net/ipv6/route.c

0001 // SPDX-License-Identifier: GPL-2.0-or-later
0002 /*
0003  *  Linux INET6 implementation
0004  *  FIB front-end.
0005  *
0006  *  Authors:
0007  *  Pedro Roque     <roque@di.fc.ul.pt>
0008  */
0009
0010 /*  Changes:
0011  *
0012  *  YOSHIFUJI Hideaki @USAGI
0013  *      reworked default router selection.
0014  *      - respect outgoing interface
0015  *      - select from (probably) reachable routers (i.e.
0016  *      routers in REACHABLE, STALE, DELAY or PROBE states).
0017  *      - always select the same router if it is (probably)
0018  *      reachable.  otherwise, round-robin the list.
0019  *  Ville Nuorvala
0020  *      Fixed routing subtrees.
0021  */
0022
0023 #define pr_fmt(fmt) "IPv6: " fmt
0024
0025 #include <linux/capability.h>
0026 #include <linux/errno.h>
0027 #include <linux/export.h>
0028 #include <linux/types.h>
0029 #include <linux/times.h>
0030 #include <linux/socket.h>
0031 #include <linux/sockios.h>
0032 #include <linux/net.h>
0033 #include <linux/route.h>
0034 #include <linux/netdevice.h>
0035 #include <linux/in6.h>
0036 #include <linux/mroute6.h>
0037 #include <linux/init.h>
0038 #include <linux/if_arp.h>
0039 #include <linux/proc_fs.h>
0040 #include <linux/seq_file.h>
0041 #include <linux/nsproxy.h>
0042 #include <linux/slab.h>
0043 #include <linux/jhash.h>
0044 #include <linux/siphash.h>
0045 #include <net/net_namespace.h>
0046 #include <net/snmp.h>
0047 #include <net/ipv6.h>
0048 #include <net/ip6_fib.h>
0049 #include <net/ip6_route.h>
0050 #include <net/ndisc.h>
0051 #include <net/addrconf.h>
0052 #include <net/tcp.h>
0053 #include <linux/rtnetlink.h>
0054 #include <net/dst.h>
0055 #include <net/dst_metadata.h>
0056 #include <net/xfrm.h>
0057 #include <net/netevent.h>
0058 #include <net/netlink.h>
0059 #include <net/rtnh.h>
0060 #include <net/lwtunnel.h>
0061 #include <net/ip_tunnels.h>
0062 #include <net/l3mdev.h>
0063 #include <net/ip.h>
0064 #include <linux/uaccess.h>
0065 #include <linux/btf_ids.h>
0066
0067 #ifdef CONFIG_SYSCTL
0068 #include <linux/sysctl.h>
0069 #endif
0070
0071 static int ip6_rt_type_to_error(u8 fib6_type);
0072
0073 #define CREATE_TRACE_POINTS
0074 #include <trace/events/fib6.h>
0075 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
0076 #undef CREATE_TRACE_POINTS
0077
0078 enum rt6_nud_state {
0079     RT6_NUD_FAIL_HARD = -3,
0080     RT6_NUD_FAIL_PROBE = -2,
0081     RT6_NUD_FAIL_DO_RR = -1,
0082     RT6_NUD_SUCCEED = 1
0083 };
0084
0085 INDIRECT_CALLABLE_SCOPE
0086 struct dst_entry    *ip6_dst_check(struct dst_entry *dst, u32 cookie);
0087 static unsigned int  ip6_default_advmss(const struct dst_entry *dst);
0088 INDIRECT_CALLABLE_SCOPE
0089 unsigned int        ip6_mtu(const struct dst_entry *dst);
0090 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
0091 static void     ip6_dst_destroy(struct dst_entry *);
0092 static void     ip6_dst_ifdown(struct dst_entry *,
0093                        struct net_device *dev, int how);
0094 static int       ip6_dst_gc(struct dst_ops *ops);
0095
0096 static int      ip6_pkt_discard(struct sk_buff *skb);
0097 static int      ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
0098 static int      ip6_pkt_prohibit(struct sk_buff *skb);
0099 static int      ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
0100 static void     ip6_link_failure(struct sk_buff *skb);
0101 static void     ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
0102                        struct sk_buff *skb, u32 mtu,
0103                        bool confirm_neigh);
0104 static void     rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
0105                     struct sk_buff *skb);
0106 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
0107                int strict);
0108 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
0109 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
0110              struct fib6_info *rt, struct dst_entry *dst,
0111              struct in6_addr *dest, struct in6_addr *src,
0112              int iif, int type, u32 portid, u32 seq,
0113              unsigned int flags);
0114 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
0115                        const struct in6_addr *daddr,
0116                        const struct in6_addr *saddr);
0117
0118 #ifdef CONFIG_IPV6_ROUTE_INFO
0119 static struct fib6_info *rt6_add_route_info(struct net *net,
0120                        const struct in6_addr *prefix, int prefixlen,
0121                        const struct in6_addr *gwaddr,
0122                        struct net_device *dev,
0123                        unsigned int pref);
0124 static struct fib6_info *rt6_get_route_info(struct net *net,
0125                        const struct in6_addr *prefix, int prefixlen,
0126                        const struct in6_addr *gwaddr,
0127                        struct net_device *dev);
0128 #endif
0129
0130 struct uncached_list {
0131     spinlock_t      lock;
0132     struct list_head    head;
0133     struct list_head    quarantine;
0134 };
0135
0136 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
0137
0138 void rt6_uncached_list_add(struct rt6_info *rt)
0139 {
0140     struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
0141
0142     rt->rt6i_uncached_list = ul;
0143
0144     spin_lock_bh(&ul->lock);
0145     list_add_tail(&rt->rt6i_uncached, &ul->head);
0146     spin_unlock_bh(&ul->lock);
0147 }
0148
0149 void rt6_uncached_list_del(struct rt6_info *rt)
0150 {
0151     if (!list_empty(&rt->rt6i_uncached)) {
0152         struct uncached_list *ul = rt->rt6i_uncached_list;
0153
0154         spin_lock_bh(&ul->lock);
0155         list_del_init(&rt->rt6i_uncached);
0156         spin_unlock_bh(&ul->lock);
0157     }
0158 }
0159
0160 static void rt6_uncached_list_flush_dev(struct net_device *dev)
0161 {
0162     int cpu;
0163
0164     for_each_possible_cpu(cpu) {
0165         struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
0166         struct rt6_info *rt, *safe;
0167
0168         if (list_empty(&ul->head))
0169             continue;
0170
0171         spin_lock_bh(&ul->lock);
0172         list_for_each_entry_safe(rt, safe, &ul->head, rt6i_uncached) {
0173             struct inet6_dev *rt_idev = rt->rt6i_idev;
0174             struct net_device *rt_dev = rt->dst.dev;
0175             bool handled = false;
0176
0177             if (rt_idev->dev == dev) {
0178                 rt->rt6i_idev = in6_dev_get(blackhole_netdev);
0179                 in6_dev_put(rt_idev);
0180                 handled = true;
0181             }
0182
0183             if (rt_dev == dev) {
0184                 rt->dst.dev = blackhole_netdev;
0185                 netdev_ref_replace(rt_dev, blackhole_netdev,
0186                            &rt->dst.dev_tracker,
0187                            GFP_ATOMIC);
0188                 handled = true;
0189             }
0190             if (handled)
0191                 list_move(&rt->rt6i_uncached,
0192                       &ul->quarantine);
0193         }
0194         spin_unlock_bh(&ul->lock);
0195     }
0196 }
0197
0198 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
0199                          struct sk_buff *skb,
0200                          const void *daddr)
0201 {
0202     if (!ipv6_addr_any(p))
0203         return (const void *) p;
0204     else if (skb)
0205         return &ipv6_hdr(skb)->daddr;
0206     return daddr;
0207 }
0208
0209 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
0210                    struct net_device *dev,
0211                    struct sk_buff *skb,
0212                    const void *daddr)
0213 {
0214     struct neighbour *n;
0215
0216     daddr = choose_neigh_daddr(gw, skb, daddr);
0217     n = __ipv6_neigh_lookup(dev, daddr);
0218     if (n)
0219         return n;
0220
0221     n = neigh_create(&nd_tbl, daddr, dev);
0222     return IS_ERR(n) ? NULL : n;
0223 }
0224
0225 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
0226                           struct sk_buff *skb,
0227                           const void *daddr)
0228 {
0229     const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
0230
0231     return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
0232                 dst->dev, skb, daddr);
0233 }
0234
0235 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
0236 {
0237     struct net_device *dev = dst->dev;
0238     struct rt6_info *rt = (struct rt6_info *)dst;
0239
0240     daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
0241     if (!daddr)
0242         return;
0243     if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
0244         return;
0245     if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
0246         return;
0247     __ipv6_confirm_neigh(dev, daddr);
0248 }
0249
0250 static struct dst_ops ip6_dst_ops_template = {
0251     .family         =   AF_INET6,
0252     .gc         =   ip6_dst_gc,
0253     .gc_thresh      =   1024,
0254     .check          =   ip6_dst_check,
0255     .default_advmss     =   ip6_default_advmss,
0256     .mtu            =   ip6_mtu,
0257     .cow_metrics        =   dst_cow_metrics_generic,
0258     .destroy        =   ip6_dst_destroy,
0259     .ifdown         =   ip6_dst_ifdown,
0260     .negative_advice    =   ip6_negative_advice,
0261     .link_failure       =   ip6_link_failure,
0262     .update_pmtu        =   ip6_rt_update_pmtu,
0263     .redirect       =   rt6_do_redirect,
0264     .local_out      =   __ip6_local_out,
0265     .neigh_lookup       =   ip6_dst_neigh_lookup,
0266     .confirm_neigh      =   ip6_confirm_neigh,
0267 };
0268
0269 static struct dst_ops ip6_dst_blackhole_ops = {
0270     .family         = AF_INET6,
0271     .default_advmss     = ip6_default_advmss,
0272     .neigh_lookup       = ip6_dst_neigh_lookup,
0273     .check          = ip6_dst_check,
0274     .destroy        = ip6_dst_destroy,
0275     .cow_metrics        = dst_cow_metrics_generic,
0276     .update_pmtu        = dst_blackhole_update_pmtu,
0277     .redirect       = dst_blackhole_redirect,
0278     .mtu            = dst_blackhole_mtu,
0279 };
0280
0281 static const u32 ip6_template_metrics[RTAX_MAX] = {
0282     [RTAX_HOPLIMIT - 1] = 0,
0283 };
0284
0285 static const struct fib6_info fib6_null_entry_template = {
0286     .fib6_flags = (RTF_REJECT | RTF_NONEXTHOP),
0287     .fib6_protocol  = RTPROT_KERNEL,
0288     .fib6_metric    = ~(u32)0,
0289     .fib6_ref   = REFCOUNT_INIT(1),
0290     .fib6_type  = RTN_UNREACHABLE,
0291     .fib6_metrics   = (struct dst_metrics *)&dst_default_metrics,
0292 };
0293
0294 static const struct rt6_info ip6_null_entry_template = {
0295     .dst = {
0296         .__refcnt   = ATOMIC_INIT(1),
0297         .__use      = 1,
0298         .obsolete   = DST_OBSOLETE_FORCE_CHK,
0299         .error      = -ENETUNREACH,
0300         .input      = ip6_pkt_discard,
0301         .output     = ip6_pkt_discard_out,
0302     },
0303     .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
0304 };
0305
0306 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
0307
0308 static const struct rt6_info ip6_prohibit_entry_template = {
0309     .dst = {
0310         .__refcnt   = ATOMIC_INIT(1),
0311         .__use      = 1,
0312         .obsolete   = DST_OBSOLETE_FORCE_CHK,
0313         .error      = -EACCES,
0314         .input      = ip6_pkt_prohibit,
0315         .output     = ip6_pkt_prohibit_out,
0316     },
0317     .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
0318 };
0319
0320 static const struct rt6_info ip6_blk_hole_entry_template = {
0321     .dst = {
0322         .__refcnt   = ATOMIC_INIT(1),
0323         .__use      = 1,
0324         .obsolete   = DST_OBSOLETE_FORCE_CHK,
0325         .error      = -EINVAL,
0326         .input      = dst_discard,
0327         .output     = dst_discard_out,
0328     },
0329     .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
0330 };
0331
0332 #endif
0333
0334 static void rt6_info_init(struct rt6_info *rt)
0335 {
0336     memset_after(rt, 0, dst);
0337     INIT_LIST_HEAD(&rt->rt6i_uncached);
0338 }
0339
0340 /* allocate dst with ip6_dst_ops */
0341 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
0342                    int flags)
0343 {
0344     struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
0345                     1, DST_OBSOLETE_FORCE_CHK, flags);
0346
0347     if (rt) {
0348         rt6_info_init(rt);
0349         atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
0350     }
0351
0352     return rt;
0353 }
0354 EXPORT_SYMBOL(ip6_dst_alloc);
0355
0356 static void ip6_dst_destroy(struct dst_entry *dst)
0357 {
0358     struct rt6_info *rt = (struct rt6_info *)dst;
0359     struct fib6_info *from;
0360     struct inet6_dev *idev;
0361
0362     ip_dst_metrics_put(dst);
0363     rt6_uncached_list_del(rt);
0364
0365     idev = rt->rt6i_idev;
0366     if (idev) {
0367         rt->rt6i_idev = NULL;
0368         in6_dev_put(idev);
0369     }
0370
0371     from = xchg((__force struct fib6_info **)&rt->from, NULL);
0372     fib6_info_release(from);
0373 }
0374
0375 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
0376                int how)
0377 {
0378     struct rt6_info *rt = (struct rt6_info *)dst;
0379     struct inet6_dev *idev = rt->rt6i_idev;
0380
0381     if (idev && idev->dev != blackhole_netdev) {
0382         struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
0383
0384         if (blackhole_idev) {
0385             rt->rt6i_idev = blackhole_idev;
0386             in6_dev_put(idev);
0387         }
0388     }
0389 }
0390
0391 static bool __rt6_check_expired(const struct rt6_info *rt)
0392 {
0393     if (rt->rt6i_flags & RTF_EXPIRES)
0394         return time_after(jiffies, rt->dst.expires);
0395     else
0396         return false;
0397 }
0398
0399 static bool rt6_check_expired(const struct rt6_info *rt)
0400 {
0401     struct fib6_info *from;
0402
0403     from = rcu_dereference(rt->from);
0404
0405     if (rt->rt6i_flags & RTF_EXPIRES) {
0406         if (time_after(jiffies, rt->dst.expires))
0407             return true;
0408     } else if (from) {
0409         return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
0410             fib6_check_expired(from);
0411     }
0412     return false;
0413 }
0414
0415 void fib6_select_path(const struct net *net, struct fib6_result *res,
0416               struct flowi6 *fl6, int oif, bool have_oif_match,
0417               const struct sk_buff *skb, int strict)
0418 {
0419     struct fib6_info *sibling, *next_sibling;
0420     struct fib6_info *match = res->f6i;
0421
0422     if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
0423         goto out;
0424
0425     if (match->nh && have_oif_match && res->nh)
0426         return;
0427
0428     /* We might have already computed the hash for ICMPv6 errors. In such
0429      * case it will always be non-zero. Otherwise now is the time to do it.
0430      */
0431     if (!fl6->mp_hash &&
0432         (!match->nh || nexthop_is_multipath(match->nh)))
0433         fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
0434
0435     if (unlikely(match->nh)) {
0436         nexthop_path_fib6_result(res, fl6->mp_hash);
0437         return;
0438     }
0439
0440     if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
0441         goto out;
0442
0443     list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
0444                  fib6_siblings) {
0445         const struct fib6_nh *nh = sibling->fib6_nh;
0446         int nh_upper_bound;
0447
0448         nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
0449         if (fl6->mp_hash > nh_upper_bound)
0450             continue;
0451         if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
0452             break;
0453         match = sibling;
0454         break;
0455     }
0456
0457 out:
0458     res->f6i = match;
0459     res->nh = match->fib6_nh;
0460 }
0461
0462 /*
0463  *  Route lookup. rcu_read_lock() should be held.
0464  */
0465
0466 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
0467                    const struct in6_addr *saddr, int oif, int flags)
0468 {
0469     const struct net_device *dev;
0470
0471     if (nh->fib_nh_flags & RTNH_F_DEAD)
0472         return false;
0473
0474     dev = nh->fib_nh_dev;
0475     if (oif) {
0476         if (dev->ifindex == oif)
0477             return true;
0478     } else {
0479         if (ipv6_chk_addr(net, saddr, dev,
0480                   flags & RT6_LOOKUP_F_IFACE))
0481             return true;
0482     }
0483
0484     return false;
0485 }
0486
0487 struct fib6_nh_dm_arg {
0488     struct net      *net;
0489     const struct in6_addr   *saddr;
0490     int         oif;
0491     int         flags;
0492     struct fib6_nh      *nh;
0493 };
0494
0495 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
0496 {
0497     struct fib6_nh_dm_arg *arg = _arg;
0498
0499     arg->nh = nh;
0500     return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
0501                   arg->flags);
0502 }
0503
0504 /* returns fib6_nh from nexthop or NULL */
0505 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
0506                     struct fib6_result *res,
0507                     const struct in6_addr *saddr,
0508                     int oif, int flags)
0509 {
0510     struct fib6_nh_dm_arg arg = {
0511         .net   = net,
0512         .saddr = saddr,
0513         .oif   = oif,
0514         .flags = flags,
0515     };
0516
0517     if (nexthop_is_blackhole(nh))
0518         return NULL;
0519
0520     if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
0521         return arg.nh;
0522
0523     return NULL;
0524 }
0525
0526 static void rt6_device_match(struct net *net, struct fib6_result *res,
0527                  const struct in6_addr *saddr, int oif, int flags)
0528 {
0529     struct fib6_info *f6i = res->f6i;
0530     struct fib6_info *spf6i;
0531     struct fib6_nh *nh;
0532
0533     if (!oif && ipv6_addr_any(saddr)) {
0534         if (unlikely(f6i->nh)) {
0535             nh = nexthop_fib6_nh(f6i->nh);
0536             if (nexthop_is_blackhole(f6i->nh))
0537                 goto out_blackhole;
0538         } else {
0539             nh = f6i->fib6_nh;
0540         }
0541         if (!(nh->fib_nh_flags & RTNH_F_DEAD))
0542             goto out;
0543     }
0544
0545     for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
0546         bool matched = false;
0547
0548         if (unlikely(spf6i->nh)) {
0549             nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
0550                           oif, flags);
0551             if (nh)
0552                 matched = true;
0553         } else {
0554             nh = spf6i->fib6_nh;
0555             if (__rt6_device_match(net, nh, saddr, oif, flags))
0556                 matched = true;
0557         }
0558         if (matched) {
0559             res->f6i = spf6i;
0560             goto out;
0561         }
0562     }
0563
0564     if (oif && flags & RT6_LOOKUP_F_IFACE) {
0565         res->f6i = net->ipv6.fib6_null_entry;
0566         nh = res->f6i->fib6_nh;
0567         goto out;
0568     }
0569
0570     if (unlikely(f6i->nh)) {
0571         nh = nexthop_fib6_nh(f6i->nh);
0572         if (nexthop_is_blackhole(f6i->nh))
0573             goto out_blackhole;
0574     } else {
0575         nh = f6i->fib6_nh;
0576     }
0577
0578     if (nh->fib_nh_flags & RTNH_F_DEAD) {
0579         res->f6i = net->ipv6.fib6_null_entry;
0580         nh = res->f6i->fib6_nh;
0581     }
0582 out:
0583     res->nh = nh;
0584     res->fib6_type = res->f6i->fib6_type;
0585     res->fib6_flags = res->f6i->fib6_flags;
0586     return;
0587
0588 out_blackhole:
0589     res->fib6_flags |= RTF_REJECT;
0590     res->fib6_type = RTN_BLACKHOLE;
0591     res->nh = nh;
0592 }
0593
0594 #ifdef CONFIG_IPV6_ROUTER_PREF
0595 struct __rt6_probe_work {
0596     struct work_struct work;
0597     struct in6_addr target;
0598     struct net_device *dev;
0599     netdevice_tracker dev_tracker;
0600 };
0601
0602 static void rt6_probe_deferred(struct work_struct *w)
0603 {
0604     struct in6_addr mcaddr;
0605     struct __rt6_probe_work *work =
0606         container_of(w, struct __rt6_probe_work, work);
0607
0608     addrconf_addr_solict_mult(&work->target, &mcaddr);
0609     ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
0610     netdev_put(work->dev, &work->dev_tracker);
0611     kfree(work);
0612 }
0613
0614 static void rt6_probe(struct fib6_nh *fib6_nh)
0615 {
0616     struct __rt6_probe_work *work = NULL;
0617     const struct in6_addr *nh_gw;
0618     unsigned long last_probe;
0619     struct neighbour *neigh;
0620     struct net_device *dev;
0621     struct inet6_dev *idev;
0622
0623     /*
0624      * Okay, this does not seem to be appropriate
0625      * for now, however, we need to check if it
0626      * is really so; aka Router Reachability Probing.
0627      *
0628      * Router Reachability Probe MUST be rate-limited
0629      * to no more than one per minute.
0630      */
0631     if (!fib6_nh->fib_nh_gw_family)
0632         return;
0633
0634     nh_gw = &fib6_nh->fib_nh_gw6;
0635     dev = fib6_nh->fib_nh_dev;
0636     rcu_read_lock_bh();
0637     last_probe = READ_ONCE(fib6_nh->last_probe);
0638     idev = __in6_dev_get(dev);
0639     neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
0640     if (neigh) {
0641         if (neigh->nud_state & NUD_VALID)
0642             goto out;
0643
0644         write_lock(&neigh->lock);
0645         if (!(neigh->nud_state & NUD_VALID) &&
0646             time_after(jiffies,
0647                    neigh->updated + idev->cnf.rtr_probe_interval)) {
0648             work = kmalloc(sizeof(*work), GFP_ATOMIC);
0649             if (work)
0650                 __neigh_set_probe_once(neigh);
0651         }
0652         write_unlock(&neigh->lock);
0653     } else if (time_after(jiffies, last_probe +
0654                        idev->cnf.rtr_probe_interval)) {
0655         work = kmalloc(sizeof(*work), GFP_ATOMIC);
0656     }
0657
0658     if (!work || cmpxchg(&fib6_nh->last_probe,
0659                  last_probe, jiffies) != last_probe) {
0660         kfree(work);
0661     } else {
0662         INIT_WORK(&work->work, rt6_probe_deferred);
0663         work->target = *nh_gw;
0664         netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
0665         work->dev = dev;
0666         schedule_work(&work->work);
0667     }
0668
0669 out:
0670     rcu_read_unlock_bh();
0671 }
0672 #else
0673 static inline void rt6_probe(struct fib6_nh *fib6_nh)
0674 {
0675 }
0676 #endif
0677
0678 /*
0679  * Default Router Selection (RFC 2461 6.3.6)
0680  */
0681 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
0682 {
0683     enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
0684     struct neighbour *neigh;
0685
0686     rcu_read_lock_bh();
0687     neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
0688                       &fib6_nh->fib_nh_gw6);
0689     if (neigh) {
0690         read_lock(&neigh->lock);
0691         if (neigh->nud_state & NUD_VALID)
0692             ret = RT6_NUD_SUCCEED;
0693 #ifdef CONFIG_IPV6_ROUTER_PREF
0694         else if (!(neigh->nud_state & NUD_FAILED))
0695             ret = RT6_NUD_SUCCEED;
0696         else
0697             ret = RT6_NUD_FAIL_PROBE;
0698 #endif
0699         read_unlock(&neigh->lock);
0700     } else {
0701         ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
0702               RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
0703     }
0704     rcu_read_unlock_bh();
0705
0706     return ret;
0707 }
0708
0709 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
0710                int strict)
0711 {
0712     int m = 0;
0713
0714     if (!oif || nh->fib_nh_dev->ifindex == oif)
0715         m = 2;
0716
0717     if (!m && (strict & RT6_LOOKUP_F_IFACE))
0718         return RT6_NUD_FAIL_HARD;
0719 #ifdef CONFIG_IPV6_ROUTER_PREF
0720     m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
0721 #endif
0722     if ((strict & RT6_LOOKUP_F_REACHABLE) &&
0723         !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
0724         int n = rt6_check_neigh(nh);
0725         if (n < 0)
0726             return n;
0727     }
0728     return m;
0729 }
0730
0731 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
0732                int oif, int strict, int *mpri, bool *do_rr)
0733 {
0734     bool match_do_rr = false;
0735     bool rc = false;
0736     int m;
0737
0738     if (nh->fib_nh_flags & RTNH_F_DEAD)
0739         goto out;
0740
0741     if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
0742         nh->fib_nh_flags & RTNH_F_LINKDOWN &&
0743         !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
0744         goto out;
0745
0746     m = rt6_score_route(nh, fib6_flags, oif, strict);
0747     if (m == RT6_NUD_FAIL_DO_RR) {
0748         match_do_rr = true;
0749         m = 0; /* lowest valid score */
0750     } else if (m == RT6_NUD_FAIL_HARD) {
0751         goto out;
0752     }
0753
0754     if (strict & RT6_LOOKUP_F_REACHABLE)
0755         rt6_probe(nh);
0756
0757     /* note that m can be RT6_NUD_FAIL_PROBE at this point */
0758     if (m > *mpri) {
0759         *do_rr = match_do_rr;
0760         *mpri = m;
0761         rc = true;
0762     }
0763 out:
0764     return rc;
0765 }
0766
0767 struct fib6_nh_frl_arg {
0768     u32     flags;
0769     int     oif;
0770     int     strict;
0771     int     *mpri;
0772     bool        *do_rr;
0773     struct fib6_nh  *nh;
0774 };
0775
0776 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
0777 {
0778     struct fib6_nh_frl_arg *arg = _arg;
0779
0780     arg->nh = nh;
0781     return find_match(nh, arg->flags, arg->oif, arg->strict,
0782               arg->mpri, arg->do_rr);
0783 }
0784
0785 static void __find_rr_leaf(struct fib6_info *f6i_start,
0786                struct fib6_info *nomatch, u32 metric,
0787                struct fib6_result *res, struct fib6_info **cont,
0788                int oif, int strict, bool *do_rr, int *mpri)
0789 {
0790     struct fib6_info *f6i;
0791
0792     for (f6i = f6i_start;
0793          f6i && f6i != nomatch;
0794          f6i = rcu_dereference(f6i->fib6_next)) {
0795         bool matched = false;
0796         struct fib6_nh *nh;
0797
0798         if (cont && f6i->fib6_metric != metric) {
0799             *cont = f6i;
0800             return;
0801         }
0802
0803         if (fib6_check_expired(f6i))
0804             continue;
0805
0806         if (unlikely(f6i->nh)) {
0807             struct fib6_nh_frl_arg arg = {
0808                 .flags  = f6i->fib6_flags,
0809                 .oif    = oif,
0810                 .strict = strict,
0811                 .mpri   = mpri,
0812                 .do_rr  = do_rr
0813             };
0814
0815             if (nexthop_is_blackhole(f6i->nh)) {
0816                 res->fib6_flags = RTF_REJECT;
0817                 res->fib6_type = RTN_BLACKHOLE;
0818                 res->f6i = f6i;
0819                 res->nh = nexthop_fib6_nh(f6i->nh);
0820                 return;
0821             }
0822             if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
0823                              &arg)) {
0824                 matched = true;
0825                 nh = arg.nh;
0826             }
0827         } else {
0828             nh = f6i->fib6_nh;
0829             if (find_match(nh, f6i->fib6_flags, oif, strict,
0830                        mpri, do_rr))
0831                 matched = true;
0832         }
0833         if (matched) {
0834             res->f6i = f6i;
0835             res->nh = nh;
0836             res->fib6_flags = f6i->fib6_flags;
0837             res->fib6_type = f6i->fib6_type;
0838         }
0839     }
0840 }
0841
0842 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
0843              struct fib6_info *rr_head, int oif, int strict,
0844              bool *do_rr, struct fib6_result *res)
0845 {
0846     u32 metric = rr_head->fib6_metric;
0847     struct fib6_info *cont = NULL;
0848     int mpri = -1;
0849
0850     __find_rr_leaf(rr_head, NULL, metric, res, &cont,
0851                oif, strict, do_rr, &mpri);
0852
0853     __find_rr_leaf(leaf, rr_head, metric, res, &cont,
0854                oif, strict, do_rr, &mpri);
0855
0856     if (res->f6i || !cont)
0857         return;
0858
0859     __find_rr_leaf(cont, NULL, metric, res, NULL,
0860                oif, strict, do_rr, &mpri);
0861 }
0862
0863 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
0864                struct fib6_result *res, int strict)
0865 {
0866     struct fib6_info *leaf = rcu_dereference(fn->leaf);
0867     struct fib6_info *rt0;
0868     bool do_rr = false;
0869     int key_plen;
0870
0871     /* make sure this function or its helpers sets f6i */
0872     res->f6i = NULL;
0873
0874     if (!leaf || leaf == net->ipv6.fib6_null_entry)
0875         goto out;
0876
0877     rt0 = rcu_dereference(fn->rr_ptr);
0878     if (!rt0)
0879         rt0 = leaf;
0880
0881     /* Double check to make sure fn is not an intermediate node
0882      * and fn->leaf does not points to its child's leaf
0883      * (This might happen if all routes under fn are deleted from
0884      * the tree and fib6_repair_tree() is called on the node.)
