1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	Linux INET6 implementation
   4 *	FIB front-end.
   5 *
   6 *	Authors:
   7 *	Pedro Roque		<roque@di.fc.ul.pt>
   8 */
   9
  10/*	Changes:
  11 *
  12 *	YOSHIFUJI Hideaki @USAGI
  13 *		reworked default router selection.
  14 *		- respect outgoing interface
  15 *		- select from (probably) reachable routers (i.e.
  16 *		routers in REACHABLE, STALE, DELAY or PROBE states).
  17 *		- always select the same router if it is (probably)
  18 *		reachable.  otherwise, round-robin the list.
  19 *	Ville Nuorvala
  20 *		Fixed routing subtrees.
  21 */
  22
  23#define pr_fmt(fmt) "IPv6: " fmt
  24
  25#include <linux/capability.h>
  26#include <linux/errno.h>
  27#include <linux/export.h>
  28#include <linux/types.h>
  29#include <linux/times.h>
  30#include <linux/socket.h>
  31#include <linux/sockios.h>
  32#include <linux/net.h>
  33#include <linux/route.h>
  34#include <linux/netdevice.h>
  35#include <linux/in6.h>
  36#include <linux/mroute6.h>
  37#include <linux/init.h>
  38#include <linux/if_arp.h>
  39#include <linux/proc_fs.h>
  40#include <linux/seq_file.h>
  41#include <linux/nsproxy.h>
  42#include <linux/slab.h>
  43#include <linux/jhash.h>
  44#include <linux/siphash.h>
  45#include <net/net_namespace.h>
  46#include <net/snmp.h>
  47#include <net/ipv6.h>
  48#include <net/ip6_fib.h>
  49#include <net/ip6_route.h>
  50#include <net/ndisc.h>
  51#include <net/addrconf.h>
  52#include <net/tcp.h>
  53#include <linux/rtnetlink.h>
  54#include <net/dst.h>
  55#include <net/dst_metadata.h>
  56#include <net/xfrm.h>
  57#include <net/netevent.h>
  58#include <net/netlink.h>
  59#include <net/rtnh.h>
  60#include <net/lwtunnel.h>
  61#include <net/ip_tunnels.h>
  62#include <net/l3mdev.h>
  63#include <net/ip.h>
  64#include <linux/uaccess.h>
  65#include <linux/btf_ids.h>
  66
  67#ifdef CONFIG_SYSCTL
  68#include <linux/sysctl.h>
  69#endif
  70
  71static int ip6_rt_type_to_error(u8 fib6_type);
  72
  73#define CREATE_TRACE_POINTS
  74#include <trace/events/fib6.h>
  75EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
  76#undef CREATE_TRACE_POINTS
  77
  78enum rt6_nud_state {
  79	RT6_NUD_FAIL_HARD = -3,
  80	RT6_NUD_FAIL_PROBE = -2,
  81	RT6_NUD_FAIL_DO_RR = -1,
  82	RT6_NUD_SUCCEED = 1
  83};
  84
  85INDIRECT_CALLABLE_SCOPE
  86struct dst_entry	*ip6_dst_check(struct dst_entry *dst, u32 cookie);
  87static unsigned int	 ip6_default_advmss(const struct dst_entry *dst);
  88INDIRECT_CALLABLE_SCOPE
  89unsigned int		ip6_mtu(const struct dst_entry *dst);
  90static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  91static void		ip6_dst_destroy(struct dst_entry *);
  92static void		ip6_dst_ifdown(struct dst_entry *,
  93				       struct net_device *dev);
  94static void		 ip6_dst_gc(struct dst_ops *ops);
  95
  96static int		ip6_pkt_discard(struct sk_buff *skb);
  97static int		ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  98static int		ip6_pkt_prohibit(struct sk_buff *skb);
  99static int		ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
 100static void		ip6_link_failure(struct sk_buff *skb);
 101static void		ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
 102					   struct sk_buff *skb, u32 mtu,
 103					   bool confirm_neigh);
 104static void		rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
 105					struct sk_buff *skb);
 106static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
 107			   int strict);
 108static size_t rt6_nlmsg_size(struct fib6_info *f6i);
 109static int rt6_fill_node(struct net *net, struct sk_buff *skb,
 110			 struct fib6_info *rt, struct dst_entry *dst,
 111			 struct in6_addr *dest, struct in6_addr *src,
 112			 int iif, int type, u32 portid, u32 seq,
 113			 unsigned int flags);
 114static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
 115					   const struct in6_addr *daddr,
 116					   const struct in6_addr *saddr);
 117
 118#ifdef CONFIG_IPV6_ROUTE_INFO
 119static struct fib6_info *rt6_add_route_info(struct net *net,
 120					   const struct in6_addr *prefix, int prefixlen,
 121					   const struct in6_addr *gwaddr,
 122					   struct net_device *dev,
 123					   unsigned int pref);
 124static struct fib6_info *rt6_get_route_info(struct net *net,
 125					   const struct in6_addr *prefix, int prefixlen,
