_laddrhash = fib_info_hash_alloc(bytes);
1409 if (!new_info_hash || !new_laddrhash) {
1410 fib_info_hash_free(new_info_hash, bytes);
1411 fib_info_hash_free(new_laddrhash, bytes);
1413 fib_info_hash_move(new_info_hash, new_laddrhash, new_size);
1415 if (!fib_info_hash_size)
1419 fi = kzalloc(struct_size(fi, fib_nh, nhs), GFP_KERNEL);
1422 fi->fib_metrics = ip_fib_metrics_init(fi->fib_net, cfg->fc_mx,
1423 cfg->fc_mx_len, extack);
1424 if (IS_ERR(fi->fib_metrics)) {
1425 err = PTR_ERR(fi->fib_metrics);
1427 return ERR_PTR(err);
1432 fi->fib_protocol = cfg->fc_protocol;
1433 fi->fib_scope = cfg->fc_scope;
1434 fi->fib_flags = cfg->fc_flags;
1435 fi->fib_priority = cfg->fc_priority;
1436 fi->fib_prefsrc = cfg->fc_prefsrc;
1437 fi->fib_type = cfg->fc_type;
1438 fi->fib_tb_id = cfg->fc_table;
1442 if (!nexthop_get(nh)) {
1443 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1450 change_nexthops(fi) {
1451 nexthop_nh->nh_parent = fi;
1452 } endfor_nexthops(fi)
1455 err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg,
1458 err = fib_nh_init(net, fi->fib_nh, cfg, 1, extack);
1464 if (fib_props[cfg->fc_type].error) {
1465 if (cfg->fc_gw_family || cfg->fc_oif || cfg->fc_mp) {
1466 NL_SET_ERR_MSG(extack,
1467 "Gateway, device and multipath can not be specified for this route type");
1472 switch (cfg->fc_type) {
1480 NL_SET_ERR_MSG(extack, "Invalid route type");
1485 if (cfg->fc_scope > RT_SCOPE_HOST) {
1486 NL_SET_ERR_MSG(extack, "Invalid scope");
1491 err = fib_check_nexthop(fi->nh, cfg->fc_scope, extack);
1494 } else if (cfg->fc_scope == RT_SCOPE_HOST) {
1495 struct fib_nh *nh = fi->fib_nh;
1497 /* Local address is added. */
1499 NL_SET_ERR_MSG(extack,
1500 "Route with host scope can not have multiple nexthops");
1503 if (nh->fib_nh_gw_family) {
1504 NL_SET_ERR_MSG(extack,
1505 "Route with host scope can not have a gateway");
1508 nh->fib_nh_scope = RT_SCOPE_NOWHERE;
1509 nh->fib_nh_dev = dev_get_by_index(net, nh->fib_nh_oif);
1511 if (!nh->fib_nh_dev)
1516 change_nexthops(fi) {
1517 err = fib_check_nh(cfg->fc_nlinfo.nl_net, nexthop_nh,
1518 cfg->fc_table, cfg->fc_scope,
1522 if (nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN)
1524 } endfor_nexthops(fi)
1525 if (linkdown == fi->fib_nhs)
1526 fi->fib_flags |= RTNH_F_LINKDOWN;
1529 if (fi->fib_prefsrc && !fib_valid_prefsrc(cfg, fi->fib_prefsrc)) {
1530 NL_SET_ERR_MSG(extack, "Invalid prefsrc address");
1535 change_nexthops(fi) {
1536 fib_info_update_nhc_saddr(net, &nexthop_nh->nh_common,
1538 if (nexthop_nh->fib_nh_gw_family == AF_INET6)
1539 fi->fib_nh_is_v6 = true;
1540 } endfor_nexthops(fi)
1546 ofi = fib_find_info(fi);
1555 refcount_set(&fi->fib_clntref, 1);
1556 spin_lock_bh(&fib_info_lock);
1557 hlist_add_head(&fi->fib_hash,
1558 &fib_info_hash[fib_info_hashfn(fi)]);
1559 if (fi->fib_prefsrc) {
1560 struct hlist_head *head;
1562 head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)];
1563 hlist_add_head(&fi->fib_lhash, head);
1566 list_add(&fi->nh_list, &nh->fi_list);
1568 change_nexthops(fi) {
1569 struct hlist_head *head;
1572 if (!nexthop_nh->fib_nh_dev)
1574 hash = fib_devindex_hashfn(nexthop_nh->fib_nh_dev->ifindex);
1575 head = &fib_info_devhash[hash];
1576 hlist_add_head(&nexthop_nh->nh_hash, head);
1577 } endfor_nexthops(fi)
1579 spin_unlock_bh(&fib_info_lock);
1591 return ERR_PTR(err);
1594 int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nhc,
1595 u8 rt_family, unsigned char *flags, bool skip_oif)
1597 if (nhc->nhc_flags & RTNH_F_DEAD)
1598 *flags |= RTNH_F_DEAD;