0885      */
0886     key_plen = rt0->fib6_dst.plen;
0887 #ifdef CONFIG_IPV6_SUBTREES
0888     if (rt0->fib6_src.plen)
0889         key_plen = rt0->fib6_src.plen;
0890 #endif
0891     if (fn->fn_bit != key_plen)
0892         goto out;
0893
0894     find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
0895     if (do_rr) {
0896         struct fib6_info *next = rcu_dereference(rt0->fib6_next);
0897
0898         /* no entries matched; do round-robin */
0899         if (!next || next->fib6_metric != rt0->fib6_metric)
0900             next = leaf;
0901
0902         if (next != rt0) {
0903             spin_lock_bh(&leaf->fib6_table->tb6_lock);
0904             /* make sure next is not being deleted from the tree */
0905             if (next->fib6_node)
0906                 rcu_assign_pointer(fn->rr_ptr, next);
0907             spin_unlock_bh(&leaf->fib6_table->tb6_lock);
0908         }
0909     }
0910
0911 out:
0912     if (!res->f6i) {
0913         res->f6i = net->ipv6.fib6_null_entry;
0914         res->nh = res->f6i->fib6_nh;
0915         res->fib6_flags = res->f6i->fib6_flags;
0916         res->fib6_type = res->f6i->fib6_type;
0917     }
0918 }
0919
0920 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
0921 {
0922     return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
0923            res->nh->fib_nh_gw_family;
0924 }
0925
0926 #ifdef CONFIG_IPV6_ROUTE_INFO
0927 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
0928           const struct in6_addr *gwaddr)
0929 {
0930     struct net *net = dev_net(dev);
0931     struct route_info *rinfo = (struct route_info *) opt;
0932     struct in6_addr prefix_buf, *prefix;
0933     unsigned int pref;
0934     unsigned long lifetime;
0935     struct fib6_info *rt;
0936
0937     if (len < sizeof(struct route_info)) {
0938         return -EINVAL;
0939     }
0940
0941     /* Sanity check for prefix_len and length */
0942     if (rinfo->length > 3) {
0943         return -EINVAL;
0944     } else if (rinfo->prefix_len > 128) {
0945         return -EINVAL;
0946     } else if (rinfo->prefix_len > 64) {
0947         if (rinfo->length < 2) {
0948             return -EINVAL;
0949         }
0950     } else if (rinfo->prefix_len > 0) {
0951         if (rinfo->length < 1) {
0952             return -EINVAL;
0953         }
0954     }
0955
0956     pref = rinfo->route_pref;
0957     if (pref == ICMPV6_ROUTER_PREF_INVALID)
0958         return -EINVAL;
0959
0960     lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
0961
0962     if (rinfo->length == 3)
0963         prefix = (struct in6_addr *)rinfo->prefix;
0964     else {
0965         /* this function is safe */
0966         ipv6_addr_prefix(&prefix_buf,
0967                  (struct in6_addr *)rinfo->prefix,
0968                  rinfo->prefix_len);
0969         prefix = &prefix_buf;
0970     }
0971
0972     if (rinfo->prefix_len == 0)
0973         rt = rt6_get_dflt_router(net, gwaddr, dev);
0974     else
0975         rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
0976                     gwaddr, dev);
0977
0978     if (rt && !lifetime) {
0979         ip6_del_rt(net, rt, false);
0980         rt = NULL;
0981     }
0982
0983     if (!rt && lifetime)
0984         rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
0985                     dev, pref);
0986     else if (rt)
0987         rt->fib6_flags = RTF_ROUTEINFO |
0988                  (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
0989
0990     if (rt) {
0991         if (!addrconf_finite_timeout(lifetime))
0992             fib6_clean_expires(rt);
0993         else
0994             fib6_set_expires(rt, jiffies + HZ * lifetime);
0995
0996         fib6_info_release(rt);
0997     }
0998     return 0;
0999 }
1000 #endif
1001
1002 /*
1003  *  Misc support functions
1004  */
1005
1006 /* called with rcu_lock held */
1007 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1008 {
1009     struct net_device *dev = res->nh->fib_nh_dev;
1010
1011     if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1012         /* for copies of local routes, dst->dev needs to be the
1013          * device if it is a master device, the master device if
1014          * device is enslaved, and the loopback as the default
1015          */
1016         if (netif_is_l3_slave(dev) &&
1017             !rt6_need_strict(&res->f6i->fib6_dst.addr))
1018             dev = l3mdev_master_dev_rcu(dev);
1019         else if (!netif_is_l3_master(dev))
1020             dev = dev_net(dev)->loopback_dev;
1021         /* last case is netif_is_l3_master(dev) is true in which
1022          * case we want dev returned to be dev
1023          */
1024     }
1025
1026     return dev;
1027 }
1028
1029 static const int fib6_prop[RTN_MAX + 1] = {
1030     [RTN_UNSPEC]    = 0,
1031     [RTN_UNICAST]   = 0,
1032     [RTN_LOCAL] = 0,
1033     [RTN_BROADCAST] = 0,
1034     [RTN_ANYCAST]   = 0,
1035     [RTN_MULTICAST] = 0,
1036     [RTN_BLACKHOLE] = -EINVAL,
1037     [RTN_UNREACHABLE] = -EHOSTUNREACH,
1038     [RTN_PROHIBIT]  = -EACCES,
1039     [RTN_THROW] = -EAGAIN,
1040     [RTN_NAT]   = -EINVAL,
1041     [RTN_XRESOLVE]  = -EINVAL,
1042 };
1043
1044 static int ip6_rt_type_to_error(u8 fib6_type)
1045 {
1046     return fib6_prop[fib6_type];
1047 }
1048
1049 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1050 {
1051     unsigned short flags = 0;
1052
1053     if (rt->dst_nocount)
1054         flags |= DST_NOCOUNT;
1055     if (rt->dst_nopolicy)
1056         flags |= DST_NOPOLICY;
1057
1058     return flags;
1059 }
1060
1061 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1062 {
1063     rt->dst.error = ip6_rt_type_to_error(fib6_type);
1064
1065     switch (fib6_type) {
1066     case RTN_BLACKHOLE:
1067         rt->dst.output = dst_discard_out;
1068         rt->dst.input = dst_discard;
1069         break;
1070     case RTN_PROHIBIT:
1071         rt->dst.output = ip6_pkt_prohibit_out;
1072         rt->dst.input = ip6_pkt_prohibit;
1073         break;
1074     case RTN_THROW:
1075     case RTN_UNREACHABLE:
1076     default:
1077         rt->dst.output = ip6_pkt_discard_out;
1078         rt->dst.input = ip6_pkt_discard;
1079         break;
1080     }
1081 }
1082
1083 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1084 {
1085     struct fib6_info *f6i = res->f6i;
1086
1087     if (res->fib6_flags & RTF_REJECT) {
1088         ip6_rt_init_dst_reject(rt, res->fib6_type);
1089         return;
1090     }
1091
1092     rt->dst.error = 0;
1093     rt->dst.output = ip6_output;
1094
1095     if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1096         rt->dst.input = ip6_input;
1097     } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1098         rt->dst.input = ip6_mc_input;
1099     } else {
1100         rt->dst.input = ip6_forward;
1101     }
1102
1103     if (res->nh->fib_nh_lws) {
1104         rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1105         lwtunnel_set_redirect(&rt->dst);
1106     }
1107
1108     rt->dst.lastuse = jiffies;
1109 }
1110
1111 /* Caller must already hold reference to @from */
1112 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1113 {
1114     rt->rt6i_flags &= ~RTF_EXPIRES;
1115     rcu_assign_pointer(rt->from, from);
1116     ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1117 }
1118
1119 /* Caller must already hold reference to f6i in result */
1120 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1121 {
1122     const struct fib6_nh *nh = res->nh;
1123     const struct net_device *dev = nh->fib_nh_dev;
1124     struct fib6_info *f6i = res->f6i;
1125
1126     ip6_rt_init_dst(rt, res);
1127
1128     rt->rt6i_dst = f6i->fib6_dst;
1129     rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1130     rt->rt6i_flags = res->fib6_flags;
1131     if (nh->fib_nh_gw_family) {
1132         rt->rt6i_gateway = nh->fib_nh_gw6;
1133         rt->rt6i_flags |= RTF_GATEWAY;
1134     }
1135     rt6_set_from(rt, f6i);
1136 #ifdef CONFIG_IPV6_SUBTREES
1137     rt->rt6i_src = f6i->fib6_src;
1138 #endif
1139 }
1140
1141 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1142                     struct in6_addr *saddr)
1143 {
1144     struct fib6_node *pn, *sn;
1145     while (1) {
1146         if (fn->fn_flags & RTN_TL_ROOT)
1147             return NULL;
1148         pn = rcu_dereference(fn->parent);
1149         sn = FIB6_SUBTREE(pn);
1150         if (sn && sn != fn)
1151             fn = fib6_node_lookup(sn, NULL, saddr);
1152         else
1153             fn = pn;
1154         if (fn->fn_flags & RTN_RTINFO)
1155             return fn;
1156     }
1157 }
1158
1159 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1160 {
1161     struct rt6_info *rt = *prt;
1162
1163     if (dst_hold_safe(&rt->dst))
1164         return true;
1165     if (net) {
1166         rt = net->ipv6.ip6_null_entry;
1167         dst_hold(&rt->dst);
1168     } else {
1169         rt = NULL;
1170     }
1171     *prt = rt;
1172     return false;
1173 }
1174
1175 /* called with rcu_lock held */
1176 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1177 {
1178     struct net_device *dev = res->nh->fib_nh_dev;
1179     struct fib6_info *f6i = res->f6i;
1180     unsigned short flags;
1181     struct rt6_info *nrt;
1182
1183     if (!fib6_info_hold_safe(f6i))
1184         goto fallback;
1185
1186     flags = fib6_info_dst_flags(f6i);
1187     nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1188     if (!nrt) {
1189         fib6_info_release(f6i);
1190         goto fallback;
1191     }
1192
1193     ip6_rt_copy_init(nrt, res);
1194     return nrt;
1195
1196 fallback:
1197     nrt = dev_net(dev)->ipv6.ip6_null_entry;
1198     dst_hold(&nrt->dst);
1199     return nrt;
1200 }
1201
1202 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1203                          struct fib6_table *table,
1204                          struct flowi6 *fl6,
1205                          const struct sk_buff *skb,
1206                          int flags)
1207 {
1208     struct fib6_result res = {};
1209     struct fib6_node *fn;
1210     struct rt6_info *rt;
1211
1212     rcu_read_lock();
1213     fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1214 restart:
1215     res.f6i = rcu_dereference(fn->leaf);
1216     if (!res.f6i)
1217         res.f6i = net->ipv6.fib6_null_entry;
1218     else
1219         rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1220                  flags);
1221
1222     if (res.f6i == net->ipv6.fib6_null_entry) {
1223         fn = fib6_backtrack(fn, &fl6->saddr);
1224         if (fn)
1225             goto restart;
1226
1227         rt = net->ipv6.ip6_null_entry;
1228         dst_hold(&rt->dst);
1229         goto out;
1230     } else if (res.fib6_flags & RTF_REJECT) {
1231         goto do_create;
1232     }
1233
1234     fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1235              fl6->flowi6_oif != 0, skb, flags);
1236
1237     /* Search through exception table */
1238     rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1239     if (rt) {
1240         if (ip6_hold_safe(net, &rt))
1241             dst_use_noref(&rt->dst, jiffies);
1242     } else {
1243 do_create:
1244         rt = ip6_create_rt_rcu(&res);
1245     }
1246
1247 out:
1248     trace_fib6_table_lookup(net, &res, table, fl6);
1249
1250     rcu_read_unlock();
1251
1252     return rt;
1253 }
1254
1255 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1256                    const struct sk_buff *skb, int flags)
1257 {
1258     return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1259 }
1260 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1261
1262 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1263                 const struct in6_addr *saddr, int oif,
1264                 const struct sk_buff *skb, int strict)
1265 {
1266     struct flowi6 fl6 = {
1267         .flowi6_oif = oif,
1268         .daddr = *daddr,
1269     };
1270     struct dst_entry *dst;
1271     int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1272
1273     if (saddr) {
1274         memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1275         flags |= RT6_LOOKUP_F_HAS_SADDR;
1276     }
1277
1278     dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1279     if (dst->error == 0)
1280         return (struct rt6_info *) dst;
1281
1282     dst_release(dst);
1283
1284     return NULL;
1285 }
1286 EXPORT_SYMBOL(rt6_lookup);
1287
1288 /* ip6_ins_rt is called with FREE table->tb6_lock.
1289  * It takes new route entry, the addition fails by any reason the
1290  * route is released.
1291  * Caller must hold dst before calling it.
1292  */
1293
1294 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1295             struct netlink_ext_ack *extack)
1296 {
1297     int err;
1298     struct fib6_table *table;
1299
1300     table = rt->fib6_table;
1301     spin_lock_bh(&table->tb6_lock);
1302     err = fib6_add(&table->tb6_root, rt, info, extack);
1303     spin_unlock_bh(&table->tb6_lock);
1304
1305     return err;
1306 }
1307
1308 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1309 {
1310     struct nl_info info = { .nl_net = net, };
1311
1312     return __ip6_ins_rt(rt, &info, NULL);
1313 }
1314
1315 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1316                        const struct in6_addr *daddr,
1317                        const struct in6_addr *saddr)
1318 {
1319     struct fib6_info *f6i = res->f6i;
1320     struct net_device *dev;
1321     struct rt6_info *rt;
1322
1323     /*
1324      *  Clone the route.
1325      */
1326
1327     if (!fib6_info_hold_safe(f6i))
1328         return NULL;
1329
1330     dev = ip6_rt_get_dev_rcu(res);
1331     rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1332     if (!rt) {
1333         fib6_info_release(f6i);
1334         return NULL;
1335     }
1336
1337     ip6_rt_copy_init(rt, res);
1338     rt->rt6i_flags |= RTF_CACHE;
1339     rt->rt6i_dst.addr = *daddr;
1340     rt->rt6i_dst.plen = 128;
1341
1342     if (!rt6_is_gw_or_nonexthop(res)) {
1343         if (f6i->fib6_dst.plen != 128 &&
1344             ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1345             rt->rt6i_flags |= RTF_ANYCAST;
1346 #ifdef CONFIG_IPV6_SUBTREES
1347         if (rt->rt6i_src.plen && saddr) {
1348             rt->rt6i_src.addr = *saddr;
1349             rt->rt6i_src.plen = 128;
1350         }
1351 #endif
1352     }
1353
1354     return rt;
1355 }
1356
1357 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1358 {
1359     struct fib6_info *f6i = res->f6i;
1360     unsigned short flags = fib6_info_dst_flags(f6i);
1361     struct net_device *dev;
1362     struct rt6_info *pcpu_rt;
1363
1364     if (!fib6_info_hold_safe(f6i))
1365         return NULL;
1366
1367     rcu_read_lock();
1368     dev = ip6_rt_get_dev_rcu(res);
1369     pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1370     rcu_read_unlock();
1371     if (!pcpu_rt) {
1372         fib6_info_release(f6i);
1373         return NULL;
1374     }
1375     ip6_rt_copy_init(pcpu_rt, res);
1376     pcpu_rt->rt6i_flags |= RTF_PCPU;
1377
1378     if (f6i->nh)
1379         pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1380
1381     return pcpu_rt;
1382 }
1383
1384 static bool rt6_is_valid(const struct rt6_info *rt6)
1385 {
1386     return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1387 }
1388
1389 /* It should be called with rcu_read_lock() acquired */
1390 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1391 {
1392     struct rt6_info *pcpu_rt;
1393
1394     pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1395
1396     if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1397         struct rt6_info *prev, **p;
1398
1399         p = this_cpu_ptr(res->nh->rt6i_pcpu);
1400         prev = xchg(p, NULL);
1401         if (prev) {
1402             dst_dev_put(&prev->dst);
1403             dst_release(&prev->dst);
1404         }
1405
1406         pcpu_rt = NULL;
1407     }
1408
1409     return pcpu_rt;
1410 }
1411
1412 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1413                         const struct fib6_result *res)
1414 {
1415     struct rt6_info *pcpu_rt, *prev, **p;
1416
1417     pcpu_rt = ip6_rt_pcpu_alloc(res);
1418     if (!pcpu_rt)
1419         return NULL;
1420
1421     p = this_cpu_ptr(res->nh->rt6i_pcpu);
1422     prev = cmpxchg(p, NULL, pcpu_rt);
1423     BUG_ON(prev);
1424
1425     if (res->f6i->fib6_destroying) {
1426         struct fib6_info *from;
1427
1428         from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1429         fib6_info_release(from);
1430     }
1431
1432     return pcpu_rt;
1433 }
1434
1435 /* exception hash table implementation
1436  */
1437 static DEFINE_SPINLOCK(rt6_exception_lock);
1438
1439 /* Remove rt6_ex from hash table and free the memory
1440  * Caller must hold rt6_exception_lock
1441  */
1442 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1443                  struct rt6_exception *rt6_ex)
1444 {
1445     struct fib6_info *from;
1446     struct net *net;
1447
1448     if (!bucket || !rt6_ex)
1449         return;
1450
1451     net = dev_net(rt6_ex->rt6i->dst.dev);
1452     net->ipv6.rt6_stats->fib_rt_cache--;
1453
1454     /* purge completely the exception to allow releasing the held resources:
1455      * some [sk] cache may keep the dst around for unlimited time
1456      */
1457     from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1458     fib6_info_release(from);
1459     dst_dev_put(&rt6_ex->rt6i->dst);
1460
1461     hlist_del_rcu(&rt6_ex->hlist);
1462     dst_release(&rt6_ex->rt6i->dst);
1463     kfree_rcu(rt6_ex, rcu);
1464     WARN_ON_ONCE(!bucket->depth);
1465     bucket->depth--;
1466 }
1467
1468 /* Remove oldest rt6_ex in bucket and free the memory
1469  * Caller must hold rt6_exception_lock
1470  */
1471 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1472 {
1473     struct rt6_exception *rt6_ex, *oldest = NULL;
1474
1475     if (!bucket)
1476         return;
1477
1478     hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1479         if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1480             oldest = rt6_ex;
1481     }
1482     rt6_remove_exception(bucket, oldest);
1483 }
1484
1485 static u32 rt6_exception_hash(const struct in6_addr *dst,
1486                   const struct in6_addr *src)
1487 {
1488     static siphash_aligned_key_t rt6_exception_key;
1489     struct {
1490         struct in6_addr dst;
1491         struct in6_addr src;
1492     } __aligned(SIPHASH_ALIGNMENT) combined = {
1493         .dst = *dst,
1494     };
1495     u64 val;
1496
1497     net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1498
1499 #ifdef CONFIG_IPV6_SUBTREES
1500     if (src)
1501         combined.src = *src;
1502 #endif
1503     val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1504
1505     return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1506 }
1507
1508 /* Helper function to find the cached rt in the hash table
1509  * and update bucket pointer to point to the bucket for this
1510  * (daddr, saddr) pair
1511  * Caller must hold rt6_exception_lock
1512  */
1513 static struct rt6_exception *
1514 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1515                   const struct in6_addr *daddr,
1516                   const struct in6_addr *saddr)
1517 {
1518     struct rt6_exception *rt6_ex;
1519     u32 hval;
1520
1521     if (!(*bucket) || !daddr)
1522         return NULL;
1523
1524     hval = rt6_exception_hash(daddr, saddr);
1525     *bucket += hval;
1526
1527     hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1528         struct rt6_info *rt6 = rt6_ex->rt6i;
1529         bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1530
1531 #ifdef CONFIG_IPV6_SUBTREES
1532         if (matched && saddr)
1533             matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1534 #endif
1535         if (matched)
1536             return rt6_ex;
1537     }
1538     return NULL;
1539 }
1540
1541 /* Helper function to find the cached rt in the hash table
1542  * and update bucket pointer to point to the bucket for this
1543  * (daddr, saddr) pair
1544  * Caller must hold rcu_read_lock()
1545  */
1546 static struct rt6_exception *
1547 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1548              const struct in6_addr *daddr,
1549              const struct in6_addr *saddr)
1550 {
1551     struct rt6_exception *rt6_ex;
1552     u32 hval;
1553
1554     WARN_ON_ONCE(!rcu_read_lock_held());
1555
1556     if (!(*bucket) || !daddr)
1557         return NULL;
1558
1559     hval = rt6_exception_hash(daddr, saddr);
1560     *bucket += hval;
1561
1562     hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1563         struct rt6_info *rt6 = rt6_ex->rt6i;
1564         bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1565
1566 #ifdef CONFIG_IPV6_SUBTREES
1567         if (matched && saddr)
1568             matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1569 #endif
1570         if (matched)
1571             return rt6_ex;
1572     }
1573     return NULL;
1574 }
1575
1576 static unsigned int fib6_mtu(const struct fib6_result *res)
1577 {
1578     const struct fib6_nh *nh = res->nh;
1579     unsigned int mtu;
1580
1581     if (res->f6i->fib6_pmtu) {
1582         mtu = res->f6i->fib6_pmtu;
1583     } else {
1584         struct net_device *dev = nh->fib_nh_dev;
1585         struct inet6_dev *idev;
1586
1587         rcu_read_lock();
1588         idev = __in6_dev_get(dev);
1589         mtu = idev->cnf.mtu6;
1590         rcu_read_unlock();
1591     }
1592
1593     mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1594
1595     return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1596 }
1597
1598 #define FIB6_EXCEPTION_BUCKET_FLUSHED  0x1UL
1599
1600 /* used when the flushed bit is not relevant, only access to the bucket
1601  * (ie., all bucket users except rt6_insert_exception);
1602  *
1603  * called under rcu lock; sometimes called with rt6_exception_lock held
1604  */
1605 static
1606 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1607                                spinlock_t *lock)
1608 {
1609     struct rt6_exception_bucket *bucket;
1610
1611     if (lock)
1612         bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1613                            lockdep_is_held(lock));
1614     else
1615         bucket = rcu_dereference(nh->rt6i_exception_bucket);
1616
1617     /* remove bucket flushed bit if set */
1618     if (bucket) {
1619         unsigned long p = (unsigned long)bucket;
1620
1621         p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1622         bucket = (struct rt6_exception_bucket *)p;
1623     }
1624
1625     return bucket;
1626 }
1627
1628 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1629 {
1630     unsigned long p = (unsigned long)bucket;
1631
1632     return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1633 }
1634
1635 /* called with rt6_exception_lock held */
1636 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1637                           spinlock_t *lock)
1638 {
1639     struct rt6_exception_bucket *bucket;
1640     unsigned long p;
1641
1642     bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1643                        lockdep_is_held(lock));
1644
1645     p = (unsigned long)bucket;
1646     p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1647     bucket = (struct rt6_exception_bucket *)p;
1648     rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1649 }
1650
1651 static int rt6_insert_exception(struct rt6_info *nrt,
1652                 const struct fib6_result *res)
1653 {
1654     struct net *net = dev_net(nrt->dst.dev);
1655     struct rt6_exception_bucket *bucket;
1656     struct fib6_info *f6i = res->f6i;
1657     struct in6_addr *src_key = NULL;
1658     struct rt6_exception *rt6_ex;
1659     struct fib6_nh *nh = res->nh;
1660     int max_depth;
1661     int err = 0;
1662
1663     spin_lock_bh(&rt6_exception_lock);
1664
1665     bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1666                       lockdep_is_held(&rt6_exception_lock));
1667     if (!bucket) {
1668         bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1669                  GFP_ATOMIC);
1670         if (!bucket) {
1671             err = -ENOMEM;
1672             goto out;
1673         }
1674         rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1675     } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1676         err = -EINVAL;
1677         goto out;
1678     }
1679
1680 #ifdef CONFIG_IPV6_SUBTREES
1681     /* fib6_src.plen != 0 indicates f6i is in subtree
1682      * and exception table is indexed by a hash of
1683      * both fib6_dst and fib6_src.