 126					   const struct in6_addr *gwaddr,
 127					   struct net_device *dev);
 128#endif
 129
 130struct uncached_list {
 131	spinlock_t		lock;
 132	struct list_head	head;
 133	struct list_head	quarantine;
 134};
 135
 136static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 137
 138void rt6_uncached_list_add(struct rt6_info *rt)
 139{
 140	struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 141
 142	rt->dst.rt_uncached_list = ul;
 143
 144	spin_lock_bh(&ul->lock);
 145	list_add_tail(&rt->dst.rt_uncached, &ul->head);
 146	spin_unlock_bh(&ul->lock);
 147}
 148
 149void rt6_uncached_list_del(struct rt6_info *rt)
 150{
 151	if (!list_empty(&rt->dst.rt_uncached)) {
 152		struct uncached_list *ul = rt->dst.rt_uncached_list;
 153
 154		spin_lock_bh(&ul->lock);
 155		list_del_init(&rt->dst.rt_uncached);
 156		spin_unlock_bh(&ul->lock);
 157	}
 158}
 159
 160static void rt6_uncached_list_flush_dev(struct net_device *dev)
 161{
 162	int cpu;
 163
 164	for_each_possible_cpu(cpu) {
 165		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 166		struct rt6_info *rt, *safe;
 167
 168		if (list_empty(&ul->head))
 169			continue;
 170
 171		spin_lock_bh(&ul->lock);
 172		list_for_each_entry_safe(rt, safe, &ul->head, dst.rt_uncached) {
 173			struct inet6_dev *rt_idev = rt->rt6i_idev;
 174			struct net_device *rt_dev = rt->dst.dev;
 175			bool handled = false;
 176
 177			if (rt_idev->dev == dev) {
 178				rt->rt6i_idev = in6_dev_get(blackhole_netdev);
 179				in6_dev_put(rt_idev);
 180				handled = true;
 181			}
 182
 183			if (rt_dev == dev) {
 184				rt->dst.dev = blackhole_netdev;
 185				netdev_ref_replace(rt_dev, blackhole_netdev,
 186						   &rt->dst.dev_tracker,
 187						   GFP_ATOMIC);
 188				handled = true;
 189			}
 190			if (handled)
 191				list_move(&rt->dst.rt_uncached,
 192					  &ul->quarantine);
 193		}
 194		spin_unlock_bh(&ul->lock);
 195	}
 196}
 197
 198static inline const void *choose_neigh_daddr(const struct in6_addr *p,
 199					     struct sk_buff *skb,
 200					     const void *daddr)
 201{
 202	if (!ipv6_addr_any(p))
 203		return (const void *) p;
 204	else if (skb)
 205		return &ipv6_hdr(skb)->daddr;
 206	return daddr;
 207}
 208
 209struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
 210				   struct net_device *dev,
 211				   struct sk_buff *skb,
 212				   const void *daddr)
 213{
 214	struct neighbour *n;
 215
 216	daddr = choose_neigh_daddr(gw, skb, daddr);
 217	n = __ipv6_neigh_lookup(dev, daddr);
 218	if (n)
 219		return n;
 220
 221	n = neigh_create(&nd_tbl, daddr, dev);
 222	return IS_ERR(n) ? NULL : n;
 223}
 224
 225static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
 226					      struct sk_buff *skb,
 227					      const void *daddr)
 228{
 229	const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
 230
 231	return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
 232				dst->dev, skb, daddr);
 233}
 234
 235static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
 236{
 237	struct net_device *dev = dst->dev;
 238	struct rt6_info *rt = (struct rt6_info *)dst;
 239
 240	daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
 241	if (!daddr)
 242		return;
 243	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 244		return;
 245	if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
 246		return;
 247	__ipv6_confirm_neigh(dev, daddr);
 248}
 249
 250static struct dst_ops ip6_dst_ops_template = {
 251	.family			=	AF_INET6,
 252	.gc			=	ip6_dst_gc,
 253	.gc_thresh		=	1024,
 254	.check			=	ip6_dst_check,
 255	.default_advmss		=	ip6_default_advmss,
 256	.mtu			=	ip6_mtu,
 257	.cow_metrics		=	dst_cow_metrics_generic,
 258	.destroy		=	ip6_dst_destroy,
 259	.ifdown			=	ip6_dst_ifdown,
 260	.negative_advice	=	ip6_negative_advice,
 261	.link_failure		=	ip6_link_failure,
 262	.update_pmtu		=	ip6_rt_update_pmtu,
 263	.redirect		=	rt6_do_redirect,
 264	.local_out		=	__ip6_local_out,
 265	.neigh_lookup		=	ip6_dst_neigh_lookup,
 266	.confirm_neigh		=	ip6_confirm_neigh,
 267};
 268
 269static struct dst_ops ip6_dst_blackhole_ops = {
 270	.family			= AF_INET6,
 271	.default_advmss		= ip6_default_advmss,
 272	.neigh_lookup		= ip6_dst_neigh_lookup,
 273	.check			= ip6_dst_check,
 274	.destroy		= ip6_dst_destroy,
 275	.cow_metrics		= dst_cow_metrics_generic,
 276	.update_pmtu		= dst_blackhole_update_pmtu,