1600 if (nhc->nhc_flags & RTNH_F_LINKDOWN) {
1601 *flags |= RTNH_F_LINKDOWN;
1604 switch (nhc->nhc_family) {
1606 if (ip_ignore_linkdown(nhc->nhc_dev))
1607 *flags |= RTNH_F_DEAD;
1610 if (ip6_ignore_linkdown(nhc->nhc_dev))
1611 *flags |= RTNH_F_DEAD;
1617 switch (nhc->nhc_gw_family) {
1619 if (nla_put_in_addr(skb, RTA_GATEWAY, nhc->nhc_gw.ipv4))
1620 goto nla_put_failure;
1623 /* if gateway family does not match nexthop family
1624 * gateway is encoded as RTA_VIA
1626 if (rt_family != nhc->nhc_gw_family) {
1627 int alen = sizeof(struct in6_addr);
1631 nla = nla_reserve(skb, RTA_VIA, alen + 2);
1633 goto nla_put_failure;
1635 via = nla_data(nla);
1636 via->rtvia_family = AF_INET6;
1637 memcpy(via->rtvia_addr, &nhc->nhc_gw.ipv6, alen);
1638 } else if (nla_put_in6_addr(skb, RTA_GATEWAY,
1639 &nhc->nhc_gw.ipv6) < 0) {
1640 goto nla_put_failure;
1645 *flags |= (nhc->nhc_flags &
1646 (RTNH_F_ONLINK | RTNH_F_OFFLOAD | RTNH_F_TRAP));
1648 if (!skip_oif && nhc->nhc_dev &&
1649 nla_put_u32(skb, RTA_OIF, nhc->nhc_dev->ifindex))
1650 goto nla_put_failure;
1652 if (nhc->nhc_lwtstate &&
1653 lwtunnel_fill_encap(skb, nhc->nhc_lwtstate,
1654 RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
1655 goto nla_put_failure;
1662 EXPORT_SYMBOL_GPL(fib_nexthop_info);
1664 #if IS_ENABLED(CONFIG_IP_ROUTE_MULTIPATH) || IS_ENABLED(CONFIG_IPV6)
1665 int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nhc,
1666 int nh_weight, u8 rt_family)
1668 const struct net_device *dev = nhc->nhc_dev;
1669 struct rtnexthop *rtnh;
1670 unsigned char flags = 0;
1672 rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
1674 goto nla_put_failure;
1676 rtnh->rtnh_hops = nh_weight - 1;
1677 rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
1679 if (fib_nexthop_info(skb, nhc, rt_family, &flags, true) < 0)
1680 goto nla_put_failure;
1682 rtnh->rtnh_flags = flags;
1684 /* length of rtnetlink header + attributes */
1685 rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
1692 EXPORT_SYMBOL_GPL(fib_add_nexthop);
1695 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1696 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1700 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
1702 goto nla_put_failure;
1704 if (unlikely(fi->nh)) {
1705 if (nexthop_mpath_fill_node(skb, fi->nh, AF_INET) < 0)
1706 goto nla_put_failure;
1711 if (fib_add_nexthop(skb, &nh->nh_common, nh->fib_nh_weight,
1713 goto nla_put_failure;
1714 #ifdef CONFIG_IP_ROUTE_CLASSID
1715 if (nh->nh_tclassid &&
1716 nla_put_u32(skb, RTA_FLOW, nh->nh_tclassid))
1717 goto nla_put_failure;
1719 } endfor_nexthops(fi);
1722 nla_nest_end(skb, mp);
1730 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1736 int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event,
1737 const struct fib_rt_info *fri, unsigned int flags)
1739 unsigned int nhs = fib_info_num_path(fri->fi);
1740 struct fib_info *fi = fri->fi;
1741 u32 tb_id = fri->tb_id;
1742 struct nlmsghdr *nlh;
1745 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1749 rtm = nlmsg_data(nlh);
1750 rtm->rtm_family = AF_INET;
1751 rtm->rtm_dst_len = fri->dst_len;
1752 rtm->rtm_src_len = 0;
1753 rtm->rtm_tos = fri->tos;
1755 rtm->rtm_table = tb_id;
1757 rtm->rtm_table = RT_TABLE_COMPAT;
1758 if (nla_put_u32(skb, RTA_TABLE, tb_id))
1759 goto nla_put_failure;
1760 rtm->rtm_type = fri->type;
1761 rtm->rtm_flags = fi->fib_flags;
1762 rtm->rtm_scope = fi->fib_scope;
1763 rtm->rtm_protocol = fi->fib_protocol;
1765 if (rtm->rtm_dst_len &&
1766 nla_put_in_addr(skb, RTA_DST, fri->dst))
1767 goto nla_put_failure;