1684      * Otherwise, the exception table is indexed by
1685      * a hash of only fib6_dst.
1686      */
1687     if (f6i->fib6_src.plen)
1688         src_key = &nrt->rt6i_src.addr;
1689 #endif
1690     /* rt6_mtu_change() might lower mtu on f6i.
1691      * Only insert this exception route if its mtu
1692      * is less than f6i's mtu value.
1693      */
1694     if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1695         err = -EINVAL;
1696         goto out;
1697     }
1698
1699     rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1700                            src_key);
1701     if (rt6_ex)
1702         rt6_remove_exception(bucket, rt6_ex);
1703
1704     rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1705     if (!rt6_ex) {
1706         err = -ENOMEM;
1707         goto out;
1708     }
1709     rt6_ex->rt6i = nrt;
1710     rt6_ex->stamp = jiffies;
1711     hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1712     bucket->depth++;
1713     net->ipv6.rt6_stats->fib_rt_cache++;
1714
1715     /* Randomize max depth to avoid some side channels attacks. */
1716     max_depth = FIB6_MAX_DEPTH + prandom_u32_max(FIB6_MAX_DEPTH);
1717     while (bucket->depth > max_depth)
1718         rt6_exception_remove_oldest(bucket);
1719
1720 out:
1721     spin_unlock_bh(&rt6_exception_lock);
1722
1723     /* Update fn->fn_sernum to invalidate all cached dst */
1724     if (!err) {
1725         spin_lock_bh(&f6i->fib6_table->tb6_lock);
1726         fib6_update_sernum(net, f6i);
1727         spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1728         fib6_force_start_gc(net);
1729     }
1730
1731     return err;
1732 }
1733
1734 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1735 {
1736     struct rt6_exception_bucket *bucket;
1737     struct rt6_exception *rt6_ex;
1738     struct hlist_node *tmp;
1739     int i;
1740
1741     spin_lock_bh(&rt6_exception_lock);
1742
1743     bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1744     if (!bucket)
1745         goto out;
1746
1747     /* Prevent rt6_insert_exception() to recreate the bucket list */
1748     if (!from)
1749         fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1750
1751     for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1752         hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1753             if (!from ||
1754                 rcu_access_pointer(rt6_ex->rt6i->from) == from)
1755                 rt6_remove_exception(bucket, rt6_ex);
1756         }
1757         WARN_ON_ONCE(!from && bucket->depth);
1758         bucket++;
1759     }
1760 out:
1761     spin_unlock_bh(&rt6_exception_lock);
1762 }
1763
1764 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1765 {
1766     struct fib6_info *f6i = arg;
1767
1768     fib6_nh_flush_exceptions(nh, f6i);
1769
1770     return 0;
1771 }
1772
1773 void rt6_flush_exceptions(struct fib6_info *f6i)
1774 {
1775     if (f6i->nh)
1776         nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1777                      f6i);
1778     else
1779         fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1780 }
1781
1782 /* Find cached rt in the hash table inside passed in rt
1783  * Caller has to hold rcu_read_lock()
1784  */
1785 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1786                        const struct in6_addr *daddr,
1787                        const struct in6_addr *saddr)
1788 {
1789     const struct in6_addr *src_key = NULL;
1790     struct rt6_exception_bucket *bucket;
1791     struct rt6_exception *rt6_ex;
1792     struct rt6_info *ret = NULL;
1793
1794 #ifdef CONFIG_IPV6_SUBTREES
1795     /* fib6i_src.plen != 0 indicates f6i is in subtree
1796      * and exception table is indexed by a hash of
1797      * both fib6_dst and fib6_src.
1798      * However, the src addr used to create the hash
1799      * might not be exactly the passed in saddr which
1800      * is a /128 addr from the flow.
1801      * So we need to use f6i->fib6_src to redo lookup
1802      * if the passed in saddr does not find anything.
1803      * (See the logic in ip6_rt_cache_alloc() on how
1804      * rt->rt6i_src is updated.)
1805      */
1806     if (res->f6i->fib6_src.plen)
1807         src_key = saddr;
1808 find_ex:
1809 #endif
1810     bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1811     rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1812
1813     if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1814         ret = rt6_ex->rt6i;
1815
1816 #ifdef CONFIG_IPV6_SUBTREES
1817     /* Use fib6_src as src_key and redo lookup */
1818     if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1819         src_key = &res->f6i->fib6_src.addr;
1820         goto find_ex;
1821     }
1822 #endif
1823
1824     return ret;
1825 }
1826
1827 /* Remove the passed in cached rt from the hash table that contains it */
1828 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1829                     const struct rt6_info *rt)
1830 {
1831     const struct in6_addr *src_key = NULL;
1832     struct rt6_exception_bucket *bucket;
1833     struct rt6_exception *rt6_ex;
1834     int err;
1835
1836     if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1837         return -ENOENT;
1838
1839     spin_lock_bh(&rt6_exception_lock);
1840     bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1841
1842 #ifdef CONFIG_IPV6_SUBTREES
1843     /* rt6i_src.plen != 0 indicates 'from' is in subtree
1844      * and exception table is indexed by a hash of
1845      * both rt6i_dst and rt6i_src.
1846      * Otherwise, the exception table is indexed by
1847      * a hash of only rt6i_dst.
1848      */
1849     if (plen)
1850         src_key = &rt->rt6i_src.addr;
1851 #endif
1852     rt6_ex = __rt6_find_exception_spinlock(&bucket,
1853                            &rt->rt6i_dst.addr,
1854                            src_key);
1855     if (rt6_ex) {
1856         rt6_remove_exception(bucket, rt6_ex);
1857         err = 0;
1858     } else {
1859         err = -ENOENT;
1860     }
1861
1862     spin_unlock_bh(&rt6_exception_lock);
1863     return err;
1864 }
1865
1866 struct fib6_nh_excptn_arg {
1867     struct rt6_info *rt;
1868     int     plen;
1869 };
1870
1871 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1872 {
1873     struct fib6_nh_excptn_arg *arg = _arg;
1874     int err;
1875
1876     err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1877     if (err == 0)
1878         return 1;
1879
1880     return 0;
1881 }
1882
1883 static int rt6_remove_exception_rt(struct rt6_info *rt)
1884 {
1885     struct fib6_info *from;
1886
1887     from = rcu_dereference(rt->from);
1888     if (!from || !(rt->rt6i_flags & RTF_CACHE))
1889         return -EINVAL;
1890
1891     if (from->nh) {
1892         struct fib6_nh_excptn_arg arg = {
1893             .rt = rt,
1894             .plen = from->fib6_src.plen
1895         };
1896         int rc;
1897
1898         /* rc = 1 means an entry was found */
1899         rc = nexthop_for_each_fib6_nh(from->nh,
1900                           rt6_nh_remove_exception_rt,
1901                           &arg);
1902         return rc ? 0 : -ENOENT;
1903     }
1904
1905     return fib6_nh_remove_exception(from->fib6_nh,
1906                     from->fib6_src.plen, rt);
1907 }
1908
1909 /* Find rt6_ex which contains the passed in rt cache and
1910  * refresh its stamp
1911  */
1912 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1913                      const struct rt6_info *rt)
1914 {
1915     const struct in6_addr *src_key = NULL;
1916     struct rt6_exception_bucket *bucket;
1917     struct rt6_exception *rt6_ex;
1918
1919     bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1920 #ifdef CONFIG_IPV6_SUBTREES
1921     /* rt6i_src.plen != 0 indicates 'from' is in subtree
1922      * and exception table is indexed by a hash of
1923      * both rt6i_dst and rt6i_src.
1924      * Otherwise, the exception table is indexed by
1925      * a hash of only rt6i_dst.
1926      */
1927     if (plen)
1928         src_key = &rt->rt6i_src.addr;
1929 #endif
1930     rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1931     if (rt6_ex)
1932         rt6_ex->stamp = jiffies;
1933 }
1934
1935 struct fib6_nh_match_arg {
1936     const struct net_device *dev;
1937     const struct in6_addr   *gw;
1938     struct fib6_nh      *match;
1939 };
1940
1941 /* determine if fib6_nh has given device and gateway */
1942 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1943 {
1944     struct fib6_nh_match_arg *arg = _arg;
1945
1946     if (arg->dev != nh->fib_nh_dev ||
1947         (arg->gw && !nh->fib_nh_gw_family) ||
1948         (!arg->gw && nh->fib_nh_gw_family) ||
1949         (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1950         return 0;
1951
1952     arg->match = nh;
1953
1954     /* found a match, break the loop */
1955     return 1;
1956 }
1957
1958 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1959 {
1960     struct fib6_info *from;
1961     struct fib6_nh *fib6_nh;
1962
1963     rcu_read_lock();
1964
1965     from = rcu_dereference(rt->from);
1966     if (!from || !(rt->rt6i_flags & RTF_CACHE))
1967         goto unlock;
1968
1969     if (from->nh) {
1970         struct fib6_nh_match_arg arg = {
1971             .dev = rt->dst.dev,
1972             .gw = &rt->rt6i_gateway,
1973         };
1974
1975         nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1976
1977         if (!arg.match)
1978             goto unlock;
1979         fib6_nh = arg.match;
1980     } else {
1981         fib6_nh = from->fib6_nh;
1982     }
1983     fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1984 unlock:
1985     rcu_read_unlock();
1986 }
1987
1988 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1989                      struct rt6_info *rt, int mtu)
1990 {
1991     /* If the new MTU is lower than the route PMTU, this new MTU will be the
1992      * lowest MTU in the path: always allow updating the route PMTU to
1993      * reflect PMTU decreases.
1994      *
1995      * If the new MTU is higher, and the route PMTU is equal to the local
1996      * MTU, this means the old MTU is the lowest in the path, so allow
1997      * updating it: if other nodes now have lower MTUs, PMTU discovery will
1998      * handle this.
1999      */
2000
2001     if (dst_mtu(&rt->dst) >= mtu)
2002         return true;
2003
2004     if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2005         return true;
2006
2007     return false;
2008 }
2009
2010 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2011                        const struct fib6_nh *nh, int mtu)
2012 {
2013     struct rt6_exception_bucket *bucket;
2014     struct rt6_exception *rt6_ex;
2015     int i;
2016
2017     bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2018     if (!bucket)
2019         return;
2020
2021     for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2022         hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2023             struct rt6_info *entry = rt6_ex->rt6i;
2024
2025             /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2026              * route), the metrics of its rt->from have already
2027              * been updated.
2028              */
2029             if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2030                 rt6_mtu_change_route_allowed(idev, entry, mtu))
2031                 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2032         }
2033         bucket++;
2034     }
2035 }
2036
2037 #define RTF_CACHE_GATEWAY   (RTF_GATEWAY | RTF_CACHE)
2038
2039 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2040                         const struct in6_addr *gateway)
2041 {
2042     struct rt6_exception_bucket *bucket;
2043     struct rt6_exception *rt6_ex;
2044     struct hlist_node *tmp;
2045     int i;
2046
2047     if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2048         return;
2049
2050     spin_lock_bh(&rt6_exception_lock);
2051     bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2052     if (bucket) {
2053         for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2054             hlist_for_each_entry_safe(rt6_ex, tmp,
2055                           &bucket->chain, hlist) {
2056                 struct rt6_info *entry = rt6_ex->rt6i;
2057
2058                 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2059                     RTF_CACHE_GATEWAY &&
2060                     ipv6_addr_equal(gateway,
2061                             &entry->rt6i_gateway)) {
2062                     rt6_remove_exception(bucket, rt6_ex);
2063                 }
2064             }
2065             bucket++;
2066         }
2067     }
2068
2069     spin_unlock_bh(&rt6_exception_lock);
2070 }
2071
2072 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2073                       struct rt6_exception *rt6_ex,
2074                       struct fib6_gc_args *gc_args,
2075                       unsigned long now)
2076 {
2077     struct rt6_info *rt = rt6_ex->rt6i;
2078
2079     /* we are pruning and obsoleting aged-out and non gateway exceptions
2080      * even if others have still references to them, so that on next
2081      * dst_check() such references can be dropped.
2082      * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2083      * expired, independently from their aging, as per RFC 8201 section 4
2084      */
2085     if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2086         if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2087             RT6_TRACE("aging clone %p\n", rt);
2088             rt6_remove_exception(bucket, rt6_ex);
2089             return;
2090         }
2091     } else if (time_after(jiffies, rt->dst.expires)) {
2092         RT6_TRACE("purging expired route %p\n", rt);
2093         rt6_remove_exception(bucket, rt6_ex);
2094         return;
2095     }
2096
2097     if (rt->rt6i_flags & RTF_GATEWAY) {
2098         struct neighbour *neigh;
2099
2100         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2101
2102         if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2103             RT6_TRACE("purging route %p via non-router but gateway\n",
2104                   rt);
2105             rt6_remove_exception(bucket, rt6_ex);
2106             return;
2107         }
2108     }
2109
2110     gc_args->more++;
2111 }
2112
2113 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2114                    struct fib6_gc_args *gc_args,
2115                    unsigned long now)
2116 {
2117     struct rt6_exception_bucket *bucket;
2118     struct rt6_exception *rt6_ex;
2119     struct hlist_node *tmp;
2120     int i;
2121
2122     if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2123         return;
2124
2125     rcu_read_lock_bh();
2126     spin_lock(&rt6_exception_lock);
2127     bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2128     if (bucket) {
2129         for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2130             hlist_for_each_entry_safe(rt6_ex, tmp,
2131                           &bucket->chain, hlist) {
2132                 rt6_age_examine_exception(bucket, rt6_ex,
2133                               gc_args, now);
2134             }
2135             bucket++;
2136         }
2137     }
2138     spin_unlock(&rt6_exception_lock);
2139     rcu_read_unlock_bh();
2140 }
2141
2142 struct fib6_nh_age_excptn_arg {
2143     struct fib6_gc_args *gc_args;
2144     unsigned long       now;
2145 };
2146
2147 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2148 {
2149     struct fib6_nh_age_excptn_arg *arg = _arg;
2150
2151     fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2152     return 0;
2153 }
2154
2155 void rt6_age_exceptions(struct fib6_info *f6i,
2156             struct fib6_gc_args *gc_args,
2157             unsigned long now)
2158 {
2159     if (f6i->nh) {
2160         struct fib6_nh_age_excptn_arg arg = {
2161             .gc_args = gc_args,
2162             .now = now
2163         };
2164
2165         nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2166                      &arg);
2167     } else {
2168         fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2169     }
2170 }
2171
2172 /* must be called with rcu lock held */
2173 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2174               struct flowi6 *fl6, struct fib6_result *res, int strict)
2175 {
2176     struct fib6_node *fn, *saved_fn;
2177
2178     fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2179     saved_fn = fn;
2180
2181 redo_rt6_select:
2182     rt6_select(net, fn, oif, res, strict);
2183     if (res->f6i == net->ipv6.fib6_null_entry) {
2184         fn = fib6_backtrack(fn, &fl6->saddr);
2185         if (fn)
2186             goto redo_rt6_select;
2187         else if (strict & RT6_LOOKUP_F_REACHABLE) {
2188             /* also consider unreachable route */
2189             strict &= ~RT6_LOOKUP_F_REACHABLE;
2190             fn = saved_fn;
2191             goto redo_rt6_select;
2192         }
2193     }
2194
2195     trace_fib6_table_lookup(net, res, table, fl6);
2196
2197     return 0;
2198 }
2199
2200 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2201                    int oif, struct flowi6 *fl6,
2202                    const struct sk_buff *skb, int flags)
2203 {
2204     struct fib6_result res = {};
2205     struct rt6_info *rt = NULL;
2206     int strict = 0;
2207
2208     WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2209              !rcu_read_lock_held());
2210
2211     strict |= flags & RT6_LOOKUP_F_IFACE;
2212     strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2213     if (net->ipv6.devconf_all->forwarding == 0)
2214         strict |= RT6_LOOKUP_F_REACHABLE;
2215
2216     rcu_read_lock();
2217
2218     fib6_table_lookup(net, table, oif, fl6, &res, strict);
2219     if (res.f6i == net->ipv6.fib6_null_entry)
2220         goto out;
2221
2222     fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2223
2224     /*Search through exception table */
2225     rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2226     if (rt) {
2227         goto out;
2228     } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2229                 !res.nh->fib_nh_gw_family)) {
2230         /* Create a RTF_CACHE clone which will not be
2231          * owned by the fib6 tree.  It is for the special case where
2232          * the daddr in the skb during the neighbor look-up is different
2233          * from the fl6->daddr used to look-up route here.
2234          */
2235         rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2236
2237         if (rt) {
2238             /* 1 refcnt is taken during ip6_rt_cache_alloc().
2239              * As rt6_uncached_list_add() does not consume refcnt,
2240              * this refcnt is always returned to the caller even
2241              * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2242              */
2243             rt6_uncached_list_add(rt);
2244             rcu_read_unlock();
2245
2246             return rt;
2247         }
2248     } else {
2249         /* Get a percpu copy */
2250         local_bh_disable();
2251         rt = rt6_get_pcpu_route(&res);
2252
2253         if (!rt)
2254             rt = rt6_make_pcpu_route(net, &res);
2255
2256         local_bh_enable();
2257     }
2258 out:
2259     if (!rt)
2260         rt = net->ipv6.ip6_null_entry;
2261     if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2262         ip6_hold_safe(net, &rt);
2263     rcu_read_unlock();
2264
2265     return rt;
2266 }
2267 EXPORT_SYMBOL_GPL(ip6_pol_route);
2268
2269 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2270                         struct fib6_table *table,
2271                         struct flowi6 *fl6,
2272                         const struct sk_buff *skb,
2273                         int flags)
2274 {
2275     return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2276 }
2277
2278 struct dst_entry *ip6_route_input_lookup(struct net *net,
2279                      struct net_device *dev,
2280                      struct flowi6 *fl6,
2281                      const struct sk_buff *skb,
2282                      int flags)
2283 {
2284     if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2285         flags |= RT6_LOOKUP_F_IFACE;
2286
2287     return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2288 }
2289 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2290
2291 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2292                   struct flow_keys *keys,
2293                   struct flow_keys *flkeys)
2294 {
2295     const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2296     const struct ipv6hdr *key_iph = outer_iph;
2297     struct flow_keys *_flkeys = flkeys;
2298     const struct ipv6hdr *inner_iph;
2299     const struct icmp6hdr *icmph;
2300     struct ipv6hdr _inner_iph;
2301     struct icmp6hdr _icmph;
2302
2303     if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2304         goto out;
2305
2306     icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2307                    sizeof(_icmph), &_icmph);
2308     if (!icmph)
2309         goto out;
2310
2311     if (!icmpv6_is_err(icmph->icmp6_type))
2312         goto out;
2313
2314     inner_iph = skb_header_pointer(skb,
2315                        skb_transport_offset(skb) + sizeof(*icmph),
2316                        sizeof(_inner_iph), &_inner_iph);
2317     if (!inner_iph)
2318         goto out;
2319
2320     key_iph = inner_iph;
2321     _flkeys = NULL;
2322 out:
2323     if (_flkeys) {
2324         keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2325         keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2326         keys->tags.flow_label = _flkeys->tags.flow_label;
2327         keys->basic.ip_proto = _flkeys->basic.ip_proto;
2328     } else {
2329         keys->addrs.v6addrs.src = key_iph->saddr;
2330         keys->addrs.v6addrs.dst = key_iph->daddr;
2331         keys->tags.flow_label = ip6_flowlabel(key_iph);
2332         keys->basic.ip_proto = key_iph->nexthdr;
2333     }
2334 }
2335
2336 static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2337                        const struct sk_buff *skb,
2338                        bool *p_has_inner)
2339 {
2340     u32 hash_fields = ip6_multipath_hash_fields(net);
2341     struct flow_keys keys, hash_keys;
2342
2343     if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2344         return 0;
2345
2346     memset(&hash_keys, 0, sizeof(hash_keys));
2347     skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2348
2349     hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2350     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2351         hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2352     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2353         hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2354     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2355         hash_keys.basic.ip_proto = keys.basic.ip_proto;
2356     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2357         hash_keys.tags.flow_label = keys.tags.flow_label;
2358     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2359         hash_keys.ports.src = keys.ports.src;
2360     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2361         hash_keys.ports.dst = keys.ports.dst;
2362
2363     *p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2364     return flow_hash_from_keys(&hash_keys);
2365 }
2366
2367 static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2368                        const struct sk_buff *skb,
2369                        bool has_inner)
2370 {
2371     u32 hash_fields = ip6_multipath_hash_fields(net);
2372     struct flow_keys keys, hash_keys;
2373
2374     /* We assume the packet carries an encapsulation, but if none was
2375      * encountered during dissection of the outer flow, then there is no
2376      * point in calling the flow dissector again.