 277	.redirect		= dst_blackhole_redirect,
 278	.mtu			= dst_blackhole_mtu,
 279};
 280
 281static const u32 ip6_template_metrics[RTAX_MAX] = {
 282	[RTAX_HOPLIMIT - 1] = 0,
 283};
 284
 285static const struct fib6_info fib6_null_entry_template = {
 286	.fib6_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 287	.fib6_protocol  = RTPROT_KERNEL,
 288	.fib6_metric	= ~(u32)0,
 289	.fib6_ref	= REFCOUNT_INIT(1),
 290	.fib6_type	= RTN_UNREACHABLE,
 291	.fib6_metrics	= (struct dst_metrics *)&dst_default_metrics,
 292};
 293
 294static const struct rt6_info ip6_null_entry_template = {
 295	.dst = {
 296		.__rcuref	= RCUREF_INIT(1),
 297		.__use		= 1,
 298		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 299		.error		= -ENETUNREACH,
 300		.input		= ip6_pkt_discard,
 301		.output		= ip6_pkt_discard_out,
 302	},
 303	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 304};
 305
 306#ifdef CONFIG_IPV6_MULTIPLE_TABLES
 307
 308static const struct rt6_info ip6_prohibit_entry_template = {
 309	.dst = {
 310		.__rcuref	= RCUREF_INIT(1),
 311		.__use		= 1,
 312		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 313		.error		= -EACCES,
 314		.input		= ip6_pkt_prohibit,
 315		.output		= ip6_pkt_prohibit_out,
 316	},
 317	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 318};
 319
 320static const struct rt6_info ip6_blk_hole_entry_template = {
 321	.dst = {
 322		.__rcuref	= RCUREF_INIT(1),
 323		.__use		= 1,
 324		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 325		.error		= -EINVAL,
 326		.input		= dst_discard,
 327		.output		= dst_discard_out,
 328	},
 329	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 330};
 331
 332#endif
 333
 334static void rt6_info_init(struct rt6_info *rt)
 335{
 336	memset_after(rt, 0, dst);
 337}
 338
 339/* allocate dst with ip6_dst_ops */
 340struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
 341			       int flags)
 342{
 343	struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 344					DST_OBSOLETE_FORCE_CHK, flags);
 345
 346	if (rt) {
 347		rt6_info_init(rt);
 348		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
 349	}
 350
 351	return rt;
 352}
 353EXPORT_SYMBOL(ip6_dst_alloc);
 354
 355static void ip6_dst_destroy(struct dst_entry *dst)
 356{
 357	struct rt6_info *rt = (struct rt6_info *)dst;
 358	struct fib6_info *from;
 359	struct inet6_dev *idev;
 360
 361	ip_dst_metrics_put(dst);
 362	rt6_uncached_list_del(rt);
 363
 364	idev = rt->rt6i_idev;
 365	if (idev) {
 366		rt->rt6i_idev = NULL;
 367		in6_dev_put(idev);
 368	}
 369
 370	from = xchg((__force struct fib6_info **)&rt->from, NULL);
 371	fib6_info_release(from);
 372}
 373
 374static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
 375{
 376	struct rt6_info *rt = (struct rt6_info *)dst;
 377	struct inet6_dev *idev = rt->rt6i_idev;
 378
 379	if (idev && idev->dev != blackhole_netdev) {
 380		struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);
 381
 382		if (blackhole_idev) {
 383			rt->rt6i_idev = blackhole_idev;
 384			in6_dev_put(idev);
 385		}
 386	}
 387}
 388
 389static bool __rt6_check_expired(const struct rt6_info *rt)
 390{
 391	if (rt->rt6i_flags & RTF_EXPIRES)
 392		return time_after(jiffies, rt->dst.expires);
 393	else
 394		return false;
 395}
 396
 397static bool rt6_check_expired(const struct rt6_info *rt)
 398{
 399	struct fib6_info *from;
 400
 401	from = rcu_dereference(rt->from);
 402
 403	if (rt->rt6i_flags & RTF_EXPIRES) {
 404		if (time_after(jiffies, rt->dst.expires))
 405			return true;
 406	} else if (from) {
 407		return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
 408			fib6_check_expired(from);
 409	}
 410	return false;
 411}
 412
 413void fib6_select_path(const struct net *net, struct fib6_result *res,
 414		      struct flowi6 *fl6, int oif, bool have_oif_match,
 415		      const struct sk_buff *skb, int strict)
 416{
 417	struct fib6_info *sibling, *next_sibling;
 418	struct fib6_info *match = res->f6i;
 419
 420	if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
 421		goto out;
 422
 423	if (match->nh && have_oif_match && res->nh)
 424		return;
 425
 426	if (skb)
 427		IP6CB(skb)->flags |= IP6SKB_MULTIPATH;
 428
 429	/* We might have already computed the hash for ICMPv6 errors. In such
 430	 * case it will always be non-zero. Otherwise now is the time to do it.