1768 if (fi->fib_priority &&
1769 nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority))
1770 goto nla_put_failure;
1771 if (rtnetlink_put_metrics(skb, fi->fib_metrics->metrics) < 0)
1772 goto nla_put_failure;
1774 if (fi->fib_prefsrc &&
1775 nla_put_in_addr(skb, RTA_PREFSRC, fi->fib_prefsrc))
1776 goto nla_put_failure;
1779 if (nla_put_u32(skb, RTA_NH_ID, fi->nh->id))
1780 goto nla_put_failure;
1781 if (nexthop_is_blackhole(fi->nh))
1782 rtm->rtm_type = RTN_BLACKHOLE;
1783 if (!fi->fib_net->ipv4.sysctl_nexthop_compat_mode)
1788 const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
1789 unsigned char flags = 0;
1791 if (fib_nexthop_info(skb, nhc, AF_INET, &flags, false) < 0)
1792 goto nla_put_failure;
1794 rtm->rtm_flags = flags;
1795 #ifdef CONFIG_IP_ROUTE_CLASSID
1796 if (nhc->nhc_family == AF_INET) {
1799 nh = container_of(nhc, struct fib_nh, nh_common);
1800 if (nh->nh_tclassid &&
1801 nla_put_u32(skb, RTA_FLOW, nh->nh_tclassid))
1802 goto nla_put_failure;
1806 if (fib_add_multipath(skb, fi) < 0)
1807 goto nla_put_failure;
1812 rtm->rtm_flags |= RTM_F_OFFLOAD;
1814 rtm->rtm_flags |= RTM_F_TRAP;
1815 if (fri->offload_failed)
1816 rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
1818 nlmsg_end(skb, nlh);
1822 nlmsg_cancel(skb, nlh);
1828 * - local address disappeared -> we must delete all the entries
1830 * - device went down -> we must shutdown all nexthops going via it.
1832 int fib_sync_down_addr(struct net_device *dev, __be32 local)
1835 unsigned int hash = fib_laddr_hashfn(local);
1836 struct hlist_head *head = &fib_info_laddrhash[hash];
1837 int tb_id = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
1838 struct net *net = dev_net(dev);
1839 struct fib_info *fi;
1841 if (!fib_info_laddrhash || local == 0)
1844 hlist_for_each_entry(fi, head, fib_lhash) {
1845 if (!net_eq(fi->fib_net, net) ||
1846 fi->fib_tb_id != tb_id)
1848 if (fi->fib_prefsrc == local) {
1849 fi->fib_flags |= RTNH_F_DEAD;
1856 static int call_fib_nh_notifiers(struct fib_nh *nh,
1857 enum fib_event_type event_type)
1859 bool ignore_link_down = ip_ignore_linkdown(nh->fib_nh_dev);
1860 struct fib_nh_notifier_info info = {
1864 switch (event_type) {
1865 case FIB_EVENT_NH_ADD:
1866 if (nh->fib_nh_flags & RTNH_F_DEAD)
1868 if (ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN)
1870 return call_fib4_notifiers(dev_net(nh->fib_nh_dev), event_type,
1872 case FIB_EVENT_NH_DEL:
1873 if ((ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN) ||
1874 (nh->fib_nh_flags & RTNH_F_DEAD))
1875 return call_fib4_notifiers(dev_net(nh->fib_nh_dev),
1876 event_type, &info.info);
1884 /* Update the PMTU of exceptions when:
1885 * - the new MTU of the first hop becomes smaller than the PMTU
1886 * - the old MTU was the same as the PMTU, and it limited discovery of
1887 * larger MTUs on the path. With that limit raised, we can now
1888 * discover larger MTUs
1889 * A special case is locked exceptions, for which the PMTU is smaller
1890 * than the minimal accepted PMTU:
1891 * - if the new MTU is greater than the PMTU, don't make any change
1892 * - otherwise, unlock and set PMTU
1894 void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig)
1896 struct fnhe_hash_bucket *bucket;
1899 bucket = rcu_dereference_protected(nhc->nhc_exceptions, 1);
1903 for (i = 0; i < FNHE_HASH_SIZE; i++) {
1904 struct fib_nh_exception *fnhe;
1906 for (fnhe = rcu_dereference_protected(bucket[i].chain, 1);
1908 fnhe = rcu_dereference_protected(fnhe->fnhe_next, 1)) {
1909 if (fnhe->fnhe_mtu_locked) {