2377      */
2378     if (!has_inner)
2379         return 0;
2380
2381     if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2382         return 0;
2383
2384     memset(&hash_keys, 0, sizeof(hash_keys));
2385     skb_flow_dissect_flow_keys(skb, &keys, 0);
2386
2387     if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2388         return 0;
2389
2390     if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2391         hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2392         if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2393             hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2394         if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2395             hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2396     } else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2397         hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2398         if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2399             hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2400         if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2401             hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2402         if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2403             hash_keys.tags.flow_label = keys.tags.flow_label;
2404     }
2405
2406     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2407         hash_keys.basic.ip_proto = keys.basic.ip_proto;
2408     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2409         hash_keys.ports.src = keys.ports.src;
2410     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2411         hash_keys.ports.dst = keys.ports.dst;
2412
2413     return flow_hash_from_keys(&hash_keys);
2414 }
2415
2416 static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2417                      const struct sk_buff *skb)
2418 {
2419     u32 mhash, mhash_inner;
2420     bool has_inner = true;
2421
2422     mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2423     mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2424
2425     return jhash_2words(mhash, mhash_inner, 0);
2426 }
2427
2428 static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2429                      const struct flowi6 *fl6)
2430 {
2431     u32 hash_fields = ip6_multipath_hash_fields(net);
2432     struct flow_keys hash_keys;
2433
2434     if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2435         return 0;
2436
2437     memset(&hash_keys, 0, sizeof(hash_keys));
2438     hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2439     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2440         hash_keys.addrs.v6addrs.src = fl6->saddr;
2441     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2442         hash_keys.addrs.v6addrs.dst = fl6->daddr;
2443     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2444         hash_keys.basic.ip_proto = fl6->flowi6_proto;
2445     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2446         hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2447     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2448         hash_keys.ports.src = fl6->fl6_sport;
2449     if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2450         hash_keys.ports.dst = fl6->fl6_dport;
2451
2452     return flow_hash_from_keys(&hash_keys);
2453 }
2454
2455 /* if skb is set it will be used and fl6 can be NULL */
2456 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2457                const struct sk_buff *skb, struct flow_keys *flkeys)
2458 {
2459     struct flow_keys hash_keys;
2460     u32 mhash = 0;
2461
2462     switch (ip6_multipath_hash_policy(net)) {
2463     case 0:
2464         memset(&hash_keys, 0, sizeof(hash_keys));
2465         hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2466         if (skb) {
2467             ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2468         } else {
2469             hash_keys.addrs.v6addrs.src = fl6->saddr;
2470             hash_keys.addrs.v6addrs.dst = fl6->daddr;
2471             hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2472             hash_keys.basic.ip_proto = fl6->flowi6_proto;
2473         }
2474         mhash = flow_hash_from_keys(&hash_keys);
2475         break;
2476     case 1:
2477         if (skb) {
2478             unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2479             struct flow_keys keys;
2480
2481             /* short-circuit if we already have L4 hash present */
2482             if (skb->l4_hash)
2483                 return skb_get_hash_raw(skb) >> 1;
2484
2485             memset(&hash_keys, 0, sizeof(hash_keys));
2486
2487             if (!flkeys) {
2488                 skb_flow_dissect_flow_keys(skb, &keys, flag);
2489                 flkeys = &keys;
2490             }
2491             hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2492             hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2493             hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2494             hash_keys.ports.src = flkeys->ports.src;
2495             hash_keys.ports.dst = flkeys->ports.dst;
2496             hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2497         } else {
2498             memset(&hash_keys, 0, sizeof(hash_keys));
2499             hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2500             hash_keys.addrs.v6addrs.src = fl6->saddr;
2501             hash_keys.addrs.v6addrs.dst = fl6->daddr;
2502             hash_keys.ports.src = fl6->fl6_sport;
2503             hash_keys.ports.dst = fl6->fl6_dport;
2504             hash_keys.basic.ip_proto = fl6->flowi6_proto;
2505         }
2506         mhash = flow_hash_from_keys(&hash_keys);
2507         break;
2508     case 2:
2509         memset(&hash_keys, 0, sizeof(hash_keys));
2510         hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2511         if (skb) {
2512             struct flow_keys keys;
2513
2514             if (!flkeys) {
2515                 skb_flow_dissect_flow_keys(skb, &keys, 0);
2516                 flkeys = &keys;
2517             }
2518
2519             /* Inner can be v4 or v6 */
2520             if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2521                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2522                 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2523                 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2524             } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2525                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2526                 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2527                 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2528                 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2529                 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2530             } else {
2531                 /* Same as case 0 */
2532                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2533                 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2534             }
2535         } else {
2536             /* Same as case 0 */
2537             hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2538             hash_keys.addrs.v6addrs.src = fl6->saddr;
2539             hash_keys.addrs.v6addrs.dst = fl6->daddr;
2540             hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2541             hash_keys.basic.ip_proto = fl6->flowi6_proto;
2542         }
2543         mhash = flow_hash_from_keys(&hash_keys);
2544         break;
2545     case 3:
2546         if (skb)
2547             mhash = rt6_multipath_custom_hash_skb(net, skb);
2548         else
2549             mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2550         break;
2551     }
2552
2553     return mhash >> 1;
2554 }
2555
2556 /* Called with rcu held */
2557 void ip6_route_input(struct sk_buff *skb)
2558 {
2559     const struct ipv6hdr *iph = ipv6_hdr(skb);
2560     struct net *net = dev_net(skb->dev);
2561     int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2562     struct ip_tunnel_info *tun_info;
2563     struct flowi6 fl6 = {
2564         .flowi6_iif = skb->dev->ifindex,
2565         .daddr = iph->daddr,
2566         .saddr = iph->saddr,
2567         .flowlabel = ip6_flowinfo(iph),
2568         .flowi6_mark = skb->mark,
2569         .flowi6_proto = iph->nexthdr,
2570     };
2571     struct flow_keys *flkeys = NULL, _flkeys;
2572
2573     tun_info = skb_tunnel_info(skb);
2574     if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2575         fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2576
2577     if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2578         flkeys = &_flkeys;
2579
2580     if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2581         fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2582     skb_dst_drop(skb);
2583     skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2584                               &fl6, skb, flags));
2585 }
2586
2587 INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2588                          struct fib6_table *table,
2589                          struct flowi6 *fl6,
2590                          const struct sk_buff *skb,
2591                          int flags)
2592 {
2593     return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2594 }
2595
2596 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2597                            const struct sock *sk,
2598                            struct flowi6 *fl6, int flags)
2599 {
2600     bool any_src;
2601
2602     if (ipv6_addr_type(&fl6->daddr) &
2603         (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2604         struct dst_entry *dst;
2605
2606         /* This function does not take refcnt on the dst */
2607         dst = l3mdev_link_scope_lookup(net, fl6);
2608         if (dst)
2609             return dst;
2610     }
2611
2612     fl6->flowi6_iif = LOOPBACK_IFINDEX;
2613
2614     flags |= RT6_LOOKUP_F_DST_NOREF;
2615     any_src = ipv6_addr_any(&fl6->saddr);
2616     if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2617         (fl6->flowi6_oif && any_src))
2618         flags |= RT6_LOOKUP_F_IFACE;
2619
2620     if (!any_src)
2621         flags |= RT6_LOOKUP_F_HAS_SADDR;
2622     else if (sk)
2623         flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2624
2625     return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2626 }
2627 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2628
2629 struct dst_entry *ip6_route_output_flags(struct net *net,
2630                      const struct sock *sk,
2631                      struct flowi6 *fl6,
2632                      int flags)
2633 {
2634     struct dst_entry *dst;
2635     struct rt6_info *rt6;
2636
2637     rcu_read_lock();
2638     dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2639     rt6 = (struct rt6_info *)dst;
2640     /* For dst cached in uncached_list, refcnt is already taken. */
2641     if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2642         dst = &net->ipv6.ip6_null_entry->dst;
2643         dst_hold(dst);
2644     }
2645     rcu_read_unlock();
2646
2647     return dst;
2648 }
2649 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2650
2651 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2652 {
2653     struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2654     struct net_device *loopback_dev = net->loopback_dev;
2655     struct dst_entry *new = NULL;
2656
2657     rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2658                DST_OBSOLETE_DEAD, 0);
2659     if (rt) {
2660         rt6_info_init(rt);
2661         atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2662
2663         new = &rt->dst;
2664         new->__use = 1;
2665         new->input = dst_discard;
2666         new->output = dst_discard_out;
2667
2668         dst_copy_metrics(new, &ort->dst);
2669
2670         rt->rt6i_idev = in6_dev_get(loopback_dev);
2671         rt->rt6i_gateway = ort->rt6i_gateway;
2672         rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2673
2674         memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2675 #ifdef CONFIG_IPV6_SUBTREES
2676         memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2677 #endif
2678     }
2679
2680     dst_release(dst_orig);
2681     return new ? new : ERR_PTR(-ENOMEM);
2682 }
2683
2684 /*
2685  *  Destination cache support functions
2686  */
2687
2688 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2689 {
2690     u32 rt_cookie = 0;
2691
2692     if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2693         return false;
2694
2695     if (fib6_check_expired(f6i))
2696         return false;
2697
2698     return true;
2699 }
2700
2701 static struct dst_entry *rt6_check(struct rt6_info *rt,
2702                    struct fib6_info *from,
2703                    u32 cookie)
2704 {
2705     u32 rt_cookie = 0;
2706
2707     if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2708         rt_cookie != cookie)
2709         return NULL;
2710
2711     if (rt6_check_expired(rt))
2712         return NULL;
2713
2714     return &rt->dst;
2715 }
2716
2717 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2718                         struct fib6_info *from,
2719                         u32 cookie)
2720 {
2721     if (!__rt6_check_expired(rt) &&
2722         rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2723         fib6_check(from, cookie))
2724         return &rt->dst;
2725     else
2726         return NULL;
2727 }
2728
2729 INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2730                             u32 cookie)
2731 {
2732     struct dst_entry *dst_ret;
2733     struct fib6_info *from;
2734     struct rt6_info *rt;
2735
2736     rt = container_of(dst, struct rt6_info, dst);
2737
2738     if (rt->sernum)
2739         return rt6_is_valid(rt) ? dst : NULL;
2740
2741     rcu_read_lock();
2742
2743     /* All IPV6 dsts are created with ->obsolete set to the value
2744      * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2745      * into this function always.
2746      */
2747
2748     from = rcu_dereference(rt->from);
2749
2750     if (from && (rt->rt6i_flags & RTF_PCPU ||
2751         unlikely(!list_empty(&rt->rt6i_uncached))))
2752         dst_ret = rt6_dst_from_check(rt, from, cookie);
2753     else
2754         dst_ret = rt6_check(rt, from, cookie);
2755
2756     rcu_read_unlock();
2757
2758     return dst_ret;
2759 }
2760 EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2761
2762 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2763 {
2764     struct rt6_info *rt = (struct rt6_info *) dst;
2765
2766     if (rt) {
2767         if (rt->rt6i_flags & RTF_CACHE) {
2768             rcu_read_lock();
2769             if (rt6_check_expired(rt)) {
2770                 rt6_remove_exception_rt(rt);
2771                 dst = NULL;
2772             }
2773             rcu_read_unlock();
2774         } else {
2775             dst_release(dst);
2776             dst = NULL;
2777         }
2778     }
2779     return dst;
2780 }
2781
2782 static void ip6_link_failure(struct sk_buff *skb)
2783 {
2784     struct rt6_info *rt;
2785
2786     icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2787
2788     rt = (struct rt6_info *) skb_dst(skb);
2789     if (rt) {
2790         rcu_read_lock();
2791         if (rt->rt6i_flags & RTF_CACHE) {
2792             rt6_remove_exception_rt(rt);
2793         } else {
2794             struct fib6_info *from;
2795             struct fib6_node *fn;
2796
2797             from = rcu_dereference(rt->from);
2798             if (from) {
2799                 fn = rcu_dereference(from->fib6_node);
2800                 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2801                     WRITE_ONCE(fn->fn_sernum, -1);
2802             }
2803         }
2804         rcu_read_unlock();
2805     }
2806 }
2807
2808 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2809 {
2810     if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2811         struct fib6_info *from;
2812
2813         rcu_read_lock();
2814         from = rcu_dereference(rt0->from);
2815         if (from)
2816             rt0->dst.expires = from->expires;
2817         rcu_read_unlock();
2818     }
2819
2820     dst_set_expires(&rt0->dst, timeout);
2821     rt0->rt6i_flags |= RTF_EXPIRES;
2822 }
2823
2824 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2825 {
2826     struct net *net = dev_net(rt->dst.dev);
2827
2828     dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2829     rt->rt6i_flags |= RTF_MODIFIED;
2830     rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2831 }
2832
2833 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2834 {
2835     return !(rt->rt6i_flags & RTF_CACHE) &&
2836         (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2837 }
2838
2839 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2840                  const struct ipv6hdr *iph, u32 mtu,
2841                  bool confirm_neigh)
2842 {
2843     const struct in6_addr *daddr, *saddr;
2844     struct rt6_info *rt6 = (struct rt6_info *)dst;
2845
2846     /* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2847      * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2848      * [see also comment in rt6_mtu_change_route()]
2849      */
2850
2851     if (iph) {
2852         daddr = &iph->daddr;
2853         saddr = &iph->saddr;
2854     } else if (sk) {
2855         daddr = &sk->sk_v6_daddr;
2856         saddr = &inet6_sk(sk)->saddr;
2857     } else {
2858         daddr = NULL;
2859         saddr = NULL;
2860     }
2861
2862     if (confirm_neigh)
2863         dst_confirm_neigh(dst, daddr);
2864
2865     if (mtu < IPV6_MIN_MTU)
2866         return;
2867     if (mtu >= dst_mtu(dst))
2868         return;
2869
2870     if (!rt6_cache_allowed_for_pmtu(rt6)) {
2871         rt6_do_update_pmtu(rt6, mtu);
2872         /* update rt6_ex->stamp for cache */
2873         if (rt6->rt6i_flags & RTF_CACHE)
2874             rt6_update_exception_stamp_rt(rt6);
2875     } else if (daddr) {
2876         struct fib6_result res = {};
2877         struct rt6_info *nrt6;
2878
2879         rcu_read_lock();
2880         res.f6i = rcu_dereference(rt6->from);
2881         if (!res.f6i)
2882             goto out_unlock;
2883
2884         res.fib6_flags = res.f6i->fib6_flags;
2885         res.fib6_type = res.f6i->fib6_type;
2886
2887         if (res.f6i->nh) {
2888             struct fib6_nh_match_arg arg = {
2889                 .dev = dst->dev,
2890                 .gw = &rt6->rt6i_gateway,
2891             };
2892
2893             nexthop_for_each_fib6_nh(res.f6i->nh,
2894                          fib6_nh_find_match, &arg);
2895
2896             /* fib6_info uses a nexthop that does not have fib6_nh
2897              * using the dst->dev + gw. Should be impossible.
2898              */
2899             if (!arg.match)
2900                 goto out_unlock;
2901
2902             res.nh = arg.match;
2903         } else {
2904             res.nh = res.f6i->fib6_nh;
2905         }
2906
2907         nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2908         if (nrt6) {
2909             rt6_do_update_pmtu(nrt6, mtu);
2910             if (rt6_insert_exception(nrt6, &res))
2911                 dst_release_immediate(&nrt6->dst);
2912         }
2913 out_unlock:
2914         rcu_read_unlock();
2915     }
2916 }
2917
2918 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2919                    struct sk_buff *skb, u32 mtu,
2920                    bool confirm_neigh)
2921 {
2922     __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2923                  confirm_neigh);
2924 }
2925
2926 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2927              int oif, u32 mark, kuid_t uid)
2928 {
2929     const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2930     struct dst_entry *dst;
2931     struct flowi6 fl6 = {
2932         .flowi6_oif = oif,
2933         .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2934         .daddr = iph->daddr,
2935         .saddr = iph->saddr,
2936         .flowlabel = ip6_flowinfo(iph),
2937         .flowi6_uid = uid,
2938     };
2939
2940     dst = ip6_route_output(net, NULL, &fl6);
2941     if (!dst->error)
2942         __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2943     dst_release(dst);
2944 }
2945 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2946
2947 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2948 {
2949     int oif = sk->sk_bound_dev_if;
2950     struct dst_entry *dst;
2951
2952     if (!oif && skb->dev)
2953         oif = l3mdev_master_ifindex(skb->dev);
2954
2955     ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2956
2957     dst = __sk_dst_get(sk);
2958     if (!dst || !dst->obsolete ||
2959         dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2960         return;
2961
2962     bh_lock_sock(sk);
2963     if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2964         ip6_datagram_dst_update(sk, false);
2965     bh_unlock_sock(sk);
2966 }
2967 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2968
2969 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2970                const struct flowi6 *fl6)
2971 {
2972 #ifdef CONFIG_IPV6_SUBTREES
2973     struct ipv6_pinfo *np = inet6_sk(sk);
2974 #endif
2975
2976     ip6_dst_store(sk, dst,
2977               ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2978               &sk->sk_v6_daddr : NULL,
2979 #ifdef CONFIG_IPV6_SUBTREES
2980               ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2981               &np->saddr :
2982 #endif
2983               NULL);
2984 }
2985
2986 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2987                   struct flowi6 *fl6,
2988                   const struct in6_addr *gw,
2989                   struct rt6_info **ret)
2990 {
2991     const struct fib6_nh *nh = res->nh;
2992
2993     if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2994         fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2995         return false;
2996
2997     /* rt_cache's gateway might be different from its 'parent'
2998      * in the case of an ip redirect.
2999      * So we keep searching in the exception table if the gateway
3000      * is different.
3001      */
3002     if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3003         struct rt6_info *rt_cache;
3004
3005         rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3006         if (rt_cache &&
3007             ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3008             *ret = rt_cache;
3009             return true;
3010         }
3011         return false;
3012     }
3013     return true;
3014 }
3015
3016 struct fib6_nh_rd_arg {
3017     struct fib6_result  *res;
3018     struct flowi6       *fl6;
3019     const struct in6_addr   *gw;
3020     struct rt6_info     **ret;
3021 };
3022
3023 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3024 {
3025     struct fib6_nh_rd_arg *arg = _arg;
3026
3027     arg->res->nh = nh;
3028     return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3029 }
3030
3031 /* Handle redirects */
3032 struct ip6rd_flowi {
3033     struct flowi6 fl6;
3034     struct in6_addr gateway;
3035 };
3036
3037 INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3038                          struct fib6_table *table,
3039                          struct flowi6 *fl6,
3040                          const struct sk_buff *skb,
3041                          int flags)
3042 {
3043     struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3044     struct rt6_info *ret = NULL;
3045     struct fib6_result res = {};
3046     struct fib6_nh_rd_arg arg = {
3047         .res = &res,
3048         .fl6 = fl6,
3049         .gw  = &rdfl->gateway,
3050         .ret = &ret
3051     };
3052     struct fib6_info *rt;
3053     struct fib6_node *fn;
3054
3055     /* Get the "current" route for this destination and
3056      * check if the redirect has come from appropriate router.
3057      *
3058      * RFC 4861 specifies that redirects should only be
3059      * accepted if they come from the nexthop to the target.
3060      * Due to the way the routes are chosen, this notion
3061      * is a bit fuzzy and one might need to check all possible
3062      * routes.
3063      */
3064
3065     rcu_read_lock();
3066     fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3067 restart:
3068     for_each_fib6_node_rt_rcu(fn) {
3069         res.f6i = rt;
3070         if (fib6_check_expired(rt))
3071             continue;
3072         if (rt->fib6_flags & RTF_REJECT)
3073             break;
3074         if (unlikely(rt->nh)) {
3075             if (nexthop_is_blackhole(rt->nh))
3076                 continue;
3077             /* on match, res->nh is filled in and potentially ret */
3078             if (nexthop_for_each_fib6_nh(rt->nh,
3079                              fib6_nh_redirect_match,
3080                              &arg))
3081                 goto out;
3082         } else {
3083             res.nh = rt->fib6_nh;
3084             if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3085                           &ret))
3086                 goto out;
3087         }
3088     }
3089
3090     if (!rt)
3091         rt = net->ipv6.fib6_null_entry;
3092     else if (rt->fib6_flags & RTF_REJECT) {
3093         ret = net->ipv6.ip6_null_entry;
3094         goto out;
3095     }
3096
3097     if (rt == net->ipv6.fib6_null_entry) {
3098         fn = fib6_backtrack(fn, &fl6->saddr);
3099         if (fn)
3100             goto restart;
3101     }
3102
3103     res.f6i = rt;
3104     res.nh = rt->fib6_nh;
3105 out:
3106     if (ret) {
3107         ip6_hold_safe(net, &ret);
3108     } else {
3109         res.fib6_flags = res.f6i->fib6_flags;
3110         res.fib6_type = res.f6i->fib6_type;
3111         ret = ip6_create_rt_rcu(&res);
3112     }
3113
3114     rcu_read_unlock();
3115
3116     trace_fib6_table_lookup(net, &res, table, fl6);
3117     return ret;
3118 };
3119
3120 static struct dst_entry *ip6_route_redirect(struct net *net,
3121                         const struct flowi6 *fl6,
3122                         const struct sk_buff *skb,
3123                         const struct in6_addr *gateway)
3124 {
3125     int flags = RT6_LOOKUP_F_HAS_SADDR;
3126     struct ip6rd_flowi rdfl;
3127
3128     rdfl.fl6 = *fl6;
3129     rdfl.gateway = *gateway;
3130
3131     return fib6_rule_lookup(net, &rdfl.fl6, skb,
3132                 flags, __ip6_route_redirect);
3133 }
3134
3135 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3136           kuid_t uid)
3137 {
3138     const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3139     struct dst_entry *dst;
3140     struct flowi6 fl6 = {
3141         .flowi6_iif = LOOPBACK_IFINDEX,
3142         .flowi6_oif = oif,
3143         .flowi6_mark = mark,
3144         .daddr = iph->daddr,
3145         .saddr = iph->saddr,
3146         .flowlabel = ip6_flowinfo(iph),
3147         .flowi6_uid = uid,
3148     };
3149
3150     dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3151     rt6_do_redirect(dst, NULL, skb);
3152     dst_release(dst);
3153 }
3154 EXPORT_SYMBOL_GPL(ip6_redirect);
3155
3156 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3157 {
3158     const struct ipv6hdr *iph = ipv6_hdr(skb);
3159     const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3160     struct dst_entry *dst;
3161     struct flowi6 fl6 = {
3162         .flowi6_iif = LOOPBACK_IFINDEX,
3163         .flowi6_oif = oif,
3164         .daddr = msg->dest,
3165         .saddr = iph->daddr,
3166         .flowi6_uid = sock_net_uid(net, NULL),
3167     };
3168
3169     dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3170     rt6_do_redirect(dst, NULL, skb);
3171     dst_release(dst);
3172 }
3173
3174 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3175 {
3176     ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3177              sk->sk_uid);
3178 }
3179 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3180
3181 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3182 {
3183     struct net_device *dev = dst->dev;
3184     unsigned int mtu = dst_mtu(dst);
3185     struct net *net = dev_net(dev);
3186
3187     mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3188
3189     if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3190         mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3191
3192     /*
3193      * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3194      * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3195      * IPV6_MAXPLEN is also valid and means: "any MSS,
3196      * rely only on pmtu discovery"
3197      */
3198     if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3199         mtu = IPV6_MAXPLEN;
3200     return mtu;
3201 }
3202
3203 INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3204 {
3205     return ip6_dst_mtu_maybe_forward(dst, false);
3206 }
3207 EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3208
3209 /* MTU selection:
3210  * 1. mtu on route is locked - use it
3211  * 2. mtu from nexthop exception
3212  * 3. mtu from egress device
3213  *
3214  * based on ip6_dst_mtu_forward and exception logic of
3215  * rt6_find_cached_rt; called with rcu_read_lock
3216  */
3217 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3218               const struct in6_addr *daddr,
3219               const struct in6_addr *saddr)
3220 {
3221     const struct fib6_nh *nh = res->nh;
3222     struct fib6_info *f6i = res->f6i;
3223     struct inet6_dev *idev;
3224     struct rt6_info *rt;
3225     u32 mtu = 0;
3226
3227     if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3228         mtu = f6i->fib6_pmtu;
3229         if (mtu)
3230             goto out;
3231     }
3232
3233     rt = rt6_find_cached_rt(res, daddr, saddr);
3234     if (unlikely(rt)) {
3235         mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3236     } else {
3237         struct net_device *dev = nh->fib_nh_dev;
3238
3239         mtu = IPV6_MIN_MTU;
3240         idev = __in6_dev_get(dev);
3241         if (idev && idev->cnf.mtu6 > mtu)
3242             mtu = idev->cnf.mtu6;
3243     }
3244
3245     mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3246 out:
3247     return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3248 }
3249
3250 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3251                   struct flowi6 *fl6)
3252 {
3253     struct dst_entry *dst;
3254     struct rt6_info *rt;
3255     struct inet6_dev *idev = in6_dev_get(dev);
3256     struct net *net = dev_net(dev);
3257
3258     if (unlikely(!idev))
3259         return ERR_PTR(-ENODEV);
3260
3261     rt = ip6_dst_alloc(net, dev, 0);
3262     if (unlikely(!rt)) {
3263         in6_dev_put(idev);
3264         dst = ERR_PTR(-ENOMEM);
3265         goto out;
3266     }
3267
3268     rt->dst.input = ip6_input;
3269     rt->dst.output  = ip6_output;
3270     rt->rt6i_gateway  = fl6->daddr;
3271     rt->rt6i_dst.addr = fl6->daddr;
3272     rt->rt6i_dst.plen = 128;
3273     rt->rt6i_idev     = idev;
3274     dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3275
3276     /* Add this dst into uncached_list so that rt6_disable_ip() can
3277      * do proper release of the net_device
3278      */
3279     rt6_uncached_list_add(rt);
3280
3281     dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3282
3283 out:
3284     return dst;
3285 }
3286
3287 static int ip6_dst_gc(struct dst_ops *ops)
3288 {
3289     struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3290     int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3291     int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3292     int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3293     int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3294     unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3295     unsigned int val;
3296     int entries;
3297
3298     entries = dst_entries_get_fast(ops);
3299     if (entries > rt_max_size)
3300         entries = dst_entries_get_slow(ops);
3301
3302     if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3303         entries <= rt_max_size)
3304         goto out;
3305
3306     fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3307     entries = dst_entries_get_slow(ops);
3308     if (entries < ops->gc_thresh)
3309         atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3310 out:
3311     val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3312     atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3313     return entries > rt_max_size;
3314 }
3315
3316 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3317                    const struct in6_addr *gw_addr, u32 tbid,
3318                    int flags, struct fib6_result *res)
3319 {
3320     struct flowi6 fl6 = {
3321         .flowi6_oif = cfg->fc_ifindex,
3322         .daddr = *gw_addr,
3323         .saddr = cfg->fc_prefsrc,
3324     };
3325     struct fib6_table *table;
3326     int err;
3327
3328     table = fib6_get_table(net, tbid);
3329     if (!table)
3330         return -EINVAL;
3331
3332     if (!ipv6_addr_any(&cfg->fc_prefsrc))
3333         flags |= RT6_LOOKUP_F_HAS_SADDR;
3334
3335     flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3336
3337     err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3338     if (!err && res->f6i != net->ipv6.fib6_null_entry)
3339         fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3340                  cfg->fc_ifindex != 0, NULL, flags);
3341
3342     return err;
3343 }
3344
3345 static int ip6_route_check_nh_onlink(struct net *net,
3346                      struct fib6_config *cfg,
3347                      const struct net_device *dev,
3348                      struct netlink_ext_ack *extack)
3349 {
3350     u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3351     const struct in6_addr *gw_addr = &cfg->fc_gateway;
3352     struct fib6_result res = {};
3353     int err;
3354
3355     err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3356     if (!err && !(res.fib6_flags & RTF_REJECT) &&
3357         /* ignore match if it is the default route */
3358         !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3359         (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3360         NL_SET_ERR_MSG(extack,
3361                    "Nexthop has invalid gateway or device mismatch");
3362         err = -EINVAL;
3363     }
3364
3365     return err;
3366 }
3367
3368 static int ip6_route_check_nh(struct net *net,
3369                   struct fib6_config *cfg,
3370                   struct net_device **_dev,
3371                   struct inet6_dev **idev)
3372 {
3373     const struct in6_addr *gw_addr = &cfg->fc_gateway;
3374     struct net_device *dev = _dev ? *_dev : NULL;
3375     int flags = RT6_LOOKUP_F_IFACE;
3376     struct fib6_result res = {};
3377     int err = -EHOSTUNREACH;
3378
3379     if (cfg->fc_table) {
3380         err = ip6_nh_lookup_table(net, cfg, gw_addr,
3381                       cfg->fc_table, flags, &res);
3382         /* gw_addr can not require a gateway or resolve to a reject
3383          * route. If a device is given, it must match the result.