 431	 */
 432	if (!fl6->mp_hash &&
 433	    (!match->nh || nexthop_is_multipath(match->nh)))
 434		fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
 435
 436	if (unlikely(match->nh)) {
 437		nexthop_path_fib6_result(res, fl6->mp_hash);
 438		return;
 439	}
 440
 441	if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
 442		goto out;
 443
 444	list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
 445				 fib6_siblings) {
 446		const struct fib6_nh *nh = sibling->fib6_nh;
 447		int nh_upper_bound;
 448
 449		nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
 450		if (fl6->mp_hash > nh_upper_bound)
 451			continue;
 452		if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
 453			break;
 454		match = sibling;
 455		break;
 456	}
 457
 458out:
 459	res->f6i = match;
 460	res->nh = match->fib6_nh;
 461}
 462
 463/*
 464 *	Route lookup. rcu_read_lock() should be held.
 465 */
 466
 467static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
 468			       const struct in6_addr *saddr, int oif, int flags)
 469{
 470	const struct net_device *dev;
 471
 472	if (nh->fib_nh_flags & RTNH_F_DEAD)
 473		return false;
 474
 475	dev = nh->fib_nh_dev;
 476	if (oif) {
 477		if (dev->ifindex == oif)
 478			return true;
 479	} else {
 480		if (ipv6_chk_addr(net, saddr, dev,
 481				  flags & RT6_LOOKUP_F_IFACE))
 482			return true;
 483	}
 484
 485	return false;
 486}
 487
 488struct fib6_nh_dm_arg {
 489	struct net		*net;
 490	const struct in6_addr	*saddr;
 491	int			oif;
 492	int			flags;
 493	struct fib6_nh		*nh;
 494};
 495
 496static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
 497{
 498	struct fib6_nh_dm_arg *arg = _arg;
 499
 500	arg->nh = nh;
 501	return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
 502				  arg->flags);
 503}
 504
 505/* returns fib6_nh from nexthop or NULL */
 506static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
 507					struct fib6_result *res,
 508					const struct in6_addr *saddr,
 509					int oif, int flags)
 510{
 511	struct fib6_nh_dm_arg arg = {
 512		.net   = net,
 513		.saddr = saddr,
 514		.oif   = oif,
 515		.flags = flags,
 516	};
 517
 518	if (nexthop_is_blackhole(nh))
 519		return NULL;
 520
 521	if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
 522		return arg.nh;
 523
 524	return NULL;
 525}
 526
 527static void rt6_device_match(struct net *net, struct fib6_result *res,
 528			     const struct in6_addr *saddr, int oif, int flags)
 529{
 530	struct fib6_info *f6i = res->f6i;
 531	struct fib6_info *spf6i;
 532	struct fib6_nh *nh;
 533
 534	if (!oif && ipv6_addr_any(saddr)) {
 535		if (unlikely(f6i->nh)) {
 536			nh = nexthop_fib6_nh(f6i->nh);
 537			if (nexthop_is_blackhole(f6i->nh))
 538				goto out_blackhole;
 539		} else {
 540			nh = f6i->fib6_nh;
 541		}
 542		if (!(nh->fib_nh_flags & RTNH_F_DEAD))
 543			goto out;
 544	}
 545
 546	for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
 547		bool matched = false;
 548
 549		if (unlikely(spf6i->nh)) {
 550			nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
 551					      oif, flags);
 552			if (nh)
 553				matched = true;
 554		} else {
 555			nh = spf6i->fib6_nh;
 556			if (__rt6_device_match(net, nh, saddr, oif, flags))
 557				matched = true;
 558		}
 559		if (matched) {
 560			res->f6i = spf6i;
 561			goto out;
 562		}
 563	}
 564
 565	if (oif && flags & RT6_LOOKUP_F_IFACE) {
 566		res->f6i = net->ipv6.fib6_null_entry;
 567		nh = res->f6i->fib6_nh;
 568		goto out;
 569	}
 570
 571	if (unlikely(f6i->nh)) {
 572		nh = nexthop_fib6_nh(f6i->nh);
 573		if (nexthop_is_blackhole(f6i->nh))
 574			goto out_blackhole;
 575	} else {
 576		nh = f6i->fib6_nh;
 577	}
 578
 579	if (nh->fib_nh_flags & RTNH_F_DEAD) {
 580		res->f6i = net->ipv6.fib6_null_entry;
 581		nh = res->f6i->fib6_nh;
 582	}
 583out:
 584	res->nh = nh;
 585	res->fib6_type = res->f6i->fib6_type;
 586	res->fib6_flags = res->f6i->fib6_flags;
 587	return;
 588
 589out_blackhole:
 590	res->fib6_flags |= RTF_REJECT;
 591	res->fib6_type = RTN_BLACKHOLE;
 592	res->nh = nh;
 593}
 594
 595#ifdef CONFIG_IPV6_ROUTER_PREF
 596struct __rt6_probe_work {
 597	struct work_struct work;
 598	struct in6_addr target;
 599	struct net_device *dev;
 600	netdevice_tracker dev_tracker;
 601};
 602
 603static void rt6_probe_deferred(struct work_struct *w)
 604{
 605	struct in6_addr mcaddr;
 606	struct __rt6_probe_work *work =
 607		container_of(w, struct __rt6_probe_work, work);
 608
 609	addrconf_addr_solict_mult(&work->target, &mcaddr);
 610	ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
 611	netdev_put(work->dev, &work->dev_tracker);
 612	kfree(work);
 613}
 614
 615static void rt6_probe(struct fib6_nh *fib6_nh)
 616{
 617	struct __rt6_probe_work *work = NULL;
 618	const struct in6_addr *nh_gw;
 619	unsigned long last_probe;
 620	struct neighbour *neigh;
 621	struct net_device *dev;
 622	struct inet6_dev *idev;
 623
 624	/*
 625	 * Okay, this does not seem to be appropriate
 626	 * for now, however, we need to check if it
 627	 * is really so; aka Router Reachability Probing.