1910 if (new <= fnhe->fnhe_pmtu) {
1911 fnhe->fnhe_pmtu = new;
1912 fnhe->fnhe_mtu_locked = false;
1914 } else if (new < fnhe->fnhe_pmtu ||
1915 orig == fnhe->fnhe_pmtu) {
1916 fnhe->fnhe_pmtu = new;
1922 void fib_sync_mtu(struct net_device *dev, u32 orig_mtu)
1924 unsigned int hash = fib_devindex_hashfn(dev->ifindex);
1925 struct hlist_head *head = &fib_info_devhash[hash];
1928 hlist_for_each_entry(nh, head, nh_hash) {
1929 if (nh->fib_nh_dev == dev)
1930 fib_nhc_update_mtu(&nh->nh_common, dev->mtu, orig_mtu);
1934 /* Event force Flags Description
1935 * NETDEV_CHANGE 0 LINKDOWN Carrier OFF, not for scope host
1936 * NETDEV_DOWN 0 LINKDOWN|DEAD Link down, not for scope host
1937 * NETDEV_DOWN 1 LINKDOWN|DEAD Last address removed
1938 * NETDEV_UNREGISTER 1 LINKDOWN|DEAD Device removed
1940 * only used when fib_nh is built into fib_info
1942 int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force)
1945 int scope = RT_SCOPE_NOWHERE;
1946 struct fib_info *prev_fi = NULL;
1947 unsigned int hash = fib_devindex_hashfn(dev->ifindex);
1948 struct hlist_head *head = &fib_info_devhash[hash];
1954 hlist_for_each_entry(nh, head, nh_hash) {
1955 struct fib_info *fi = nh->nh_parent;
1958 BUG_ON(!fi->fib_nhs);
1959 if (nh->fib_nh_dev != dev || fi == prev_fi)
1963 change_nexthops(fi) {
1964 if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD)
1966 else if (nexthop_nh->fib_nh_dev == dev &&
1967 nexthop_nh->fib_nh_scope != scope) {
1970 case NETDEV_UNREGISTER:
1971 nexthop_nh->fib_nh_flags |= RTNH_F_DEAD;
1974 nexthop_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1977 call_fib_nh_notifiers(nexthop_nh,
1981 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1982 if (event == NETDEV_UNREGISTER &&
1983 nexthop_nh->fib_nh_dev == dev) {
1988 } endfor_nexthops(fi)
1989 if (dead == fi->fib_nhs) {
1992 case NETDEV_UNREGISTER:
1993 fi->fib_flags |= RTNH_F_DEAD;
1996 fi->fib_flags |= RTNH_F_LINKDOWN;
2008 /* Must be invoked inside of an RCU protected region. */
2009 static void fib_select_default(const struct flowi4 *flp, struct fib_result *res)
2011 struct fib_info *fi = NULL, *last_resort = NULL;
2012 struct hlist_head *fa_head = res->fa_head;
2013 struct fib_table *tb = res->table;
2014 u8 slen = 32 - res->prefixlen;
2015 int order = -1, last_idx = -1;
2016 struct fib_alias *fa, *fa1 = NULL;
2017 u32 last_prio = res->fi->fib_priority;
2020 hlist_for_each_entry_rcu(fa, fa_head, fa_list) {
2021 struct fib_info *next_fi = fa->fa_info;
2022 struct fib_nh_common *nhc;
2024 if (fa->fa_slen != slen)
2026 if (fa->fa_tos && fa->fa_tos != flp->flowi4_tos)
2028 if (fa->tb_id != tb->tb_id)
2030 if (next_fi->fib_priority > last_prio &&
2031 fa->fa_tos == last_tos) {
2036 if (next_fi->fib_flags & RTNH_F_DEAD)
2038 last_tos = fa->fa_tos;
2039 last_prio = next_fi->fib_priority;
2041 if (next_fi->fib_scope != res->scope ||
2042 fa->fa_type != RTN_UNICAST)
2045 nhc = fib_info_nhc(next_fi, 0);
2046 if (!nhc->nhc_gw_family || nhc->nhc_scope != RT_SCOPE_LINK)
2049 fib_alias_accessed(fa);
2052 if (next_fi != res->fi)
2055 } else if (!fib_detect_death(fi, order, &last_resort,
2056 &last_idx, fa1->fa_default)) {
2057 fib_result_assign(res, fi);
2058 fa1->fa_default = order;
2065 if (order <= 0 || !fi) {
2067 fa1->fa_default = -1;
2071 if (!fib_detect_death(fi, order, &last_resort, &last_idx,
2073 fib_result_assign(res, fi);
2074 fa1->fa_default = order;
2079 fib_result_assign(res, last_resort);
2080 fa1->fa_default = last_idx;
2086 * Dead device goes up. We wake up dead nexthops.