3384          */
3385         if (err || res.fib6_flags & RTF_REJECT ||
3386             res.nh->fib_nh_gw_family ||
3387             (dev && dev != res.nh->fib_nh_dev))
3388             err = -EHOSTUNREACH;
3389     }
3390
3391     if (err < 0) {
3392         struct flowi6 fl6 = {
3393             .flowi6_oif = cfg->fc_ifindex,
3394             .daddr = *gw_addr,
3395         };
3396
3397         err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3398         if (err || res.fib6_flags & RTF_REJECT ||
3399             res.nh->fib_nh_gw_family)
3400             err = -EHOSTUNREACH;
3401
3402         if (err)
3403             return err;
3404
3405         fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3406                  cfg->fc_ifindex != 0, NULL, flags);
3407     }
3408
3409     err = 0;
3410     if (dev) {
3411         if (dev != res.nh->fib_nh_dev)
3412             err = -EHOSTUNREACH;
3413     } else {
3414         *_dev = dev = res.nh->fib_nh_dev;
3415         dev_hold(dev);
3416         *idev = in6_dev_get(dev);
3417     }
3418
3419     return err;
3420 }
3421
3422 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3423                struct net_device **_dev, struct inet6_dev **idev,
3424                struct netlink_ext_ack *extack)
3425 {
3426     const struct in6_addr *gw_addr = &cfg->fc_gateway;
3427     int gwa_type = ipv6_addr_type(gw_addr);
3428     bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3429     const struct net_device *dev = *_dev;
3430     bool need_addr_check = !dev;
3431     int err = -EINVAL;
3432
3433     /* if gw_addr is local we will fail to detect this in case
3434      * address is still TENTATIVE (DAD in progress). rt6_lookup()
3435      * will return already-added prefix route via interface that
3436      * prefix route was assigned to, which might be non-loopback.
3437      */
3438     if (dev &&
3439         ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3440         NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3441         goto out;
3442     }
3443
3444     if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3445         /* IPv6 strictly inhibits using not link-local
3446          * addresses as nexthop address.
3447          * Otherwise, router will not able to send redirects.
3448          * It is very good, but in some (rare!) circumstances
3449          * (SIT, PtP, NBMA NOARP links) it is handy to allow
3450          * some exceptions. --ANK
3451          * We allow IPv4-mapped nexthops to support RFC4798-type
3452          * addressing
3453          */
3454         if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3455             NL_SET_ERR_MSG(extack, "Invalid gateway address");
3456             goto out;
3457         }
3458
3459         rcu_read_lock();
3460
3461         if (cfg->fc_flags & RTNH_F_ONLINK)
3462             err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3463         else
3464             err = ip6_route_check_nh(net, cfg, _dev, idev);
3465
3466         rcu_read_unlock();
3467
3468         if (err)
3469             goto out;
3470     }
3471
3472     /* reload in case device was changed */
3473     dev = *_dev;
3474
3475     err = -EINVAL;
3476     if (!dev) {
3477         NL_SET_ERR_MSG(extack, "Egress device not specified");
3478         goto out;
3479     } else if (dev->flags & IFF_LOOPBACK) {
3480         NL_SET_ERR_MSG(extack,
3481                    "Egress device can not be loopback device for this route");
3482         goto out;
3483     }
3484
3485     /* if we did not check gw_addr above, do so now that the
3486      * egress device has been resolved.
3487      */
3488     if (need_addr_check &&
3489         ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3490         NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3491         goto out;
3492     }
3493
3494     err = 0;
3495 out:
3496     return err;
3497 }
3498
3499 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3500 {
3501     if ((flags & RTF_REJECT) ||
3502         (dev && (dev->flags & IFF_LOOPBACK) &&
3503          !(addr_type & IPV6_ADDR_LOOPBACK) &&
3504          !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3505         return true;
3506
3507     return false;
3508 }
3509
3510 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3511          struct fib6_config *cfg, gfp_t gfp_flags,
3512          struct netlink_ext_ack *extack)
3513 {
3514     struct net_device *dev = NULL;
3515     struct inet6_dev *idev = NULL;
3516     int addr_type;
3517     int err;
3518
3519     fib6_nh->fib_nh_family = AF_INET6;
3520 #ifdef CONFIG_IPV6_ROUTER_PREF
3521     fib6_nh->last_probe = jiffies;
3522 #endif
3523     if (cfg->fc_is_fdb) {
3524         fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3525         fib6_nh->fib_nh_gw_family = AF_INET6;
3526         return 0;
3527     }
3528
3529     err = -ENODEV;
3530     if (cfg->fc_ifindex) {
3531         dev = dev_get_by_index(net, cfg->fc_ifindex);
3532         if (!dev)
3533             goto out;
3534         idev = in6_dev_get(dev);
3535         if (!idev)
3536             goto out;
3537     }
3538
3539     if (cfg->fc_flags & RTNH_F_ONLINK) {
3540         if (!dev) {
3541             NL_SET_ERR_MSG(extack,
3542                        "Nexthop device required for onlink");
3543             goto out;
3544         }
3545
3546         if (!(dev->flags & IFF_UP)) {
3547             NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3548             err = -ENETDOWN;
3549             goto out;
3550         }
3551
3552         fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3553     }
3554
3555     fib6_nh->fib_nh_weight = 1;
3556
3557     /* We cannot add true routes via loopback here,
3558      * they would result in kernel looping; promote them to reject routes
3559      */
3560     addr_type = ipv6_addr_type(&cfg->fc_dst);
3561     if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3562         /* hold loopback dev/idev if we haven't done so. */
3563         if (dev != net->loopback_dev) {
3564             if (dev) {
3565                 dev_put(dev);
3566                 in6_dev_put(idev);
3567             }
3568             dev = net->loopback_dev;
3569             dev_hold(dev);
3570             idev = in6_dev_get(dev);
3571             if (!idev) {
3572                 err = -ENODEV;
3573                 goto out;
3574             }
3575         }
3576         goto pcpu_alloc;
3577     }
3578
3579     if (cfg->fc_flags & RTF_GATEWAY) {
3580         err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3581         if (err)
3582             goto out;
3583
3584         fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3585         fib6_nh->fib_nh_gw_family = AF_INET6;
3586     }
3587
3588     err = -ENODEV;
3589     if (!dev)
3590         goto out;
3591
3592     if (idev->cnf.disable_ipv6) {
3593         NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3594         err = -EACCES;
3595         goto out;
3596     }
3597
3598     if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3599         NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3600         err = -ENETDOWN;
3601         goto out;
3602     }
3603
3604     if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3605         !netif_carrier_ok(dev))
3606         fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3607
3608     err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3609                  cfg->fc_encap_type, cfg, gfp_flags, extack);
3610     if (err)
3611         goto out;
3612
3613 pcpu_alloc:
3614     fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3615     if (!fib6_nh->rt6i_pcpu) {
3616         err = -ENOMEM;
3617         goto out;
3618     }
3619
3620     fib6_nh->fib_nh_dev = dev;
3621     netdev_tracker_alloc(dev, &fib6_nh->fib_nh_dev_tracker, gfp_flags);
3622
3623     fib6_nh->fib_nh_oif = dev->ifindex;
3624     err = 0;
3625 out:
3626     if (idev)
3627         in6_dev_put(idev);
3628
3629     if (err) {
3630         lwtstate_put(fib6_nh->fib_nh_lws);
3631         fib6_nh->fib_nh_lws = NULL;
3632         dev_put(dev);
3633     }
3634
3635     return err;
3636 }
3637
3638 void fib6_nh_release(struct fib6_nh *fib6_nh)
3639 {
3640     struct rt6_exception_bucket *bucket;
3641
3642     rcu_read_lock();
3643
3644     fib6_nh_flush_exceptions(fib6_nh, NULL);
3645     bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3646     if (bucket) {
3647         rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3648         kfree(bucket);
3649     }
3650
3651     rcu_read_unlock();
3652
3653     fib6_nh_release_dsts(fib6_nh);
3654     free_percpu(fib6_nh->rt6i_pcpu);
3655
3656     fib_nh_common_release(&fib6_nh->nh_common);
3657 }
3658
3659 void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3660 {
3661     int cpu;
3662
3663     if (!fib6_nh->rt6i_pcpu)
3664         return;
3665
3666     for_each_possible_cpu(cpu) {
3667         struct rt6_info *pcpu_rt, **ppcpu_rt;
3668
3669         ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3670         pcpu_rt = xchg(ppcpu_rt, NULL);
3671         if (pcpu_rt) {
3672             dst_dev_put(&pcpu_rt->dst);
3673             dst_release(&pcpu_rt->dst);
3674         }
3675     }
3676 }
3677
3678 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3679                           gfp_t gfp_flags,
3680                           struct netlink_ext_ack *extack)
3681 {
3682     struct net *net = cfg->fc_nlinfo.nl_net;
3683     struct fib6_info *rt = NULL;
3684     struct nexthop *nh = NULL;
3685     struct fib6_table *table;
3686     struct fib6_nh *fib6_nh;
3687     int err = -EINVAL;
3688     int addr_type;
3689
3690     /* RTF_PCPU is an internal flag; can not be set by userspace */
3691     if (cfg->fc_flags & RTF_PCPU) {
3692         NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3693         goto out;
3694     }
3695
3696     /* RTF_CACHE is an internal flag; can not be set by userspace */
3697     if (cfg->fc_flags & RTF_CACHE) {
3698         NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3699         goto out;
3700     }
3701
3702     if (cfg->fc_type > RTN_MAX) {
3703         NL_SET_ERR_MSG(extack, "Invalid route type");
3704         goto out;
3705     }
3706
3707     if (cfg->fc_dst_len > 128) {
3708         NL_SET_ERR_MSG(extack, "Invalid prefix length");
3709         goto out;
3710     }
3711     if (cfg->fc_src_len > 128) {
3712         NL_SET_ERR_MSG(extack, "Invalid source address length");
3713         goto out;
3714     }
3715 #ifndef CONFIG_IPV6_SUBTREES
3716     if (cfg->fc_src_len) {
3717         NL_SET_ERR_MSG(extack,
3718                    "Specifying source address requires IPV6_SUBTREES to be enabled");
3719         goto out;
3720     }
3721 #endif
3722     if (cfg->fc_nh_id) {
3723         nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3724         if (!nh) {
3725             NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3726             goto out;
3727         }
3728         err = fib6_check_nexthop(nh, cfg, extack);
3729         if (err)
3730             goto out;
3731     }
3732
3733     err = -ENOBUFS;
3734     if (cfg->fc_nlinfo.nlh &&
3735         !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3736         table = fib6_get_table(net, cfg->fc_table);
3737         if (!table) {
3738             pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3739             table = fib6_new_table(net, cfg->fc_table);
3740         }
3741     } else {
3742         table = fib6_new_table(net, cfg->fc_table);
3743     }
3744
3745     if (!table)
3746         goto out;
3747
3748     err = -ENOMEM;
3749     rt = fib6_info_alloc(gfp_flags, !nh);
3750     if (!rt)
3751         goto out;
3752
3753     rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3754                            extack);
3755     if (IS_ERR(rt->fib6_metrics)) {
3756         err = PTR_ERR(rt->fib6_metrics);
3757         /* Do not leave garbage there. */
3758         rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3759         goto out_free;
3760     }
3761
3762     if (cfg->fc_flags & RTF_ADDRCONF)
3763         rt->dst_nocount = true;
3764
3765     if (cfg->fc_flags & RTF_EXPIRES)
3766         fib6_set_expires(rt, jiffies +
3767                 clock_t_to_jiffies(cfg->fc_expires));
3768     else
3769         fib6_clean_expires(rt);
3770
3771     if (cfg->fc_protocol == RTPROT_UNSPEC)
3772         cfg->fc_protocol = RTPROT_BOOT;
3773     rt->fib6_protocol = cfg->fc_protocol;
3774
3775     rt->fib6_table = table;
3776     rt->fib6_metric = cfg->fc_metric;
3777     rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3778     rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3779
3780     ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3781     rt->fib6_dst.plen = cfg->fc_dst_len;
3782
3783 #ifdef CONFIG_IPV6_SUBTREES
3784     ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3785     rt->fib6_src.plen = cfg->fc_src_len;
3786 #endif
3787     if (nh) {
3788         if (rt->fib6_src.plen) {
3789             NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3790             goto out_free;
3791         }
3792         if (!nexthop_get(nh)) {
3793             NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3794             goto out_free;
3795         }
3796         rt->nh = nh;
3797         fib6_nh = nexthop_fib6_nh(rt->nh);
3798     } else {
3799         err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3800         if (err)
3801             goto out;
3802
3803         fib6_nh = rt->fib6_nh;
3804
3805         /* We cannot add true routes via loopback here, they would
3806          * result in kernel looping; promote them to reject routes
3807          */
3808         addr_type = ipv6_addr_type(&cfg->fc_dst);
3809         if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3810                    addr_type))
3811             rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3812     }
3813
3814     if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3815         struct net_device *dev = fib6_nh->fib_nh_dev;
3816
3817         if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3818             NL_SET_ERR_MSG(extack, "Invalid source address");
3819             err = -EINVAL;
3820             goto out;
3821         }
3822         rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3823         rt->fib6_prefsrc.plen = 128;
3824     } else
3825         rt->fib6_prefsrc.plen = 0;
3826
3827     return rt;
3828 out:
3829     fib6_info_release(rt);
3830     return ERR_PTR(err);
3831 out_free:
3832     ip_fib_metrics_put(rt->fib6_metrics);
3833     kfree(rt);
3834     return ERR_PTR(err);
3835 }
3836
3837 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3838           struct netlink_ext_ack *extack)
3839 {
3840     struct fib6_info *rt;
3841     int err;
3842
3843     rt = ip6_route_info_create(cfg, gfp_flags, extack);
3844     if (IS_ERR(rt))
3845         return PTR_ERR(rt);
3846
3847     err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3848     fib6_info_release(rt);
3849
3850     return err;
3851 }
3852
3853 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3854 {
3855     struct net *net = info->nl_net;
3856     struct fib6_table *table;
3857     int err;
3858
3859     if (rt == net->ipv6.fib6_null_entry) {
3860         err = -ENOENT;
3861         goto out;
3862     }
3863
3864     table = rt->fib6_table;
3865     spin_lock_bh(&table->tb6_lock);
3866     err = fib6_del(rt, info);
3867     spin_unlock_bh(&table->tb6_lock);
3868
3869 out:
3870     fib6_info_release(rt);
3871     return err;
3872 }
3873
3874 int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3875 {
3876     struct nl_info info = {
3877         .nl_net = net,
3878         .skip_notify = skip_notify
3879     };
3880
3881     return __ip6_del_rt(rt, &info);
3882 }
3883
3884 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3885 {
3886     struct nl_info *info = &cfg->fc_nlinfo;
3887     struct net *net = info->nl_net;
3888     struct sk_buff *skb = NULL;
3889     struct fib6_table *table;
3890     int err = -ENOENT;
3891
3892     if (rt == net->ipv6.fib6_null_entry)
3893         goto out_put;
3894     table = rt->fib6_table;
3895     spin_lock_bh(&table->tb6_lock);
3896
3897     if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3898         struct fib6_info *sibling, *next_sibling;
3899         struct fib6_node *fn;
3900
3901         /* prefer to send a single notification with all hops */
3902         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3903         if (skb) {
3904             u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3905
3906             if (rt6_fill_node(net, skb, rt, NULL,
3907                       NULL, NULL, 0, RTM_DELROUTE,
3908                       info->portid, seq, 0) < 0) {
3909                 kfree_skb(skb);
3910                 skb = NULL;
3911             } else
3912                 info->skip_notify = 1;
3913         }
3914
3915         /* 'rt' points to the first sibling route. If it is not the
3916          * leaf, then we do not need to send a notification. Otherwise,
3917          * we need to check if the last sibling has a next route or not
3918          * and emit a replace or delete notification, respectively.
3919          */
3920         info->skip_notify_kernel = 1;
3921         fn = rcu_dereference_protected(rt->fib6_node,
3922                         lockdep_is_held(&table->tb6_lock));
3923         if (rcu_access_pointer(fn->leaf) == rt) {
3924             struct fib6_info *last_sibling, *replace_rt;
3925
3926             last_sibling = list_last_entry(&rt->fib6_siblings,
3927                                struct fib6_info,
3928                                fib6_siblings);
3929             replace_rt = rcu_dereference_protected(
3930                         last_sibling->fib6_next,
3931                         lockdep_is_held(&table->tb6_lock));
3932             if (replace_rt)
3933                 call_fib6_entry_notifiers_replace(net,
3934                                   replace_rt);
3935             else
3936                 call_fib6_multipath_entry_notifiers(net,
3937                                FIB_EVENT_ENTRY_DEL,
3938                                rt, rt->fib6_nsiblings,
3939                                NULL);
3940         }
3941         list_for_each_entry_safe(sibling, next_sibling,
3942                      &rt->fib6_siblings,
3943                      fib6_siblings) {
3944             err = fib6_del(sibling, info);
3945             if (err)
3946                 goto out_unlock;
3947         }
3948     }
3949
3950     err = fib6_del(rt, info);
3951 out_unlock:
3952     spin_unlock_bh(&table->tb6_lock);
3953 out_put:
3954     fib6_info_release(rt);
3955
3956     if (skb) {
3957         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3958                 info->nlh, gfp_any());
3959     }
3960     return err;
3961 }
3962
3963 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3964 {
3965     int rc = -ESRCH;
3966
3967     if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3968         goto out;
3969
3970     if (cfg->fc_flags & RTF_GATEWAY &&
3971         !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3972         goto out;
3973
3974     rc = rt6_remove_exception_rt(rt);
3975 out:
3976     return rc;
3977 }
3978
3979 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3980                  struct fib6_nh *nh)
3981 {
3982     struct fib6_result res = {
3983         .f6i = rt,
3984         .nh = nh,
3985     };
3986     struct rt6_info *rt_cache;
3987
3988     rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3989     if (rt_cache)
3990         return __ip6_del_cached_rt(rt_cache, cfg);
3991
3992     return 0;
3993 }
3994
3995 struct fib6_nh_del_cached_rt_arg {
3996     struct fib6_config *cfg;
3997     struct fib6_info *f6i;
3998 };
3999
4000 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4001 {
4002     struct fib6_nh_del_cached_rt_arg *arg = _arg;
4003     int rc;
4004
4005     rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4006     return rc != -ESRCH ? rc : 0;
4007 }
4008
4009 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4010 {
4011     struct fib6_nh_del_cached_rt_arg arg = {
4012         .cfg = cfg,
4013         .f6i = f6i
4014     };
4015
4016     return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4017 }
4018
4019 static int ip6_route_del(struct fib6_config *cfg,
4020              struct netlink_ext_ack *extack)
4021 {
4022     struct fib6_table *table;
4023     struct fib6_info *rt;
4024     struct fib6_node *fn;
4025     int err = -ESRCH;
4026
4027     table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4028     if (!table) {
4029         NL_SET_ERR_MSG(extack, "FIB table does not exist");
4030         return err;
4031     }
4032
4033     rcu_read_lock();
4034
4035     fn = fib6_locate(&table->tb6_root,
4036              &cfg->fc_dst, cfg->fc_dst_len,
4037              &cfg->fc_src, cfg->fc_src_len,
4038              !(cfg->fc_flags & RTF_CACHE));
4039
4040     if (fn) {
4041         for_each_fib6_node_rt_rcu(fn) {
4042             struct fib6_nh *nh;
4043
4044             if (rt->nh && cfg->fc_nh_id &&
4045                 rt->nh->id != cfg->fc_nh_id)
4046                 continue;
4047
4048             if (cfg->fc_flags & RTF_CACHE) {
4049                 int rc = 0;
4050
4051                 if (rt->nh) {
4052                     rc = ip6_del_cached_rt_nh(cfg, rt);
4053                 } else if (cfg->fc_nh_id) {
4054                     continue;
4055                 } else {
4056                     nh = rt->fib6_nh;
4057                     rc = ip6_del_cached_rt(cfg, rt, nh);
4058                 }
4059                 if (rc != -ESRCH) {
4060                     rcu_read_unlock();
4061                     return rc;
4062                 }
4063                 continue;
4064             }
4065
4066             if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4067                 continue;
4068             if (cfg->fc_protocol &&
4069                 cfg->fc_protocol != rt->fib6_protocol)
4070                 continue;
4071
4072             if (rt->nh) {
4073                 if (!fib6_info_hold_safe(rt))
4074                     continue;
4075                 rcu_read_unlock();
4076
4077                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4078             }
4079             if (cfg->fc_nh_id)
4080                 continue;
4081
4082             nh = rt->fib6_nh;
4083             if (cfg->fc_ifindex &&
4084                 (!nh->fib_nh_dev ||
4085                  nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4086                 continue;
4087             if (cfg->fc_flags & RTF_GATEWAY &&
4088                 !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4089                 continue;
4090             if (!fib6_info_hold_safe(rt))
4091                 continue;
4092             rcu_read_unlock();
4093
4094             /* if gateway was specified only delete the one hop */
4095             if (cfg->fc_flags & RTF_GATEWAY)
4096                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4097
4098             return __ip6_del_rt_siblings(rt, cfg);
4099         }
4100     }
4101     rcu_read_unlock();
4102
4103     return err;
4104 }
4105
4106 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4107 {
4108     struct netevent_redirect netevent;
4109     struct rt6_info *rt, *nrt = NULL;
4110     struct fib6_result res = {};
4111     struct ndisc_options ndopts;
4112     struct inet6_dev *in6_dev;
4113     struct neighbour *neigh;
4114     struct rd_msg *msg;
4115     int optlen, on_link;
4116     u8 *lladdr;
4117
4118     optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4119     optlen -= sizeof(*msg);
4120
4121     if (optlen < 0) {
4122         net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4123         return;
4124     }
4125
4126     msg = (struct rd_msg *)icmp6_hdr(skb);
4127
4128     if (ipv6_addr_is_multicast(&msg->dest)) {
4129         net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4130         return;
4131     }
4132
4133     on_link = 0;
4134     if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4135         on_link = 1;
4136     } else if (ipv6_addr_type(&msg->target) !=
4137            (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4138         net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4139         return;
4140     }
4141
4142     in6_dev = __in6_dev_get(skb->dev);
4143     if (!in6_dev)
4144         return;
4145     if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
4146         return;
4147
4148     /* RFC2461 8.1:
4149      *  The IP source address of the Redirect MUST be the same as the current
4150      *  first-hop router for the specified ICMP Destination Address.
4151      */
4152
4153     if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4154         net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4155         return;
4156     }
4157
4158     lladdr = NULL;
4159     if (ndopts.nd_opts_tgt_lladdr) {
4160         lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4161                          skb->dev);
4162         if (!lladdr) {
4163             net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4164             return;
4165         }
4166     }
4167
4168     rt = (struct rt6_info *) dst;
4169     if (rt->rt6i_flags & RTF_REJECT) {
4170         net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4171         return;
4172     }
4173
4174     /* Redirect received -> path was valid.
4175      * Look, redirects are sent only in response to data packets,
4176      * so that this nexthop apparently is reachable. --ANK
4177      */
4178     dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4179
4180     neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4181     if (!neigh)
4182         return;
4183
4184     /*
4185      *  We have finally decided to accept it.