 628	 *
 629	 * Router Reachability Probe MUST be rate-limited
 630	 * to no more than one per minute.
 631	 */
 632	if (!fib6_nh->fib_nh_gw_family)
 633		return;
 634
 635	nh_gw = &fib6_nh->fib_nh_gw6;
 636	dev = fib6_nh->fib_nh_dev;
 637	rcu_read_lock();
 638	last_probe = READ_ONCE(fib6_nh->last_probe);
 639	idev = __in6_dev_get(dev);
 640	neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
 641	if (neigh) {
 642		if (READ_ONCE(neigh->nud_state) & NUD_VALID)
 643			goto out;
 644
 645		write_lock_bh(&neigh->lock);
 646		if (!(neigh->nud_state & NUD_VALID) &&
 647		    time_after(jiffies,
 648			       neigh->updated + idev->cnf.rtr_probe_interval)) {
 649			work = kmalloc(sizeof(*work), GFP_ATOMIC);
 650			if (work)
 651				__neigh_set_probe_once(neigh);
 652		}
 653		write_unlock_bh(&neigh->lock);
 654	} else if (time_after(jiffies, last_probe +
 655				       idev->cnf.rtr_probe_interval)) {
 656		work = kmalloc(sizeof(*work), GFP_ATOMIC);
 657	}
 658
 659	if (!work || cmpxchg(&fib6_nh->last_probe,
 660			     last_probe, jiffies) != last_probe) {
 661		kfree(work);
 662	} else {
 663		INIT_WORK(&work->work, rt6_probe_deferred);
 664		work->target = *nh_gw;
 665		netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
 666		work->dev = dev;
 667		schedule_work(&work->work);
 668	}
 669
 670out:
 671	rcu_read_unlock();
 672}
 673#else
 674static inline void rt6_probe(struct fib6_nh *fib6_nh)
 675{
 676}
 677#endif
 678
 679/*
 680 * Default Router Selection (RFC 2461 6.3.6)
 681 */
 682static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
 683{
 684	enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 685	struct neighbour *neigh;
 686
 687	rcu_read_lock();
 688	neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
 689					  &fib6_nh->fib_nh_gw6);
 690	if (neigh) {
 691		u8 nud_state = READ_ONCE(neigh->nud_state);
 692
 693		if (nud_state & NUD_VALID)
 694			ret = RT6_NUD_SUCCEED;
 695#ifdef CONFIG_IPV6_ROUTER_PREF
 696		else if (!(nud_state & NUD_FAILED))
 697			ret = RT6_NUD_SUCCEED;
 698		else
 699			ret = RT6_NUD_FAIL_PROBE;
 700#endif
 701	} else {
 702		ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 703		      RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 704	}
 705	rcu_read_unlock();
 706
 707	return ret;
 708}
 709
 710static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
 711			   int strict)
 712{
 713	int m = 0;
 714
 715	if (!oif || nh->fib_nh_dev->ifindex == oif)
 716		m = 2;
 717
 718	if (!m && (strict & RT6_LOOKUP_F_IFACE))
 719		return RT6_NUD_FAIL_HARD;
 720#ifdef CONFIG_IPV6_ROUTER_PREF
 721	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
 722#endif
 723	if ((strict & RT6_LOOKUP_F_REACHABLE) &&
 724	    !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
 725		int n = rt6_check_neigh(nh);
 726		if (n < 0)
 727			return n;
 728	}
 729	return m;
 730}
 731
 732static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
 733		       int oif, int strict, int *mpri, bool *do_rr)
 734{
 735	bool match_do_rr = false;
 736	bool rc = false;
 737	int m;
 738
 739	if (nh->fib_nh_flags & RTNH_F_DEAD)
 740		goto out;
 741
 742	if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
 743	    nh->fib_nh_flags & RTNH_F_LINKDOWN &&
 744	    !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
 745		goto out;
 746
 747	m = rt6_score_route(nh, fib6_flags, oif, strict);
 748	if (m == RT6_NUD_FAIL_DO_RR) {
 749		match_do_rr = true;
 750		m = 0; /* lowest valid score */
 751	} else if (m == RT6_NUD_FAIL_HARD) {