2087 * It takes sense only on multipath routes.
2089 * only used when fib_nh is built into fib_info
2091 int fib_sync_up(struct net_device *dev, unsigned char nh_flags)
2093 struct fib_info *prev_fi;
2095 struct hlist_head *head;
2099 if (!(dev->flags & IFF_UP))
2102 if (nh_flags & RTNH_F_DEAD) {
2103 unsigned int flags = dev_get_flags(dev);
2105 if (flags & (IFF_RUNNING | IFF_LOWER_UP))
2106 nh_flags |= RTNH_F_LINKDOWN;
2110 hash = fib_devindex_hashfn(dev->ifindex);
2111 head = &fib_info_devhash[hash];
2114 hlist_for_each_entry(nh, head, nh_hash) {
2115 struct fib_info *fi = nh->nh_parent;
2118 BUG_ON(!fi->fib_nhs);
2119 if (nh->fib_nh_dev != dev || fi == prev_fi)
2124 change_nexthops(fi) {
2125 if (!(nexthop_nh->fib_nh_flags & nh_flags)) {
2129 if (!nexthop_nh->fib_nh_dev ||
2130 !(nexthop_nh->fib_nh_dev->flags & IFF_UP))
2132 if (nexthop_nh->fib_nh_dev != dev ||
2133 !__in_dev_get_rtnl(dev))
2136 nexthop_nh->fib_nh_flags &= ~nh_flags;
2137 call_fib_nh_notifiers(nexthop_nh, FIB_EVENT_NH_ADD);
2138 } endfor_nexthops(fi)
2141 fi->fib_flags &= ~nh_flags;
2151 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2152 static bool fib_good_nh(const struct fib_nh *nh)
2154 int state = NUD_REACHABLE;
2156 if (nh->fib_nh_scope == RT_SCOPE_LINK) {
2157 struct neighbour *n;
2161 if (likely(nh->fib_nh_gw_family == AF_INET))
2162 n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
2163 (__force u32)nh->fib_nh_gw4);
2164 else if (nh->fib_nh_gw_family == AF_INET6)
2165 n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev,
2170 state = n->nud_state;
2172 rcu_read_unlock_bh();
2175 return !!(state & NUD_VALID);
2178 void fib_select_multipath(struct fib_result *res, int hash)
2180 struct fib_info *fi = res->fi;
2181 struct net *net = fi->fib_net;
2184 if (unlikely(res->fi->nh)) {
2185 nexthop_path_fib_result(res, hash);
2189 change_nexthops(fi) {
2190 if (net->ipv4.sysctl_fib_multipath_use_neigh) {
2191 if (!fib_good_nh(nexthop_nh))
2194 res->nh_sel = nhsel;
2195 res->nhc = &nexthop_nh->nh_common;
2200 if (hash > atomic_read(&nexthop_nh->fib_nh_upper_bound))
2203 res->nh_sel = nhsel;
2204 res->nhc = &nexthop_nh->nh_common;
2206 } endfor_nexthops(fi);
2210 void fib_select_path(struct net *net, struct fib_result *res,
2211 struct flowi4 *fl4, const struct sk_buff *skb)
2213 if (fl4->flowi4_oif && !(fl4->flowi4_flags & FLOWI_FLAG_SKIP_NH_OIF))
2216 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2217 if (fib_info_num_path(res->fi) > 1) {
2218 int h = fib_multipath_hash(net, fl4, skb, NULL);
2220 fib_select_multipath(res, h);
2224 if (!res->prefixlen &&
2225 res->table->tb_num_default > 1 &&
2226 res->type == RTN_UNICAST)
2227 fib_select_default(fl4, res);
2231 fl4->saddr = fib_result_prefsrc(net, res);