4186      */
4187
4188     ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4189              NEIGH_UPDATE_F_WEAK_OVERRIDE|
4190              NEIGH_UPDATE_F_OVERRIDE|
4191              (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4192                      NEIGH_UPDATE_F_ISROUTER)),
4193              NDISC_REDIRECT, &ndopts);
4194
4195     rcu_read_lock();
4196     res.f6i = rcu_dereference(rt->from);
4197     if (!res.f6i)
4198         goto out;
4199
4200     if (res.f6i->nh) {
4201         struct fib6_nh_match_arg arg = {
4202             .dev = dst->dev,
4203             .gw = &rt->rt6i_gateway,
4204         };
4205
4206         nexthop_for_each_fib6_nh(res.f6i->nh,
4207                      fib6_nh_find_match, &arg);
4208
4209         /* fib6_info uses a nexthop that does not have fib6_nh
4210          * using the dst->dev. Should be impossible
4211          */
4212         if (!arg.match)
4213             goto out;
4214         res.nh = arg.match;
4215     } else {
4216         res.nh = res.f6i->fib6_nh;
4217     }
4218
4219     res.fib6_flags = res.f6i->fib6_flags;
4220     res.fib6_type = res.f6i->fib6_type;
4221     nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4222     if (!nrt)
4223         goto out;
4224
4225     nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4226     if (on_link)
4227         nrt->rt6i_flags &= ~RTF_GATEWAY;
4228
4229     nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4230
4231     /* rt6_insert_exception() will take care of duplicated exceptions */
4232     if (rt6_insert_exception(nrt, &res)) {
4233         dst_release_immediate(&nrt->dst);
4234         goto out;
4235     }
4236
4237     netevent.old = &rt->dst;
4238     netevent.new = &nrt->dst;
4239     netevent.daddr = &msg->dest;
4240     netevent.neigh = neigh;
4241     call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4242
4243 out:
4244     rcu_read_unlock();
4245     neigh_release(neigh);
4246 }
4247
4248 #ifdef CONFIG_IPV6_ROUTE_INFO
4249 static struct fib6_info *rt6_get_route_info(struct net *net,
4250                        const struct in6_addr *prefix, int prefixlen,
4251                        const struct in6_addr *gwaddr,
4252                        struct net_device *dev)
4253 {
4254     u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4255     int ifindex = dev->ifindex;
4256     struct fib6_node *fn;
4257     struct fib6_info *rt = NULL;
4258     struct fib6_table *table;
4259
4260     table = fib6_get_table(net, tb_id);
4261     if (!table)
4262         return NULL;
4263
4264     rcu_read_lock();
4265     fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4266     if (!fn)
4267         goto out;
4268
4269     for_each_fib6_node_rt_rcu(fn) {
4270         /* these routes do not use nexthops */
4271         if (rt->nh)
4272             continue;
4273         if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4274             continue;
4275         if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4276             !rt->fib6_nh->fib_nh_gw_family)
4277             continue;
4278         if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4279             continue;
4280         if (!fib6_info_hold_safe(rt))
4281             continue;
4282         break;
4283     }
4284 out:
4285     rcu_read_unlock();
4286     return rt;
4287 }
4288
4289 static struct fib6_info *rt6_add_route_info(struct net *net,
4290                        const struct in6_addr *prefix, int prefixlen,
4291                        const struct in6_addr *gwaddr,
4292                        struct net_device *dev,
4293                        unsigned int pref)
4294 {
4295     struct fib6_config cfg = {
4296         .fc_metric  = IP6_RT_PRIO_USER,
4297         .fc_ifindex = dev->ifindex,
4298         .fc_dst_len = prefixlen,
4299         .fc_flags   = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4300                   RTF_UP | RTF_PREF(pref),
4301         .fc_protocol = RTPROT_RA,
4302         .fc_type = RTN_UNICAST,
4303         .fc_nlinfo.portid = 0,
4304         .fc_nlinfo.nlh = NULL,
4305         .fc_nlinfo.nl_net = net,
4306     };
4307
4308     cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4309     cfg.fc_dst = *prefix;
4310     cfg.fc_gateway = *gwaddr;
4311
4312     /* We should treat it as a default route if prefix length is 0. */
4313     if (!prefixlen)
4314         cfg.fc_flags |= RTF_DEFAULT;
4315
4316     ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4317
4318     return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4319 }
4320 #endif
4321
4322 struct fib6_info *rt6_get_dflt_router(struct net *net,
4323                      const struct in6_addr *addr,
4324                      struct net_device *dev)
4325 {
4326     u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4327     struct fib6_info *rt;
4328     struct fib6_table *table;
4329
4330     table = fib6_get_table(net, tb_id);
4331     if (!table)
4332         return NULL;
4333
4334     rcu_read_lock();
4335     for_each_fib6_node_rt_rcu(&table->tb6_root) {
4336         struct fib6_nh *nh;
4337
4338         /* RA routes do not use nexthops */
4339         if (rt->nh)
4340             continue;
4341
4342         nh = rt->fib6_nh;
4343         if (dev == nh->fib_nh_dev &&
4344             ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4345             ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4346             break;
4347     }
4348     if (rt && !fib6_info_hold_safe(rt))
4349         rt = NULL;
4350     rcu_read_unlock();
4351     return rt;
4352 }
4353
4354 struct fib6_info *rt6_add_dflt_router(struct net *net,
4355                      const struct in6_addr *gwaddr,
4356                      struct net_device *dev,
4357                      unsigned int pref,
4358                      u32 defrtr_usr_metric)
4359 {
4360     struct fib6_config cfg = {
4361         .fc_table   = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4362         .fc_metric  = defrtr_usr_metric,
4363         .fc_ifindex = dev->ifindex,
4364         .fc_flags   = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4365                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4366         .fc_protocol = RTPROT_RA,
4367         .fc_type = RTN_UNICAST,
4368         .fc_nlinfo.portid = 0,
4369         .fc_nlinfo.nlh = NULL,
4370         .fc_nlinfo.nl_net = net,
4371     };
4372
4373     cfg.fc_gateway = *gwaddr;
4374
4375     if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4376         struct fib6_table *table;
4377
4378         table = fib6_get_table(dev_net(dev), cfg.fc_table);
4379         if (table)
4380             table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4381     }
4382
4383     return rt6_get_dflt_router(net, gwaddr, dev);
4384 }
4385
4386 static void __rt6_purge_dflt_routers(struct net *net,
4387                      struct fib6_table *table)
4388 {
4389     struct fib6_info *rt;
4390
4391 restart:
4392     rcu_read_lock();
4393     for_each_fib6_node_rt_rcu(&table->tb6_root) {
4394         struct net_device *dev = fib6_info_nh_dev(rt);
4395         struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4396
4397         if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4398             (!idev || idev->cnf.accept_ra != 2) &&
4399             fib6_info_hold_safe(rt)) {
4400             rcu_read_unlock();
4401             ip6_del_rt(net, rt, false);
4402             goto restart;
4403         }
4404     }
4405     rcu_read_unlock();
4406
4407     table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4408 }
4409
4410 void rt6_purge_dflt_routers(struct net *net)
4411 {
4412     struct fib6_table *table;
4413     struct hlist_head *head;
4414     unsigned int h;
4415
4416     rcu_read_lock();
4417
4418     for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4419         head = &net->ipv6.fib_table_hash[h];
4420         hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4421             if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4422                 __rt6_purge_dflt_routers(net, table);
4423         }
4424     }
4425
4426     rcu_read_unlock();
4427 }
4428
4429 static void rtmsg_to_fib6_config(struct net *net,
4430                  struct in6_rtmsg *rtmsg,
4431                  struct fib6_config *cfg)
4432 {
4433     *cfg = (struct fib6_config){
4434         .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4435              : RT6_TABLE_MAIN,
4436         .fc_ifindex = rtmsg->rtmsg_ifindex,
4437         .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4438         .fc_expires = rtmsg->rtmsg_info,
4439         .fc_dst_len = rtmsg->rtmsg_dst_len,
4440         .fc_src_len = rtmsg->rtmsg_src_len,
4441         .fc_flags = rtmsg->rtmsg_flags,
4442         .fc_type = rtmsg->rtmsg_type,
4443
4444         .fc_nlinfo.nl_net = net,
4445
4446         .fc_dst = rtmsg->rtmsg_dst,
4447         .fc_src = rtmsg->rtmsg_src,
4448         .fc_gateway = rtmsg->rtmsg_gateway,
4449     };
4450 }
4451
4452 int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4453 {
4454     struct fib6_config cfg;
4455     int err;
4456
4457     if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4458         return -EINVAL;
4459     if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4460         return -EPERM;
4461
4462     rtmsg_to_fib6_config(net, rtmsg, &cfg);
4463
4464     rtnl_lock();
4465     switch (cmd) {
4466     case SIOCADDRT:
4467         err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4468         break;
4469     case SIOCDELRT:
4470         err = ip6_route_del(&cfg, NULL);
4471         break;
4472     }
4473     rtnl_unlock();
4474     return err;
4475 }
4476
4477 /*
4478  *  Drop the packet on the floor
4479  */
4480
4481 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4482 {
4483     struct dst_entry *dst = skb_dst(skb);
4484     struct net *net = dev_net(dst->dev);
4485     struct inet6_dev *idev;
4486     SKB_DR(reason);
4487     int type;
4488
4489     if (netif_is_l3_master(skb->dev) ||
4490         dst->dev == net->loopback_dev)
4491         idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4492     else
4493         idev = ip6_dst_idev(dst);
4494
4495     switch (ipstats_mib_noroutes) {
4496     case IPSTATS_MIB_INNOROUTES:
4497         type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4498         if (type == IPV6_ADDR_ANY) {
4499             SKB_DR_SET(reason, IP_INADDRERRORS);
4500             IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4501             break;
4502         }
4503         SKB_DR_SET(reason, IP_INNOROUTES);
4504         fallthrough;
4505     case IPSTATS_MIB_OUTNOROUTES:
4506         SKB_DR_OR(reason, IP_OUTNOROUTES);
4507         IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4508         break;
4509     }
4510
4511     /* Start over by dropping the dst for l3mdev case */
4512     if (netif_is_l3_master(skb->dev))
4513         skb_dst_drop(skb);
4514
4515     icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4516     kfree_skb_reason(skb, reason);
4517     return 0;
4518 }
4519
4520 static int ip6_pkt_discard(struct sk_buff *skb)
4521 {
4522     return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4523 }
4524
4525 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4526 {
4527     skb->dev = skb_dst(skb)->dev;
4528     return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4529 }
4530
4531 static int ip6_pkt_prohibit(struct sk_buff *skb)
4532 {
4533     return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4534 }
4535
4536 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4537 {
4538     skb->dev = skb_dst(skb)->dev;
4539     return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4540 }
4541
4542 /*
4543  *  Allocate a dst for local (unicast / anycast) address.
4544  */
4545
4546 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4547                      struct inet6_dev *idev,
4548                      const struct in6_addr *addr,
4549                      bool anycast, gfp_t gfp_flags)
4550 {
4551     struct fib6_config cfg = {
4552         .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4553         .fc_ifindex = idev->dev->ifindex,
4554         .fc_flags = RTF_UP | RTF_NONEXTHOP,
4555         .fc_dst = *addr,
4556         .fc_dst_len = 128,
4557         .fc_protocol = RTPROT_KERNEL,
4558         .fc_nlinfo.nl_net = net,
4559         .fc_ignore_dev_down = true,
4560     };
4561     struct fib6_info *f6i;
4562
4563     if (anycast) {
4564         cfg.fc_type = RTN_ANYCAST;
4565         cfg.fc_flags |= RTF_ANYCAST;
4566     } else {
4567         cfg.fc_type = RTN_LOCAL;
4568         cfg.fc_flags |= RTF_LOCAL;
4569     }
4570
4571     f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4572     if (!IS_ERR(f6i)) {
4573         f6i->dst_nocount = true;
4574
4575         if (!anycast &&
4576             (net->ipv6.devconf_all->disable_policy ||
4577              idev->cnf.disable_policy))
4578             f6i->dst_nopolicy = true;
4579     }
4580
4581     return f6i;
4582 }
4583
4584 /* remove deleted ip from prefsrc entries */
4585 struct arg_dev_net_ip {
4586     struct net_device *dev;
4587     struct net *net;
4588     struct in6_addr *addr;
4589 };
4590
4591 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4592 {
4593     struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4594     struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4595     struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4596
4597     if (!rt->nh &&
4598         ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4599         rt != net->ipv6.fib6_null_entry &&
4600         ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4601         spin_lock_bh(&rt6_exception_lock);
4602         /* remove prefsrc entry */
4603         rt->fib6_prefsrc.plen = 0;
4604         spin_unlock_bh(&rt6_exception_lock);
4605     }
4606     return 0;
4607 }
4608
4609 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4610 {
4611     struct net *net = dev_net(ifp->idev->dev);
4612     struct arg_dev_net_ip adni = {
4613         .dev = ifp->idev->dev,
4614         .net = net,
4615         .addr = &ifp->addr,
4616     };
4617     fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4618 }
4619
4620 #define RTF_RA_ROUTER       (RTF_ADDRCONF | RTF_DEFAULT)
4621
4622 /* Remove routers and update dst entries when gateway turn into host. */
4623 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4624 {
4625     struct in6_addr *gateway = (struct in6_addr *)arg;
4626     struct fib6_nh *nh;
4627
4628     /* RA routes do not use nexthops */
4629     if (rt->nh)
4630         return 0;
4631
4632     nh = rt->fib6_nh;
4633     if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4634         nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4635         return -1;
4636
4637     /* Further clean up cached routes in exception table.
4638      * This is needed because cached route may have a different
4639      * gateway than its 'parent' in the case of an ip redirect.
4640      */
4641     fib6_nh_exceptions_clean_tohost(nh, gateway);
4642
4643     return 0;
4644 }
4645
4646 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4647 {
4648     fib6_clean_all(net, fib6_clean_tohost, gateway);
4649 }
4650
4651 struct arg_netdev_event {
4652     const struct net_device *dev;
4653     union {
4654         unsigned char nh_flags;
4655         unsigned long event;
4656     };
4657 };
4658
4659 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4660 {
4661     struct fib6_info *iter;
4662     struct fib6_node *fn;
4663
4664     fn = rcu_dereference_protected(rt->fib6_node,
4665             lockdep_is_held(&rt->fib6_table->tb6_lock));
4666     iter = rcu_dereference_protected(fn->leaf,
4667             lockdep_is_held(&rt->fib6_table->tb6_lock));
4668     while (iter) {
4669         if (iter->fib6_metric == rt->fib6_metric &&
4670             rt6_qualify_for_ecmp(iter))
4671             return iter;
4672         iter = rcu_dereference_protected(iter->fib6_next,
4673                 lockdep_is_held(&rt->fib6_table->tb6_lock));
4674     }
4675
4676     return NULL;
4677 }
4678
4679 /* only called for fib entries with builtin fib6_nh */
4680 static bool rt6_is_dead(const struct fib6_info *rt)
4681 {
4682     if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4683         (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4684          ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4685         return true;
4686
4687     return false;
4688 }
4689
4690 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4691 {
4692     struct fib6_info *iter;
4693     int total = 0;
4694
4695     if (!rt6_is_dead(rt))
4696         total += rt->fib6_nh->fib_nh_weight;
4697
4698     list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4699         if (!rt6_is_dead(iter))
4700             total += iter->fib6_nh->fib_nh_weight;
4701     }
4702
4703     return total;
4704 }
4705
4706 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4707 {
4708     int upper_bound = -1;
4709
4710     if (!rt6_is_dead(rt)) {
4711         *weight += rt->fib6_nh->fib_nh_weight;
4712         upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4713                             total) - 1;
4714     }
4715     atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4716 }
4717
4718 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4719 {
4720     struct fib6_info *iter;
4721     int weight = 0;
4722
4723     rt6_upper_bound_set(rt, &weight, total);
4724
4725     list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4726         rt6_upper_bound_set(iter, &weight, total);
4727 }
4728
4729 void rt6_multipath_rebalance(struct fib6_info *rt)
4730 {
4731     struct fib6_info *first;
4732     int total;
4733
4734     /* In case the entire multipath route was marked for flushing,
4735      * then there is no need to rebalance upon the removal of every
4736      * sibling route.
4737      */
4738     if (!rt->fib6_nsiblings || rt->should_flush)
4739         return;
4740
4741     /* During lookup routes are evaluated in order, so we need to
4742      * make sure upper bounds are assigned from the first sibling
4743      * onwards.
4744      */
4745     first = rt6_multipath_first_sibling(rt);
4746     if (WARN_ON_ONCE(!first))
4747         return;
4748
4749     total = rt6_multipath_total_weight(first);
4750     rt6_multipath_upper_bound_set(first, total);
4751 }
4752
4753 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4754 {
4755     const struct arg_netdev_event *arg = p_arg;
4756     struct net *net = dev_net(arg->dev);
4757
4758     if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4759         rt->fib6_nh->fib_nh_dev == arg->dev) {
4760         rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4761         fib6_update_sernum_upto_root(net, rt);
4762         rt6_multipath_rebalance(rt);
4763     }
4764
4765     return 0;
4766 }
4767
4768 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4769 {
4770     struct arg_netdev_event arg = {
4771         .dev = dev,
4772         {
4773             .nh_flags = nh_flags,
4774         },
4775     };
4776
4777     if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4778         arg.nh_flags |= RTNH_F_LINKDOWN;
4779
4780     fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4781 }
4782
4783 /* only called for fib entries with inline fib6_nh */
4784 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4785                    const struct net_device *dev)
4786 {
4787     struct fib6_info *iter;
4788
4789     if (rt->fib6_nh->fib_nh_dev == dev)
4790         return true;
4791     list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4792         if (iter->fib6_nh->fib_nh_dev == dev)
4793             return true;
4794
4795     return false;
4796 }
4797
4798 static void rt6_multipath_flush(struct fib6_info *rt)
4799 {
4800     struct fib6_info *iter;
4801
4802     rt->should_flush = 1;
4803     list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4804         iter->should_flush = 1;
4805 }
4806
4807 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4808                          const struct net_device *down_dev)
4809 {
4810     struct fib6_info *iter;
4811     unsigned int dead = 0;
4812
4813     if (rt->fib6_nh->fib_nh_dev == down_dev ||
4814         rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4815         dead++;
4816     list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4817         if (iter->fib6_nh->fib_nh_dev == down_dev ||
4818             iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4819             dead++;
4820
4821     return dead;
4822 }
4823
4824 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4825                        const struct net_device *dev,
4826                        unsigned char nh_flags)
4827 {
4828     struct fib6_info *iter;
4829
4830     if (rt->fib6_nh->fib_nh_dev == dev)
4831         rt->fib6_nh->fib_nh_flags |= nh_flags;
4832     list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4833         if (iter->fib6_nh->fib_nh_dev == dev)
4834             iter->fib6_nh->fib_nh_flags |= nh_flags;
4835 }
4836
4837 /* called with write lock held for table with rt */
4838 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4839 {
4840     const struct arg_netdev_event *arg = p_arg;
4841     const struct net_device *dev = arg->dev;
4842     struct net *net = dev_net(dev);
4843
4844     if (rt == net->ipv6.fib6_null_entry || rt->nh)
4845         return 0;
4846
4847     switch (arg->event) {
4848     case NETDEV_UNREGISTER:
4849         return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4850     case NETDEV_DOWN:
4851         if (rt->should_flush)
4852             return -1;
4853         if (!rt->fib6_nsiblings)
4854             return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4855         if (rt6_multipath_uses_dev(rt, dev)) {
4856             unsigned int count;
4857
4858             count = rt6_multipath_dead_count(rt, dev);
4859             if (rt->fib6_nsiblings + 1 == count) {
4860                 rt6_multipath_flush(rt);
4861                 return -1;
4862             }
4863             rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4864                            RTNH_F_LINKDOWN);
4865             fib6_update_sernum(net, rt);
4866             rt6_multipath_rebalance(rt);
4867         }
4868         return -2;
4869     case NETDEV_CHANGE:
4870         if (rt->fib6_nh->fib_nh_dev != dev ||
4871             rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4872             break;
4873         rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4874         rt6_multipath_rebalance(rt);
4875         break;
4876     }
4877
4878     return 0;
4879 }
4880
4881 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4882 {
4883     struct arg_netdev_event arg = {
4884         .dev = dev,
4885         {
4886             .event = event,
4887         },
4888     };
4889     struct net *net = dev_net(dev);
4890
4891     if (net->ipv6.sysctl.skip_notify_on_dev_down)
4892         fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4893     else
4894         fib6_clean_all(net, fib6_ifdown, &arg);
4895 }
4896
4897 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4898 {
4899     rt6_sync_down_dev(dev, event);
4900     rt6_uncached_list_flush_dev(dev);
4901     neigh_ifdown(&nd_tbl, dev);
4902 }
4903
4904 struct rt6_mtu_change_arg {
4905     struct net_device *dev;
4906     unsigned int mtu;
4907     struct fib6_info *f6i;
4908 };
4909
4910 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4911 {
4912     struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4913     struct fib6_info *f6i = arg->f6i;
4914
4915     /* For administrative MTU increase, there is no way to discover
4916      * IPv6 PMTU increase, so PMTU increase should be updated here.
4917      * Since RFC 1981 doesn't include administrative MTU increase
4918      * update PMTU increase is a MUST. (i.e. jumbo frame)
4919      */
4920     if (nh->fib_nh_dev == arg->dev) {
4921         struct inet6_dev *idev = __in6_dev_get(arg->dev);
4922         u32 mtu = f6i->fib6_pmtu;
4923
4924         if (mtu >= arg->mtu ||
4925             (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4926             fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4927
4928         spin_lock_bh(&rt6_exception_lock);
4929         rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4930         spin_unlock_bh(&rt6_exception_lock);
4931     }
4932
4933     return 0;
4934 }
4935
4936 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4937 {
4938     struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4939     struct inet6_dev *idev;
4940
4941     /* In IPv6 pmtu discovery is not optional,
4942        so that RTAX_MTU lock cannot disable it.
4943        We still use this lock to block changes
4944        caused by addrconf/ndisc.