 752		goto out;
 753	}
 754
 755	if (strict & RT6_LOOKUP_F_REACHABLE)
 756		rt6_probe(nh);
 757
 758	/* note that m can be RT6_NUD_FAIL_PROBE at this point */
 759	if (m > *mpri) {
 760		*do_rr = match_do_rr;
 761		*mpri = m;
 762		rc = true;
 763	}
 764out:
 765	return rc;
 766}
 767
 768struct fib6_nh_frl_arg {
 769	u32		flags;
 770	int		oif;
 771	int		strict;
 772	int		*mpri;
 773	bool		*do_rr;
 774	struct fib6_nh	*nh;
 775};
 776
 777static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
 778{
 779	struct fib6_nh_frl_arg *arg = _arg;
 780
 781	arg->nh = nh;
 782	return find_match(nh, arg->flags, arg->oif, arg->strict,
 783			  arg->mpri, arg->do_rr);
 784}
 785
 786static void __find_rr_leaf(struct fib6_info *f6i_start,
 787			   struct fib6_info *nomatch, u32 metric,
 788			   struct fib6_result *res, struct fib6_info **cont,
 789			   int oif, int strict, bool *do_rr, int *mpri)
 790{
 791	struct fib6_info *f6i;
 792
 793	for (f6i = f6i_start;
 794	     f6i && f6i != nomatch;
 795	     f6i = rcu_dereference(f6i->fib6_next)) {
 796		bool matched = false;
 797		struct fib6_nh *nh;
 798
 799		if (cont && f6i->fib6_metric != metric) {
 800			*cont = f6i;
 801			return;
 802		}
 803
 804		if (fib6_check_expired(f6i))
 805			continue;
 806
 807		if (unlikely(f6i->nh)) {
 808			struct fib6_nh_frl_arg arg = {
 809				.flags  = f6i->fib6_flags,
 810				.oif    = oif,
 811				.strict = strict,
 812				.mpri   = mpri,
 813				.do_rr  = do_rr
 814			};
 815
 816			if (nexthop_is_blackhole(f6i->nh)) {
 817				res->fib6_flags = RTF_REJECT;
 818				res->fib6_type = RTN_BLACKHOLE;
 819				res->f6i = f6i;
 820				res->nh = nexthop_fib6_nh(f6i->nh);
 821				return;
 822			}
 823			if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
 824						     &arg)) {
 825				matched = true;
 826				nh = arg.nh;
 827			}
 828		} else {
 829			nh = f6i->fib6_nh;
 830			if (find_match(nh, f6i->fib6_flags, oif, strict,
 831				       mpri, do_rr))
 832				matched = true;
 833		}
 834		if (matched) {
 835			res->f6i = f6i;
 836			res->nh = nh;
 837			res->fib6_flags = f6i->fib6_flags;
 838			res->fib6_type = f6i->fib6_type;
 839		}
 840	}
 841}
 842
 843static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
 844			 struct fib6_info *rr_head, int oif, int strict,
 845			 bool *do_rr, struct fib6_result *res)
 846{
 847	u32 metric = rr_head->fib6_metric;
 848	struct fib6_info *cont = NULL;
 849	int mpri = -1;
 850
 851	__find_rr_leaf(rr_head, NULL, metric, res, &cont,
 852		       oif, strict, do_rr, &mpri);
 853
 854	__find_rr_leaf(leaf, rr_head, metric, res, &cont,
 855		       oif, strict, do_rr, &mpri);
 856
 857	if (res->f6i || !cont)
 858		return;
 859
 860	__find_rr_leaf(cont, NULL, metric, res, NULL,
 861		       oif, strict, do_rr, &mpri);
 862}
 863
 864static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
 865		       struct fib6_result *res, int strict)
 866{
 867	struct fib6_info *leaf = rcu_dereference(fn->leaf);
 868	struct fib6_info *rt0;
 869	bool do_rr = false;
 870	int key_plen;
 871
 872	/* make sure this function or its helpers sets f6i */
 873	res->f6i = NULL;
 874
 875	if (!leaf || leaf == net->ipv6.fib6_null_entry)
 876		goto out;
 877
 878	rt0 = rcu_dereference(fn->rr_ptr);
 879	if (!rt0)
 880		rt0 = leaf;
 881
 882	/* Double check to make sure fn is not an intermediate node
 883	 * and fn->leaf does not points to its child's leaf
 884	 * (This might happen if all routes under fn are deleted from
 885	 * the tree and fib6_repair_tree() is called on the node.)
 886	 */
 887	key_plen = rt0->fib6_dst.plen;
 888#ifdef CONFIG_IPV6_SUBTREES
 889	if (rt0->fib6_src.plen)
 890		key_plen = rt0->fib6_src.plen;
 891#endif
 892	if (fn->fn_bit != key_plen)
 893		goto out;
 894
 895	find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
 896	if (do_rr) {
 897		struct fib6_info *next = rcu_dereference(rt0->fib6_next);
 898
 899		/* no entries matched; do round-robin */
 900		if (!next || next->fib6_metric != rt0->fib6_metric)
 901			next = leaf;
 902
 903		if (next != rt0) {
 904			spin_lock_bh(&leaf->fib6_table->tb6_lock);
 905			/* make sure next is not being deleted from the tree */
 906			if (next->fib6_node)
 907				rcu_assign_pointer(fn->rr_ptr, next);
 908			spin_unlock_bh(&leaf->fib6_table->tb6_lock);
 909		}
 910	}
 911
 912out:
 913	if (!res->f6i) {
 914		res->f6i = net->ipv6.fib6_null_entry;
 915		res->nh = res->f6i->fib6_nh;
 916		res->fib6_flags = res->f6i->fib6_flags;
 917		res->fib6_type = res->f6i->fib6_type;
 918	}
 919}
 920
 921static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
 922{
 923	return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
 924	       res->nh->fib_nh_gw_family;
 925}
 926
 927#ifdef CONFIG_IPV6_ROUTE_INFO
 928int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 929		  const struct in6_addr *gwaddr)
 930{
 931	struct net *net = dev_net(dev);
 932	struct route_info *rinfo = (struct route_info *) opt;
 933	struct in6_addr prefix_buf, *prefix;
 934	unsigned int pref;
 935	unsigned long lifetime;
 936	struct fib6_info *rt;
 937
 938	if (len < sizeof(struct route_info)) {
 939		return -EINVAL;
 940	}
 941
 942	/* Sanity check for prefix_len and length */
 943	if (rinfo->length > 3) {
 944		return -EINVAL;
 945	} else if (rinfo->prefix_len > 128) {
 946		return -EINVAL;
 947	} else if (rinfo->prefix_len > 64) {
 948		if (rinfo->length < 2) {
 949			return -EINVAL;
 950		}
 951	} else if (rinfo->prefix_len > 0) {
 952		if (rinfo->length < 1) {
 953			return -EINVAL;
 954		}
 955	}
 956
 957	pref = rinfo->route_pref;
 958	if (pref == ICMPV6_ROUTER_PREF_INVALID)
 959		return -EINVAL;
 960
 961	lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 962
 963	if (rinfo->length == 3)
 964		prefix = (struct in6_addr *)rinfo->prefix;
 965	else {
 966		/* this function is safe */
 967		ipv6_addr_prefix(&prefix_buf,
 968				 (struct in6_addr *)rinfo->prefix,
 969				 rinfo->prefix_len);
 970		prefix = &prefix_buf;
 971	}
 972
 973	if (rinfo->prefix_len == 0)
 974		rt = rt6_get_dflt_router(net, gwaddr, dev);
 975	else
 976		rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 977					gwaddr, dev);
 978
 979	if (rt && !lifetime) {
 980		ip6_del_rt(net, rt, false);
 981		rt = NULL;
 982	}
 983
 984	if (!rt && lifetime)
 985		rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 986					dev, pref);
 987	else if (rt)
 988		rt->fib6_flags = RTF_ROUTEINFO |
 989				 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 990
 991	if (rt) {
 992		if (!addrconf_finite_timeout(lifetime))
 993			fib6_clean_expires(rt);
 994		else
 995			fib6_set_expires(rt, jiffies + HZ * lifetime);
 996
 997		fib6_info_release(rt);
 998	}
 999	return 0;
1000}
1001#endif
1002
1003/*
1004 *	Misc support functions
1005 */
1006
1007/* called with rcu_lock held */
1008static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1009{
1010	struct net_device *dev = res->nh->fib_nh_dev;
1011
1012	if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1013		/* for copies of local routes, dst->dev needs to be the
1014		 * device if it is a master device, the master device if
1015		 * device is enslaved, and the loopback as the default
1016		 */
1017		if (netif_is_l3_slave(dev) &&
1018		    !rt6_need_strict(&res->f6i->fib6_dst.addr))
1019			dev = l3mdev_master_dev_rcu(dev);
1020		else if (!netif_is_l3_master(dev))
1021			dev = dev_net(dev)->loopback_dev;
1022		/* last case is netif_is_l3_master(dev) is true in which
1023		 * case we want dev returned to be dev
1024		 */
1025	}
1026
1027	return dev;
1028}
1029
1030static const int fib6_prop[RTN_MAX + 1] = {
1031	[RTN_UNSPEC]	= 0,