4945     */
4946
4947     idev = __in6_dev_get(arg->dev);
4948     if (!idev)
4949         return 0;
4950
4951     if (fib6_metric_locked(f6i, RTAX_MTU))
4952         return 0;
4953
4954     arg->f6i = f6i;
4955     if (f6i->nh) {
4956         /* fib6_nh_mtu_change only returns 0, so this is safe */
4957         return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4958                         arg);
4959     }
4960
4961     return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4962 }
4963
4964 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4965 {
4966     struct rt6_mtu_change_arg arg = {
4967         .dev = dev,
4968         .mtu = mtu,
4969     };
4970
4971     fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4972 }
4973
4974 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4975     [RTA_UNSPEC]        = { .strict_start_type = RTA_DPORT + 1 },
4976     [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4977     [RTA_PREFSRC]       = { .len = sizeof(struct in6_addr) },
4978     [RTA_OIF]               = { .type = NLA_U32 },
4979     [RTA_IIF]       = { .type = NLA_U32 },
4980     [RTA_PRIORITY]          = { .type = NLA_U32 },
4981     [RTA_METRICS]           = { .type = NLA_NESTED },
4982     [RTA_MULTIPATH]     = { .len = sizeof(struct rtnexthop) },
4983     [RTA_PREF]              = { .type = NLA_U8 },
4984     [RTA_ENCAP_TYPE]    = { .type = NLA_U16 },
4985     [RTA_ENCAP]     = { .type = NLA_NESTED },
4986     [RTA_EXPIRES]       = { .type = NLA_U32 },
4987     [RTA_UID]       = { .type = NLA_U32 },
4988     [RTA_MARK]      = { .type = NLA_U32 },
4989     [RTA_TABLE]     = { .type = NLA_U32 },
4990     [RTA_IP_PROTO]      = { .type = NLA_U8 },
4991     [RTA_SPORT]     = { .type = NLA_U16 },
4992     [RTA_DPORT]     = { .type = NLA_U16 },
4993     [RTA_NH_ID]     = { .type = NLA_U32 },
4994 };
4995
4996 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4997                   struct fib6_config *cfg,
4998                   struct netlink_ext_ack *extack)
4999 {
5000     struct rtmsg *rtm;
5001     struct nlattr *tb[RTA_MAX+1];
5002     unsigned int pref;
5003     int err;
5004
5005     err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5006                      rtm_ipv6_policy, extack);
5007     if (err < 0)
5008         goto errout;
5009
5010     err = -EINVAL;
5011     rtm = nlmsg_data(nlh);
5012
5013     if (rtm->rtm_tos) {
5014         NL_SET_ERR_MSG(extack,
5015                    "Invalid dsfield (tos): option not available for IPv6");
5016         goto errout;
5017     }
5018
5019     *cfg = (struct fib6_config){
5020         .fc_table = rtm->rtm_table,
5021         .fc_dst_len = rtm->rtm_dst_len,
5022         .fc_src_len = rtm->rtm_src_len,
5023         .fc_flags = RTF_UP,
5024         .fc_protocol = rtm->rtm_protocol,
5025         .fc_type = rtm->rtm_type,
5026
5027         .fc_nlinfo.portid = NETLINK_CB(skb).portid,
5028         .fc_nlinfo.nlh = nlh,
5029         .fc_nlinfo.nl_net = sock_net(skb->sk),
5030     };
5031
5032     if (rtm->rtm_type == RTN_UNREACHABLE ||
5033         rtm->rtm_type == RTN_BLACKHOLE ||
5034         rtm->rtm_type == RTN_PROHIBIT ||
5035         rtm->rtm_type == RTN_THROW)
5036         cfg->fc_flags |= RTF_REJECT;
5037
5038     if (rtm->rtm_type == RTN_LOCAL)
5039         cfg->fc_flags |= RTF_LOCAL;
5040
5041     if (rtm->rtm_flags & RTM_F_CLONED)
5042         cfg->fc_flags |= RTF_CACHE;
5043
5044     cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5045
5046     if (tb[RTA_NH_ID]) {
5047         if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
5048             tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5049             NL_SET_ERR_MSG(extack,
5050                        "Nexthop specification and nexthop id are mutually exclusive");
5051             goto errout;
5052         }
5053         cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5054     }
5055
5056     if (tb[RTA_GATEWAY]) {
5057         cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5058         cfg->fc_flags |= RTF_GATEWAY;
5059     }
5060     if (tb[RTA_VIA]) {
5061         NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5062         goto errout;
5063     }
5064
5065     if (tb[RTA_DST]) {
5066         int plen = (rtm->rtm_dst_len + 7) >> 3;
5067
5068         if (nla_len(tb[RTA_DST]) < plen)
5069             goto errout;
5070
5071         nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5072     }
5073
5074     if (tb[RTA_SRC]) {
5075         int plen = (rtm->rtm_src_len + 7) >> 3;
5076
5077         if (nla_len(tb[RTA_SRC]) < plen)
5078             goto errout;
5079
5080         nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5081     }
5082
5083     if (tb[RTA_PREFSRC])
5084         cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5085
5086     if (tb[RTA_OIF])
5087         cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5088
5089     if (tb[RTA_PRIORITY])
5090         cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5091
5092     if (tb[RTA_METRICS]) {
5093         cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5094         cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5095     }
5096
5097     if (tb[RTA_TABLE])
5098         cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5099
5100     if (tb[RTA_MULTIPATH]) {
5101         cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5102         cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5103
5104         err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5105                              cfg->fc_mp_len, extack);
5106         if (err < 0)
5107             goto errout;
5108     }
5109
5110     if (tb[RTA_PREF]) {
5111         pref = nla_get_u8(tb[RTA_PREF]);
5112         if (pref != ICMPV6_ROUTER_PREF_LOW &&
5113             pref != ICMPV6_ROUTER_PREF_HIGH)
5114             pref = ICMPV6_ROUTER_PREF_MEDIUM;
5115         cfg->fc_flags |= RTF_PREF(pref);
5116     }
5117
5118     if (tb[RTA_ENCAP])
5119         cfg->fc_encap = tb[RTA_ENCAP];
5120
5121     if (tb[RTA_ENCAP_TYPE]) {
5122         cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5123
5124         err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5125         if (err < 0)
5126             goto errout;
5127     }
5128
5129     if (tb[RTA_EXPIRES]) {
5130         unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5131
5132         if (addrconf_finite_timeout(timeout)) {
5133             cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5134             cfg->fc_flags |= RTF_EXPIRES;
5135         }
5136     }
5137
5138     err = 0;
5139 errout:
5140     return err;
5141 }
5142
5143 struct rt6_nh {
5144     struct fib6_info *fib6_info;
5145     struct fib6_config r_cfg;
5146     struct list_head next;
5147 };
5148
5149 static int ip6_route_info_append(struct net *net,
5150                  struct list_head *rt6_nh_list,
5151                  struct fib6_info *rt,
5152                  struct fib6_config *r_cfg)
5153 {
5154     struct rt6_nh *nh;
5155     int err = -EEXIST;
5156
5157     list_for_each_entry(nh, rt6_nh_list, next) {
5158         /* check if fib6_info already exists */
5159         if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5160             return err;
5161     }
5162
5163     nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5164     if (!nh)
5165         return -ENOMEM;
5166     nh->fib6_info = rt;
5167     memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5168     list_add_tail(&nh->next, rt6_nh_list);
5169
5170     return 0;
5171 }
5172
5173 static void ip6_route_mpath_notify(struct fib6_info *rt,
5174                    struct fib6_info *rt_last,
5175                    struct nl_info *info,
5176                    __u16 nlflags)
5177 {
5178     /* if this is an APPEND route, then rt points to the first route
5179      * inserted and rt_last points to last route inserted. Userspace
5180      * wants a consistent dump of the route which starts at the first
5181      * nexthop. Since sibling routes are always added at the end of
5182      * the list, find the first sibling of the last route appended
5183      */
5184     if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5185         rt = list_first_entry(&rt_last->fib6_siblings,
5186                       struct fib6_info,
5187                       fib6_siblings);
5188     }
5189
5190     if (rt)
5191         inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5192 }
5193
5194 static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5195 {
5196     bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5197     bool should_notify = false;
5198     struct fib6_info *leaf;
5199     struct fib6_node *fn;
5200
5201     rcu_read_lock();
5202     fn = rcu_dereference(rt->fib6_node);
5203     if (!fn)
5204         goto out;
5205
5206     leaf = rcu_dereference(fn->leaf);
5207     if (!leaf)
5208         goto out;
5209
5210     if (rt == leaf ||
5211         (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5212          rt6_qualify_for_ecmp(leaf)))
5213         should_notify = true;
5214 out:
5215     rcu_read_unlock();
5216
5217     return should_notify;
5218 }
5219
5220 static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5221                  struct netlink_ext_ack *extack)
5222 {
5223     if (nla_len(nla) < sizeof(*gw)) {
5224         NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5225         return -EINVAL;
5226     }
5227
5228     *gw = nla_get_in6_addr(nla);
5229
5230     return 0;
5231 }
5232
5233 static int ip6_route_multipath_add(struct fib6_config *cfg,
5234                    struct netlink_ext_ack *extack)
5235 {
5236     struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5237     struct nl_info *info = &cfg->fc_nlinfo;
5238     struct fib6_config r_cfg;
5239     struct rtnexthop *rtnh;
5240     struct fib6_info *rt;
5241     struct rt6_nh *err_nh;
5242     struct rt6_nh *nh, *nh_safe;
5243     __u16 nlflags;
5244     int remaining;
5245     int attrlen;
5246     int err = 1;
5247     int nhn = 0;
5248     int replace = (cfg->fc_nlinfo.nlh &&
5249                (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5250     LIST_HEAD(rt6_nh_list);
5251
5252     nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5253     if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5254         nlflags |= NLM_F_APPEND;
5255
5256     remaining = cfg->fc_mp_len;
5257     rtnh = (struct rtnexthop *)cfg->fc_mp;
5258
5259     /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5260      * fib6_info structs per nexthop
5261      */
5262     while (rtnh_ok(rtnh, remaining)) {
5263         memcpy(&r_cfg, cfg, sizeof(*cfg));
5264         if (rtnh->rtnh_ifindex)
5265             r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5266
5267         attrlen = rtnh_attrlen(rtnh);
5268         if (attrlen > 0) {
5269             struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5270
5271             nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5272             if (nla) {
5273                 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5274                             extack);
5275                 if (err)
5276                     goto cleanup;
5277
5278                 r_cfg.fc_flags |= RTF_GATEWAY;
5279             }
5280             r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5281
5282             /* RTA_ENCAP_TYPE length checked in
5283              * lwtunnel_valid_encap_type_attr
5284              */
5285             nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5286             if (nla)
5287                 r_cfg.fc_encap_type = nla_get_u16(nla);
5288         }
5289
5290         r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5291         rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5292         if (IS_ERR(rt)) {
5293             err = PTR_ERR(rt);
5294             rt = NULL;
5295             goto cleanup;
5296         }
5297         if (!rt6_qualify_for_ecmp(rt)) {
5298             err = -EINVAL;
5299             NL_SET_ERR_MSG(extack,
5300                        "Device only routes can not be added for IPv6 using the multipath API.");
5301             fib6_info_release(rt);
5302             goto cleanup;
5303         }
5304
5305         rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5306
5307         err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5308                         rt, &r_cfg);
5309         if (err) {
5310             fib6_info_release(rt);
5311             goto cleanup;
5312         }
5313
5314         rtnh = rtnh_next(rtnh, &remaining);
5315     }
5316
5317     if (list_empty(&rt6_nh_list)) {
5318         NL_SET_ERR_MSG(extack,
5319                    "Invalid nexthop configuration - no valid nexthops");
5320         return -EINVAL;
5321     }
5322
5323     /* for add and replace send one notification with all nexthops.
5324      * Skip the notification in fib6_add_rt2node and send one with
5325      * the full route when done
5326      */
5327     info->skip_notify = 1;
5328
5329     /* For add and replace, send one notification with all nexthops. For
5330      * append, send one notification with all appended nexthops.
5331      */
5332     info->skip_notify_kernel = 1;
5333
5334     err_nh = NULL;
5335     list_for_each_entry(nh, &rt6_nh_list, next) {
5336         err = __ip6_ins_rt(nh->fib6_info, info, extack);
5337         fib6_info_release(nh->fib6_info);
5338
5339         if (!err) {
5340             /* save reference to last route successfully inserted */
5341             rt_last = nh->fib6_info;
5342
5343             /* save reference to first route for notification */
5344             if (!rt_notif)
5345                 rt_notif = nh->fib6_info;
5346         }
5347
5348         /* nh->fib6_info is used or freed at this point, reset to NULL*/
5349         nh->fib6_info = NULL;
5350         if (err) {
5351             if (replace && nhn)
5352                 NL_SET_ERR_MSG_MOD(extack,
5353                            "multipath route replace failed (check consistency of installed routes)");
5354             err_nh = nh;
5355             goto add_errout;
5356         }
5357
5358         /* Because each route is added like a single route we remove
5359          * these flags after the first nexthop: if there is a collision,
5360          * we have already failed to add the first nexthop:
5361          * fib6_add_rt2node() has rejected it; when replacing, old
5362          * nexthops have been replaced by first new, the rest should
5363          * be added to it.
5364          */
5365         if (cfg->fc_nlinfo.nlh) {
5366             cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5367                                  NLM_F_REPLACE);
5368             cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5369         }
5370         nhn++;
5371     }
5372
5373     /* An in-kernel notification should only be sent in case the new
5374      * multipath route is added as the first route in the node, or if
5375      * it was appended to it. We pass 'rt_notif' since it is the first
5376      * sibling and might allow us to skip some checks in the replace case.
5377      */
5378     if (ip6_route_mpath_should_notify(rt_notif)) {
5379         enum fib_event_type fib_event;
5380
5381         if (rt_notif->fib6_nsiblings != nhn - 1)
5382             fib_event = FIB_EVENT_ENTRY_APPEND;
5383         else
5384             fib_event = FIB_EVENT_ENTRY_REPLACE;
5385
5386         err = call_fib6_multipath_entry_notifiers(info->nl_net,
5387                               fib_event, rt_notif,
5388                               nhn - 1, extack);
5389         if (err) {
5390             /* Delete all the siblings that were just added */
5391             err_nh = NULL;
5392             goto add_errout;
5393         }
5394     }
5395
5396     /* success ... tell user about new route */
5397     ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5398     goto cleanup;
5399
5400 add_errout:
5401     /* send notification for routes that were added so that
5402      * the delete notifications sent by ip6_route_del are
5403      * coherent
5404      */
5405     if (rt_notif)
5406         ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5407
5408     /* Delete routes that were already added */
5409     list_for_each_entry(nh, &rt6_nh_list, next) {
5410         if (err_nh == nh)
5411             break;
5412         ip6_route_del(&nh->r_cfg, extack);
5413     }
5414
5415 cleanup:
5416     list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5417         if (nh->fib6_info)
5418             fib6_info_release(nh->fib6_info);
5419         list_del(&nh->next);
5420         kfree(nh);
5421     }
5422
5423     return err;
5424 }
5425
5426 static int ip6_route_multipath_del(struct fib6_config *cfg,
5427                    struct netlink_ext_ack *extack)
5428 {
5429     struct fib6_config r_cfg;
5430     struct rtnexthop *rtnh;
5431     int last_err = 0;
5432     int remaining;
5433     int attrlen;
5434     int err;
5435
5436     remaining = cfg->fc_mp_len;
5437     rtnh = (struct rtnexthop *)cfg->fc_mp;
5438
5439     /* Parse a Multipath Entry */
5440     while (rtnh_ok(rtnh, remaining)) {
5441         memcpy(&r_cfg, cfg, sizeof(*cfg));
5442         if (rtnh->rtnh_ifindex)
5443             r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5444
5445         attrlen = rtnh_attrlen(rtnh);
5446         if (attrlen > 0) {
5447             struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5448
5449             nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5450             if (nla) {
5451                 err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5452                             extack);
5453                 if (err) {
5454                     last_err = err;
5455                     goto next_rtnh;
5456                 }
5457
5458                 r_cfg.fc_flags |= RTF_GATEWAY;
5459             }
5460         }
5461         err = ip6_route_del(&r_cfg, extack);
5462         if (err)
5463             last_err = err;
5464
5465 next_rtnh:
5466         rtnh = rtnh_next(rtnh, &remaining);
5467     }
5468
5469     return last_err;
5470 }
5471
5472 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5473                   struct netlink_ext_ack *extack)
5474 {
5475     struct fib6_config cfg;
5476     int err;
5477
5478     err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5479     if (err < 0)
5480         return err;
5481
5482     if (cfg.fc_nh_id &&
5483         !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5484         NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5485         return -EINVAL;
5486     }
5487
5488     if (cfg.fc_mp)
5489         return ip6_route_multipath_del(&cfg, extack);
5490     else {
5491         cfg.fc_delete_all_nh = 1;
5492         return ip6_route_del(&cfg, extack);
5493     }
5494 }
5495
5496 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5497                   struct netlink_ext_ack *extack)
5498 {
5499     struct fib6_config cfg;
5500     int err;
5501
5502     err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5503     if (err < 0)
5504         return err;
5505
5506     if (cfg.fc_metric == 0)
5507         cfg.fc_metric = IP6_RT_PRIO_USER;
5508
5509     if (cfg.fc_mp)
5510         return ip6_route_multipath_add(&cfg, extack);
5511     else
5512         return ip6_route_add(&cfg, GFP_KERNEL, extack);
5513 }
5514
5515 /* add the overhead of this fib6_nh to nexthop_len */
5516 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5517 {
5518     int *nexthop_len = arg;
5519
5520     *nexthop_len += nla_total_size(0)    /* RTA_MULTIPATH */
5521              + NLA_ALIGN(sizeof(struct rtnexthop))
5522              + nla_total_size(16); /* RTA_GATEWAY */
5523
5524     if (nh->fib_nh_lws) {
5525         /* RTA_ENCAP_TYPE */
5526         *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5527         /* RTA_ENCAP */
5528         *nexthop_len += nla_total_size(2);
5529     }
5530
5531     return 0;
5532 }
5533
5534 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5535 {
5536     int nexthop_len;
5537
5538     if (f6i->nh) {
5539         nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5540         nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5541                      &nexthop_len);
5542     } else {
5543         struct fib6_nh *nh = f6i->fib6_nh;
5544
5545         nexthop_len = 0;
5546         if (f6i->fib6_nsiblings) {
5547             nexthop_len = nla_total_size(0)  /* RTA_MULTIPATH */
5548                     + NLA_ALIGN(sizeof(struct rtnexthop))
5549                     + nla_total_size(16) /* RTA_GATEWAY */
5550                     + lwtunnel_get_encap_size(nh->fib_nh_lws);
5551
5552             nexthop_len *= f6i->fib6_nsiblings;
5553         }
5554         nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5555     }
5556
5557     return NLMSG_ALIGN(sizeof(struct rtmsg))
5558            + nla_total_size(16) /* RTA_SRC */
5559            + nla_total_size(16) /* RTA_DST */
5560            + nla_total_size(16) /* RTA_GATEWAY */
5561            + nla_total_size(16) /* RTA_PREFSRC */
5562            + nla_total_size(4) /* RTA_TABLE */
5563            + nla_total_size(4) /* RTA_IIF */
5564            + nla_total_size(4) /* RTA_OIF */
5565            + nla_total_size(4) /* RTA_PRIORITY */
5566            + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5567            + nla_total_size(sizeof(struct rta_cacheinfo))
5568            + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5569            + nla_total_size(1) /* RTA_PREF */
5570            + nexthop_len;
5571 }
5572
5573 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5574                  unsigned char *flags)
5575 {
5576     if (nexthop_is_multipath(nh)) {
5577         struct nlattr *mp;
5578
5579         mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5580         if (!mp)
5581             goto nla_put_failure;
5582
5583         if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5584             goto nla_put_failure;
5585
5586         nla_nest_end(skb, mp);
5587     } else {
5588         struct fib6_nh *fib6_nh;
5589
5590         fib6_nh = nexthop_fib6_nh(nh);
5591         if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5592                      flags, false) < 0)
5593             goto nla_put_failure;
5594     }
5595
5596     return 0;
5597
5598 nla_put_failure:
5599     return -EMSGSIZE;
5600 }
5601
5602 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5603              struct fib6_info *rt, struct dst_entry *dst,
5604              struct in6_addr *dest, struct in6_addr *src,
5605              int iif, int type, u32 portid, u32 seq,
5606              unsigned int flags)
5607 {
5608     struct rt6_info *rt6 = (struct rt6_info *)dst;
5609     struct rt6key *rt6_dst, *rt6_src;
5610     u32 *pmetrics, table, rt6_flags;
5611     unsigned char nh_flags = 0;
5612     struct nlmsghdr *nlh;
5613     struct rtmsg *rtm;
5614     long expires = 0;
5615
5616     nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5617     if (!nlh)
5618         return -EMSGSIZE;
5619
5620     if (rt6) {
5621         rt6_dst = &rt6->rt6i_dst;
5622         rt6_src = &rt6->rt6i_src;
5623         rt6_flags = rt6->rt6i_flags;
5624     } else {
5625         rt6_dst = &rt->fib6_dst;
5626         rt6_src = &rt->fib6_src;
5627         rt6_flags = rt->fib6_flags;
5628     }
5629
5630     rtm = nlmsg_data(nlh);
5631     rtm->rtm_family = AF_INET6;
5632     rtm->rtm_dst_len = rt6_dst->plen;
5633     rtm->rtm_src_len = rt6_src->plen;
5634     rtm->rtm_tos = 0;
5635     if (rt->fib6_table)
5636         table = rt->fib6_table->tb6_id;
5637     else
5638         table = RT6_TABLE_UNSPEC;
5639     rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5640     if (nla_put_u32(skb, RTA_TABLE, table))
5641         goto nla_put_failure;
5642
5643     rtm->rtm_type = rt->fib6_type;
5644     rtm->rtm_flags = 0;
5645     rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5646     rtm->rtm_protocol = rt->fib6_protocol;
5647
5648     if (rt6_flags & RTF_CACHE)
5649         rtm->rtm_flags |= RTM_F_CLONED;
5650
5651     if (dest) {
5652         if (nla_put_in6_addr(skb, RTA_DST, dest))
5653             goto nla_put_failure;
5654         rtm->rtm_dst_len = 128;
5655     } else if (rtm->rtm_dst_len)
5656         if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5657             goto nla_put_failure;
5658 #ifdef CONFIG_IPV6_SUBTREES
5659     if (src) {
5660         if (nla_put_in6_addr(skb, RTA_SRC, src))
5661             goto nla_put_failure;
5662         rtm->rtm_src_len = 128;
5663     } else if (rtm->rtm_src_len &&
5664            nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5665         goto nla_put_failure;
5666 #endif
5667     if (iif) {
5668 #ifdef CONFIG_IPV6_MROUTE
5669         if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5670             int err = ip6mr_get_route(net, skb, rtm, portid);
5671
5672             if (err == 0)
5673                 return 0;
5674             if (err < 0)
5675                 goto nla_put_failure;
5676         } else
5677 #endif
5678             if (nla_put_u32(skb, RTA_IIF, iif))
5679                 goto nla_put_failure;
5680     } else if (dest) {
5681         struct in6_addr saddr_buf;
5682         if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5683             nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5684             goto nla_put_failure;
5685     }
5686
5687     if (rt->fib6_prefsrc.plen) {
5688         struct in6_addr saddr_buf;
5689         saddr_buf = rt->fib6_prefsrc.addr;
5690         if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5691             goto nla_put_failure;
5692     }
5693
5694     pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5695     if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5696         goto nla_put_failure;
5697
5698     if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5699         goto nla_put_failure;
5700
5701     /* For multipath routes, walk the siblings list and add
5702      * each as a nexthop within RTA_MULTIPATH.
5703      */
5704     if (rt6) {
5705         if (rt6_flags & RTF_GATEWAY &&
5706             nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5707             goto nla_put_failure;
5708
5709         if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5710             goto nla_put_failure;
5711
5712         if (dst->lwtstate &&
5713             lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5714             goto nla_put_failure;
5715     } else if (rt->fib6_nsiblings) {
5716         struct fib6_info *sibling, *next_sibling;
5717         struct nlattr *mp;
5718
5719         mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5720         if (!mp)
5721             goto nla_put_failure;
5722
5723         if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5724                     rt->fib6_nh->fib_nh_weight, AF_INET6,
5725                     0) < 0)
5726             goto nla_put_failure;
5727
5728         list_for_each_entry_safe(sibling, next_sibling,
5729                      &rt->fib6_siblings, fib6_siblings) {
5730             if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5731                         sibling->fib6_nh->fib_nh_weight,
5732                         AF_INET6, 0) < 0)
5733                 goto nla_put_failure;
5734         }
5735
5736         nla_nest_end(skb, mp);
5737     } else if (rt->nh) {
5738         if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5739             goto nla_put_failure;
5740
5741         if (nexthop_is_blackhole(rt->nh))
5742             rtm->rtm_type = RTN_BLACKHOLE;
5743
5744         if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5745             rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5746             goto nla_put_failure;
5747
5748         rtm->rtm_flags |= nh_flags;
5749     } else {
5750         if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5751                      &nh_flags, false) < 0)
5752             goto nla_put_failure;
5753
5754         rtm->rtm_flags |= nh_flags;
5755     }
5756
5757     if (rt6_flags & RTF_EXPIRES) {
5758         expires = dst ? dst->expires : rt->expires;
5759         expires -= jiffies;
5760     }
5761
5762     if (!dst) {
5763         if (READ_ONCE(rt->offload))
5764             rtm->rtm_flags |= RTM_F_OFFLOAD;
5765         if (READ_ONCE(rt->trap))
5766             rtm->rtm_flags |= RTM_F_TRAP;
5767         if (READ_ONCE(rt->offload_failed))
5768             rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5769     }
5770
5771     if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5772         goto nla_put_failure;
5773
5774     if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5775         goto nla_put_failure;
5776
5777
5778     nlmsg_end(skb, nlh);
5779     return 0;
5780
5781 nla_put_failure:
5782     nlmsg_cancel(skb, nlh);
5783     return -EMSGSIZE;
5784 }
5785
5786 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5787 {
5788     const struct net_device *dev = arg;
5789
5790     if (nh->fib_nh_dev == dev)
5791         return 1;
5792
5793     return 0;
5794 }
5795
5796 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5797                    const struct net_device *dev)
5798 {
5799     if (f6i->nh) {
5800         struct net_device *_dev = (struct net_device *)dev;
5801
5802         return !!nexthop_for_each_fib6_nh(f6i->nh,
5803                           fib6_info_nh_uses_dev,
5804                           _dev);
5805     }
5806
5807     if (f6i->fib6_nh->fib_nh_dev == dev)
5808         return true;
5809
5810     if (f6i->fib6_nsiblings) {
5811         struct fib6_info *sibling, *next_sibling;
5812
5813         list_for_each_entry_safe(sibling, next_sibling,
5814                      &f6i->fib6_siblings, fib6_siblings) {
5815             if (sibling->fib6_nh->fib_nh_dev == dev)
5816                 return true;
5817         }
5818     }
5819
5820     return false;
5821 }
5822
5823 struct fib6_nh_exception_dump_walker {
5824     struct rt6_rtnl_dump_arg *dump;
5825     struct fib6_info *rt;
5826     unsigned int flags;
5827     unsigned int skip;
5828     unsigned int count;
5829 };
5830
5831 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5832 {
5833     struct fib6_nh_exception_dump_walker *w = arg;
5834     struct rt6_rtnl_dump_arg *dump = w->dump;
5835     struct rt6_exception_bucket *bucket;
5836     struct rt6_exception *rt6_ex;
5837     int i, err;
5838
5839     bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5840     if (!bucket)
5841         return 0;
5842
5843     for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5844         hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5845             if (w->skip) {
5846                 w->skip--;
5847                 continue;
5848             }
5849
5850             /* Expiration of entries doesn't bump sernum, insertion
5851              * does. Removal is triggered by insertion, so we can
5852              * rely on the fact that if entries change between two
5853              * partial dumps, this node is scanned again completely,
5854              * see rt6_insert_exception() and fib6_dump_table().
5855              *
5856              * Count expired entries we go through as handled
5857              * entries that we'll skip next time, in case of partial
5858              * node dump. Otherwise, if entries expire meanwhile,
5859              * we'll skip the wrong amount.
5860              */
5861             if (rt6_check_expired(rt6_ex->rt6i)) {
5862                 w->count++;
5863                 continue;
5864             }
5865
5866             err = rt6_fill_node(dump->net, dump->skb, w->rt,
5867                         &rt6_ex->rt6i->dst, NULL, NULL, 0,
5868                         RTM_NEWROUTE,
5869                         NETLINK_CB(dump->cb->skb).portid,
5870                         dump->cb->nlh->nlmsg_seq, w->flags);
5871             if (err)
5872                 return err;
5873
5874             w->count++;
5875         }
5876         bucket++;
5877     }
5878
5879     return 0;
5880 }
5881
5882 /* Return -1 if done with node, number of handled routes on partial dump */
5883 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5884 {
5885     struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5886     struct fib_dump_filter *filter = &arg->filter;
5887     unsigned int flags = NLM_F_MULTI;
5888     struct net *net = arg->net;
5889     int count = 0;
5890
5891     if (rt == net->ipv6.fib6_null_entry)
5892         return -1;
5893
5894     if ((filter->flags & RTM_F_PREFIX) &&
5895         !(rt->fib6_flags & RTF_PREFIX_RT)) {
5896         /* success since this is not a prefix route */
5897         return -1;
5898     }
5899     if (filter->filter_set &&
5900         ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5901          (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5902          (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5903         return -1;
5904     }
5905
5906     if (filter->filter_set ||
5907         !filter->dump_routes || !filter->dump_exceptions) {
5908         flags |= NLM_F_DUMP_FILTERED;
5909     }
5910
5911     if (filter->dump_routes) {
5912         if (skip) {
5913             skip--;
5914         } else {
5915             if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5916                       0, RTM_NEWROUTE,
5917                       NETLINK_CB(arg->cb->skb).portid,
5918                       arg->cb->nlh->nlmsg_seq, flags)) {
5919                 return 0;
5920             }
5921             count++;
5922         }
5923     }
5924
5925     if (filter->dump_exceptions) {
5926         struct fib6_nh_exception_dump_walker w = { .dump = arg,
5927                                .rt = rt,
5928                                .flags = flags,
5929                                .skip = skip,
5930                                .count = 0 };
5931         int err;
5932
5933         rcu_read_lock();
5934         if (rt->nh) {
5935             err = nexthop_for_each_fib6_nh(rt->nh,
5936                                rt6_nh_dump_exceptions,
5937                                &w);
5938         } else {
5939             err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5940         }
5941         rcu_read_unlock();
5942
5943         if (err)
5944             return count + w.count;
5945     }
5946
5947     return -1;
5948 }
5949
5950 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5951                     const struct nlmsghdr *nlh,
5952                     struct nlattr **tb,
5953                     struct netlink_ext_ack *extack)
5954 {
5955     struct rtmsg *rtm;
5956     int i, err;
5957
5958     if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5959         NL_SET_ERR_MSG_MOD(extack,
5960                    "Invalid header for get route request");
5961         return -EINVAL;
5962     }
5963
5964     if (!netlink_strict_get_check(skb))
5965         return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5966                           rtm_ipv6_policy, extack);
5967
5968     rtm = nlmsg_data(nlh);
5969     if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5970         (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5971         rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5972         rtm->rtm_type) {
5973         NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5974         return -EINVAL;
5975     }
5976     if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5977         NL_SET_ERR_MSG_MOD(extack,
5978                    "Invalid flags for get route request");
5979         return -EINVAL;
5980     }
5981
5982     err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5983                         rtm_ipv6_policy, extack);
5984     if (err)
5985         return err;
5986
5987     if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5988         (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5989         NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5990         return -EINVAL;
5991     }
5992
5993     for (i = 0; i <= RTA_MAX; i++) {
5994         if (!tb[i])
5995             continue;
5996
5997         switch (i) {
5998         case RTA_SRC:
5999         case RTA_DST:
6000         case RTA_IIF:
6001         case RTA_OIF:
6002         case RTA_MARK:
6003         case RTA_UID:
6004         case RTA_SPORT:
6005         case RTA_DPORT:
6006         case RTA_IP_PROTO:
6007             break;
6008         default:
6009             NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6010             return -EINVAL;
6011         }
6012     }
6013
6014     return 0;
6015 }
6016
6017 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6018                   struct netlink_ext_ack *extack)
6019 {
6020     struct net *net = sock_net(in_skb->sk);
6021     struct nlattr *tb[RTA_MAX+1];
6022     int err, iif = 0, oif = 0;
6023     struct fib6_info *from;
6024     struct dst_entry *dst;
6025     struct rt6_info *rt;
6026     struct sk_buff *skb;
6027     struct rtmsg *rtm;
6028     struct flowi6 fl6 = {};
6029     bool fibmatch;
6030
6031     err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6032     if (err < 0)
6033         goto errout;
6034
6035     err = -EINVAL;
6036     rtm = nlmsg_data(nlh);
6037     fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6038     fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6039
6040     if (tb[RTA_SRC]) {
6041         if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6042             goto errout;
6043
6044         fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6045     }
6046
6047     if (tb[RTA_DST]) {
6048         if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6049             goto errout;
6050
6051         fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6052     }
6053
6054     if (tb[RTA_IIF])
6055         iif = nla_get_u32(tb[RTA_IIF]);
6056
6057     if (tb[RTA_OIF])
6058         oif = nla_get_u32(tb[RTA_OIF]);
6059
6060     if (tb[RTA_MARK])
6061         fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6062
6063     if (tb[RTA_UID])
6064         fl6.flowi6_uid = make_kuid(current_user_ns(),
6065                        nla_get_u32(tb[RTA_UID]));
6066     else
6067         fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6068
6069     if (tb[RTA_SPORT])
6070         fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6071
6072     if (tb[RTA_DPORT])
6073         fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6074
6075     if (tb[RTA_IP_PROTO]) {
6076         err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6077                           &fl6.flowi6_proto, AF_INET6,
6078                           extack);
6079         if (err)
6080             goto errout;
6081     }
6082
6083     if (iif) {
6084         struct net_device *dev;
6085         int flags = 0;
6086
6087         rcu_read_lock();
6088
6089         dev = dev_get_by_index_rcu(net, iif);
6090         if (!dev) {
6091             rcu_read_unlock();
6092             err = -ENODEV;
6093             goto errout;
6094         }
6095
6096         fl6.flowi6_iif = iif;
6097
6098         if (!ipv6_addr_any(&fl6.saddr))
6099             flags |= RT6_LOOKUP_F_HAS_SADDR;
6100
6101         dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6102
6103         rcu_read_unlock();
6104     } else {
6105         fl6.flowi6_oif = oif;
6106
6107         dst = ip6_route_output(net, NULL, &fl6);
6108     }
6109
6110
6111     rt = container_of(dst, struct rt6_info, dst);
6112     if (rt->dst.error) {
6113         err = rt->dst.error;
6114         ip6_rt_put(rt);
6115         goto errout;
6116     }
6117
6118     if (rt == net->ipv6.ip6_null_entry) {
6119         err = rt->dst.error;
6120         ip6_rt_put(rt);
6121         goto errout;
6122     }
6123
6124     skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6125     if (!skb) {
6126         ip6_rt_put(rt);
6127         err = -ENOBUFS;
6128         goto errout;
6129     }
6130
6131     skb_dst_set(skb, &rt->dst);
6132
6133     rcu_read_lock();
6134     from = rcu_dereference(rt->from);
6135     if (from) {
6136         if (fibmatch)
6137             err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6138                         iif, RTM_NEWROUTE,
6139                         NETLINK_CB(in_skb).portid,
6140                         nlh->nlmsg_seq, 0);
6141         else
6142             err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6143                         &fl6.saddr, iif, RTM_NEWROUTE,
6144                         NETLINK_CB(in_skb).portid,
6145                         nlh->nlmsg_seq, 0);
6146     } else {
6147         err = -ENETUNREACH;
6148     }
6149     rcu_read_unlock();
6150
6151     if (err < 0) {
6152         kfree_skb(skb);
6153         goto errout;
6154     }
6155
6156     err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6157 errout:
6158     return err;
6159 }
6160
6161 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6162              unsigned int nlm_flags)
6163 {
6164     struct sk_buff *skb;
6165     struct net *net = info->nl_net;
6166     u32 seq;
6167     int err;
6168
6169     err = -ENOBUFS;
6170     seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6171
6172     skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6173     if (!skb)
6174         goto errout;
6175
6176     err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6177                 event, info->portid, seq, nlm_flags);
6178     if (err < 0) {
6179         /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6180         WARN_ON(err == -EMSGSIZE);
6181         kfree_skb(skb);
6182         goto errout;
6183     }
6184     rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6185             info->nlh, gfp_any());
6186     return;
6187 errout:
6188     if (err < 0)
6189         rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6190 }
6191
6192 void fib6_rt_update(struct net *net, struct fib6_info *rt,
6193             struct nl_info *info)
6194 {
6195     u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6196     struct sk_buff *skb;
6197     int err = -ENOBUFS;
6198
6199     skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6200     if (!skb)
6201         goto errout;
6202
6203     err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6204                 RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6205     if (err < 0) {
6206         /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6207         WARN_ON(err == -EMSGSIZE);
6208         kfree_skb(skb);
6209         goto errout;
6210     }
6211     rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6212             info->nlh, gfp_any());
6213     return;
6214 errout:
6215     if (err < 0)
6216         rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6217 }
6218
6219 void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6220                 bool offload, bool trap, bool offload_failed)
6221 {
6222     struct sk_buff *skb;
6223     int err;
6224
6225     if (READ_ONCE(f6i->offload) == offload &&
6226         READ_ONCE(f6i->trap) == trap &&
6227         READ_ONCE(f6i->offload_failed) == offload_failed)
6228         return;
6229
6230     WRITE_ONCE(f6i->offload, offload);
6231     WRITE_ONCE(f6i->trap, trap);
6232
6233     /* 2 means send notifications only if offload_failed was changed. */
6234     if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6235         READ_ONCE(f6i->offload_failed) == offload_failed)
6236         return;
6237
6238     WRITE_ONCE(f6i->offload_failed, offload_failed);
6239
6240     if (!rcu_access_pointer(f6i->fib6_node))
6241         /* The route was removed from the tree, do not send
6242          * notification.
6243          */
6244         return;
6245
6246     if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6247         return;
6248
6249     skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6250     if (!skb) {
6251         err = -ENOBUFS;
6252         goto errout;
6253     }
6254
6255     err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6256                 0, 0);
6257     if (err < 0) {
6258         /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6259         WARN_ON(err == -EMSGSIZE);
6260         kfree_skb(skb);
6261         goto errout;
6262     }
6263
6264     rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6265     return;
6266
6267 errout:
6268     rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6269 }
6270 EXPORT_SYMBOL(fib6_info_hw_flags_set);
6271
6272 static int ip6_route_dev_notify(struct notifier_block *this,
6273                 unsigned long event, void *ptr)
6274 {
6275     struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6276     struct net *net = dev_net(dev);
6277
6278     if (!(dev->flags & IFF_LOOPBACK))
6279         return NOTIFY_OK;
6280
6281     if (event == NETDEV_REGISTER) {
6282         net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6283         net->ipv6.ip6_null_entry->dst.dev = dev;
6284         net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6285 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6286         net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6287         net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6288         net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6289         net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6290 #endif
6291      } else if (event == NETDEV_UNREGISTER &&
6292             dev->reg_state != NETREG_UNREGISTERED) {
6293         /* NETDEV_UNREGISTER could be fired for multiple times by
6294          * netdev_wait_allrefs(). Make sure we only call this once.
6295          */
6296         in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6297 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6298         in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6299         in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6300 #endif
6301     }
6302
6303     return NOTIFY_OK;
6304 }
6305
6306 /*
6307  *  /proc
6308  */
6309
6310 #ifdef CONFIG_PROC_FS
6311 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6312 {
6313     struct net *net = (struct net *)seq->private;
6314     seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6315            net->ipv6.rt6_stats->fib_nodes,
6316            net->ipv6.rt6_stats->fib_route_nodes,
6317            atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6318            net->ipv6.rt6_stats->fib_rt_entries,
6319            net->ipv6.rt6_stats->fib_rt_cache,
6320            dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6321            net->ipv6.rt6_stats->fib_discarded_routes);
6322
6323     return 0;
6324 }
6325 #endif  /* CONFIG_PROC_FS */
6326
6327 #ifdef CONFIG_SYSCTL
6328
6329 static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6330                   void *buffer, size_t *lenp, loff_t *ppos)
6331 {
6332     struct net *net;
6333     int delay;
6334     int ret;
6335     if (!write)
6336         return -EINVAL;
6337
6338     net = (struct net *)ctl->extra1;
6339     delay = net->ipv6.sysctl.flush_delay;
6340     ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6341     if (ret)
6342         return ret;
6343
6344     fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6345     return 0;
6346 }
6347
6348 static struct ctl_table ipv6_route_table_template[] = {
6349     {
6350         .procname   =   "max_size",
6351         .data       =   &init_net.ipv6.sysctl.ip6_rt_max_size,
6352         .maxlen     =   sizeof(int),
6353         .mode       =   0644,
6354         .proc_handler   =   proc_dointvec,
6355     },
6356     {
6357         .procname   =   "gc_thresh",
6358         .data       =   &ip6_dst_ops_template.gc_thresh,
6359         .maxlen     =   sizeof(int),
6360         .mode       =   0644,
6361         .proc_handler   =   proc_dointvec,
6362     },
6363     {
6364         .procname   =   "flush",
6365         .data       =   &init_net.ipv6.sysctl.flush_delay,
6366         .maxlen     =   sizeof(int),
6367         .mode       =   0200,
6368         .proc_handler   =   ipv6_sysctl_rtcache_flush
6369     },
6370     {
6371         .procname   =   "gc_min_interval",
6372         .data       =   &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6373         .maxlen     =   sizeof(int),
6374         .mode       =   0644,
6375         .proc_handler   =   proc_dointvec_jiffies,
6376     },
6377     {
6378         .procname   =   "gc_timeout",
6379         .data       =   &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6380         .maxlen     =   sizeof(int),
6381         .mode       =   0644,
6382         .proc_handler   =   proc_dointvec_jiffies,
6383     },
6384     {
6385         .procname   =   "gc_interval",
6386         .data       =   &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6387         .maxlen     =   sizeof(int),
6388         .mode       =   0644,
6389         .proc_handler   =   proc_dointvec_jiffies,
6390     },
6391     {
6392         .procname   =   "gc_elasticity",
6393         .data       =   &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6394         .maxlen     =   sizeof(int),
6395         .mode       =   0644,
6396         .proc_handler   =   proc_dointvec,
6397     },
6398     {
6399         .procname   =   "mtu_expires",
6400         .data       =   &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6401         .maxlen     =   sizeof(int),
6402         .mode       =   0644,
6403         .proc_handler   =   proc_dointvec_jiffies,
6404     },
6405     {
6406         .procname   =   "min_adv_mss",
6407         .data       =   &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6408         .maxlen     =   sizeof(int),
6409         .mode       =   0644,
6410         .proc_handler   =   proc_dointvec,
6411     },
6412     {
6413         .procname   =   "gc_min_interval_ms",
6414         .data       =   &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6415         .maxlen     =   sizeof(int),
6416         .mode       =   0644,
6417         .proc_handler   =   proc_dointvec_ms_jiffies,
6418     },
6419     {
6420         .procname   =   "skip_notify_on_dev_down",
6421         .data       =   &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6422         .maxlen     =   sizeof(int),
6423         .mode       =   0644,
6424         .proc_handler   =   proc_dointvec_minmax,
6425         .extra1     =   SYSCTL_ZERO,
6426         .extra2     =   SYSCTL_ONE,
6427     },
6428     { }
6429 };
6430
6431 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6432 {
6433     struct ctl_table *table;
6434
6435     table = kmemdup(ipv6_route_table_template,
6436             sizeof(ipv6_route_table_template),
6437             GFP_KERNEL);
6438
6439     if (table) {
6440         table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6441         table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6442         table[2].data = &net->ipv6.sysctl.flush_delay;
6443         table[2].extra1 = net;
6444         table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6445         table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6446         table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6447         table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6448         table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6449         table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6450         table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6451         table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6452
6453         /* Don't export sysctls to unprivileged users */
6454         if (net->user_ns != &init_user_ns)
6455             table[1].procname = NULL;
6456     }
6457
6458     return table;
6459 }
6460 #endif
6461
6462 static int __net_init ip6_route_net_init(struct net *net)
6463 {
6464     int ret = -ENOMEM;
6465
6466     memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6467            sizeof(net->ipv6.ip6_dst_ops));
6468
6469     if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6470         goto out_ip6_dst_ops;
6471
6472     net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6473     if (!net->ipv6.fib6_null_entry)
6474         goto out_ip6_dst_entries;
6475     memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6476            sizeof(*net->ipv6.fib6_null_entry));
6477
6478     net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6479                        sizeof(*net->ipv6.ip6_null_entry),
6480                        GFP_KERNEL);
6481     if (!net->ipv6.ip6_null_entry)
6482         goto out_fib6_null_entry;
6483     net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6484     dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6485              ip6_template_metrics, true);
6486     INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6487
6488 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6489     net->ipv6.fib6_has_custom_rules = false;
6490     net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6491                            sizeof(*net->ipv6.ip6_prohibit_entry),
6492                            GFP_KERNEL);
6493     if (!net->ipv6.ip6_prohibit_entry)
6494         goto out_ip6_null_entry;
6495     net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6496     dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6497              ip6_template_metrics, true);
6498     INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6499
6500     net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6501                            sizeof(*net->ipv6.ip6_blk_hole_entry),
6502                            GFP_KERNEL);
6503     if (!net->ipv6.ip6_blk_hole_entry)
6504         goto out_ip6_prohibit_entry;
6505     net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6506     dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6507              ip6_template_metrics, true);
6508     INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6509 #ifdef CONFIG_IPV6_SUBTREES
6510     net->ipv6.fib6_routes_require_src = 0;
6511 #endif
6512 #endif
6513
6514     net->ipv6.sysctl.flush_delay = 0;
6515     net->ipv6.sysctl.ip6_rt_max_size = 4096;
6516     net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6517     net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6518     net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6519     net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6520     net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6521     net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6522     net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6523
6524     atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6525
6526     ret = 0;
6527 out:
6528     return ret;
6529
6530 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6531 out_ip6_prohibit_entry:
6532     kfree(net->ipv6.ip6_prohibit_entry);
6533 out_ip6_null_entry:
6534     kfree(net->ipv6.ip6_null_entry);
6535 #endif
6536 out_fib6_null_entry:
6537     kfree(net->ipv6.fib6_null_entry);
6538 out_ip6_dst_entries:
6539     dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6540 out_ip6_dst_ops:
6541     goto out;
6542 }
6543
6544 static void __net_exit ip6_route_net_exit(struct net *net)
6545 {
6546     kfree(net->ipv6.fib6_null_entry);
6547     kfree(net->ipv6.ip6_null_entry);
6548 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6549     kfree(net->ipv6.ip6_prohibit_entry);
6550     kfree(net->ipv6.ip6_blk_hole_entry);
6551 #endif
6552     dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6553 }
6554
6555 static int __net_init ip6_route_net_init_late(struct net *net)
6556 {
6557 #ifdef CONFIG_PROC_FS
6558     proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6559             sizeof(struct ipv6_route_iter));
6560     proc_create_net_single("rt6_stats", 0444, net->proc_net,
6561             rt6_stats_seq_show, NULL);
6562 #endif
6563     return 0;
6564 }
6565
6566 static void __net_exit ip6_route_net_exit_late(struct net *net)
6567 {
6568 #ifdef CONFIG_PROC_FS
6569     remove_proc_entry("ipv6_route", net->proc_net);
6570     remove_proc_entry("rt6_stats", net->proc_net);
6571 #endif
6572 }
6573
6574 static struct pernet_operations ip6_route_net_ops = {
6575     .init = ip6_route_net_init,
6576     .exit = ip6_route_net_exit,
6577 };
6578
6579 static int __net_init ipv6_inetpeer_init(struct net *net)
6580 {
6581     struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6582
6583     if (!bp)
6584         return -ENOMEM;
6585     inet_peer_base_init(bp);
6586     net->ipv6.peers = bp;
6587     return 0;
6588 }
6589
6590 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6591 {
6592     struct inet_peer_base *bp = net->ipv6.peers;
6593
6594     net->ipv6.peers = NULL;
6595     inetpeer_invalidate_tree(bp);
6596     kfree(bp);
6597 }
6598
6599 static struct pernet_operations ipv6_inetpeer_ops = {
6600     .init   =   ipv6_inetpeer_init,
6601     .exit   =   ipv6_inetpeer_exit,
6602 };
6603
6604 static struct pernet_operations ip6_route_net_late_ops = {
6605     .init = ip6_route_net_init_late,
6606     .exit = ip6_route_net_exit_late,
6607 };
6608
6609 static struct notifier_block ip6_route_dev_notifier = {
6610     .notifier_call = ip6_route_dev_notify,
6611     .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6612 };
6613
6614 void __init ip6_route_init_special_entries(void)
6615 {
6616     /* Registering of the loopback is done before this portion of code,
6617      * the loopback reference in rt6_info will not be taken, do it
6618      * manually for init_net */
6619     init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6620     init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6621     init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6622   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6623     init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6624     init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6625     init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6626     init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6627   #endif
6628 }
6629
6630 #if IS_BUILTIN(CONFIG_IPV6)
6631 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6632 DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6633
6634 BTF_ID_LIST(btf_fib6_info_id)
6635 BTF_ID(struct, fib6_info)
6636
6637 static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6638     .seq_ops        = &ipv6_route_seq_ops,
6639     .init_seq_private   = bpf_iter_init_seq_net,
6640     .fini_seq_private   = bpf_iter_fini_seq_net,
6641     .seq_priv_size      = sizeof(struct ipv6_route_iter),
6642 };
6643
6644 static struct bpf_iter_reg ipv6_route_reg_info = {
6645     .target         = "ipv6_route",
6646     .ctx_arg_info_size  = 1,
6647     .ctx_arg_info       = {
6648         { offsetof(struct bpf_iter__ipv6_route, rt),
6649           PTR_TO_BTF_ID_OR_NULL },
6650     },
6651     .seq_info       = &ipv6_route_seq_info,
6652 };
6653
6654 static int __init bpf_iter_register(void)
6655 {
6656     ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6657     return bpf_iter_reg_target(&ipv6_route_reg_info);
6658 }
6659
6660 static void bpf_iter_unregister(void)
6661 {
6662     bpf_iter_unreg_target(&ipv6_route_reg_info);
6663 }
6664 #endif
6665 #endif
6666
6667 int __init ip6_route_init(void)
6668 {
6669     int ret;
6670     int cpu;
6671
6672     ret = -ENOMEM;
6673     ip6_dst_ops_template.kmem_cachep =
6674         kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6675                   SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6676     if (!ip6_dst_ops_template.kmem_cachep)
6677         goto out;
6678
6679     ret = dst_entries_init(&ip6_dst_blackhole_ops);
6680     if (ret)
6681         goto out_kmem_cache;
6682
6683     ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6684     if (ret)
6685         goto out_dst_entries;
6686
6687     ret = register_pernet_subsys(&ip6_route_net_ops);
6688     if (ret)
6689         goto out_register_inetpeer;
6690
6691     ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6692
6693     ret = fib6_init();
6694     if (ret)
6695         goto out_register_subsys;
6696
6697     ret = xfrm6_init();
6698     if (ret)
6699         goto out_fib6_init;
6700
6701     ret = fib6_rules_init();
6702     if (ret)
6703         goto xfrm6_init;
6704
6705     ret = register_pernet_subsys(&ip6_route_net_late_ops);
6706     if (ret)
6707         goto fib6_rules_init;
6708
6709     ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6710                    inet6_rtm_newroute, NULL, 0);
6711     if (ret < 0)
6712         goto out_register_late_subsys;
6713
6714     ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6715                    inet6_rtm_delroute, NULL, 0);
6716     if (ret < 0)
6717         goto out_register_late_subsys;
6718
6719     ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6720                    inet6_rtm_getroute, NULL,
6721                    RTNL_FLAG_DOIT_UNLOCKED);
6722     if (ret < 0)
6723         goto out_register_late_subsys;
6724
6725     ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6726     if (ret)
6727         goto out_register_late_subsys;
6728
6729 #if IS_BUILTIN(CONFIG_IPV6)
6730 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6731     ret = bpf_iter_register();
6732     if (ret)
6733         goto out_register_late_subsys;
6734 #endif
6735 #endif
6736
6737     for_each_possible_cpu(cpu) {
6738         struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6739
6740         INIT_LIST_HEAD(&ul->head);
6741         INIT_LIST_HEAD(&ul->quarantine);
6742         spin_lock_init(&ul->lock);
6743     }
6744
6745 out:
6746     return ret;
6747
6748 out_register_late_subsys:
6749     rtnl_unregister_all(PF_INET6);
6750     unregister_pernet_subsys(&ip6_route_net_late_ops);
6751 fib6_rules_init:
6752     fib6_rules_cleanup();
6753 xfrm6_init:
6754     xfrm6_fini();
6755 out_fib6_init:
6756     fib6_gc_cleanup();
6757 out_register_subsys:
6758     unregister_pernet_subsys(&ip6_route_net_ops);
6759 out_register_inetpeer:
6760     unregister_pernet_subsys(&ipv6_inetpeer_ops);
6761 out_dst_entries:
6762     dst_entries_destroy(&ip6_dst_blackhole_ops);
6763 out_kmem_cache:
6764     kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6765     goto out;
6766 }
6767
6768 void ip6_route_cleanup(void)
6769 {
6770 #if IS_BUILTIN(CONFIG_IPV6)
6771 #if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6772     bpf_iter_unregister();
6773 #endif
6774 #endif
6775     unregister_netdevice_notifier(&ip6_route_dev_notifier);
6776     unregister_pernet_subsys(&ip6_route_net_late_ops);
6777     fib6_rules_cleanup();
6778     xfrm6_fini();
6779     fib6_gc_cleanup();
6780     unregister_pernet_subsys(&ipv6_inetpeer_ops);
6781     unregister_pernet_subsys(&ip6_route_net_ops);
6782     dst_entries_destroy(&ip6_dst_blackhole_ops);
6783     kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6784 }