1032	[RTN_UNICAST]	= 0,
1033	[RTN_LOCAL]	= 0,
1034	[RTN_BROADCAST]	= 0,
1035	[RTN_ANYCAST]	= 0,
1036	[RTN_MULTICAST]	= 0,
1037	[RTN_BLACKHOLE]	= -EINVAL,
1038	[RTN_UNREACHABLE] = -EHOSTUNREACH,
1039	[RTN_PROHIBIT]	= -EACCES,
1040	[RTN_THROW]	= -EAGAIN,
1041	[RTN_NAT]	= -EINVAL,
1042	[RTN_XRESOLVE]	= -EINVAL,
1043};
1044
1045static int ip6_rt_type_to_error(u8 fib6_type)
1046{
1047	return fib6_prop[fib6_type];
1048}
1049
1050static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1051{
1052	unsigned short flags = 0;
1053
1054	if (rt->dst_nocount)
1055		flags |= DST_NOCOUNT;
1056	if (rt->dst_nopolicy)
1057		flags |= DST_NOPOLICY;
1058
1059	return flags;
1060}
1061
1062static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1063{
1064	rt->dst.error = ip6_rt_type_to_error(fib6_type);
1065
1066	switch (fib6_type) {
1067	case RTN_BLACKHOLE:
1068		rt->dst.output = dst_discard_out;
1069		rt->dst.input = dst_discard;
1070		break;
1071	case RTN_PROHIBIT:
1072		rt->dst.output = ip6_pkt_prohibit_out;
1073		rt->dst.input = ip6_pkt_prohibit;
1074		break;
1075	case RTN_THROW:
1076	case RTN_UNREACHABLE:
1077	default:
1078		rt->dst.output = ip6_pkt_discard_out;
1079		rt->dst.input = ip6_pkt_discard;
1080		break;
1081	}
1082}
1083
1084static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1085{
1086	struct fib6_info *f6i = res->f6i;
1087
1088	if (res->fib6_flags & RTF_REJECT) {
1089		ip6_rt_init_dst_reject(rt, res->fib6_type);
1090		return;
1091	}
1092
1093	rt->dst.error = 0;
1094	rt->dst.output = ip6_output;
1095
1096	if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1097		rt->dst.input = ip6_input;
1098	} else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1099		rt->dst.input = ip6_mc_input;
1100	} else {
1101		rt->dst.input = ip6_forward;
1102	}
1103
1104	if (res->nh->fib_nh_lws) {
1105		rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1106		lwtunnel_set_redirect(&rt->dst);
1107	}
1108
1109	rt->dst.lastuse = jiffies;
1110}
1111
1112/* Caller must already hold reference to @from */
1113static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1114{
1115	rt->rt6i_flags &= ~RTF_EXPIRES;
1116	rcu_assign_pointer(rt->from, from);
1117	ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1118}
1119
1120/* Caller must already hold reference to f6i in result */
1121static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1122{
1123	const struct fib6_nh *nh = res->nh;
1124	const struct net_device *dev = nh->fib_nh_dev;
1125	struct fib6_info *f6i = res->f6i;
1126
1127	ip6_rt_init_dst(rt, res);
1128
1129	rt->rt6i_dst = f6i->fib6_dst;
1130	rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1131	rt->rt6i_flags = res->fib6_flags;
1132	if (nh->fib_nh_gw_family) {
1133		rt->rt6i_gateway = nh->fib_nh_gw6;
1134		rt->rt6i_flags |= RTF_GATEWAY;
1135	}
1136	rt6_set_from(rt, f6i);
1137#ifdef CONFIG_IPV6_SUBTREES
1138	rt->rt6i_src = f6i->fib6_src;
1139#endif
1140}
1141
1142static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1143					struct in6_addr *saddr)
1144{
1145	struct fib6_node *pn, *sn;
1146	while (1) {
1147		if (fn->fn_flags & RTN_TL_ROOT)
1148			return NULL;
1149		pn = rcu_dereference(fn->parent);
1150		sn = FIB6_SUBTREE(pn);
1151		if (sn && sn != fn)
1152			fn = fib6_node_lookup(sn, NULL, saddr);
1153		else
1154			fn = pn;
1155		if (fn->fn_flags & RTN_RTINFO)
1156			return fn;
1157	}
1158}
1159
1160static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1161{
1162	struct rt6_info *rt = *prt;
1163
1164	if (dst_hold_safe(&rt->dst))
1165		return true;
1166	if (net) {
1167		rt = net->ipv6.ip6_null_entry;
1168		dst_hold(&rt->dst);
1169	} else {
1170		rt = NULL;
1171	}
1172	*prt = rt;
1173	return false;
1174}
1175
1176/* called with rcu_lock held */
1177static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1178{
1179	struct net_device *dev = res->nh->fib_nh_dev;
1180	struct fib6_info *f6i = res->f6i;
1181	unsigned short flags;
1182	struct rt6_info *nrt;
1183
1184	if (!fib6_info_hold_safe(f6i))
1185		goto fallback;
1186
1187	flags = fib6_info_dst_flags(f6i);
1188	nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1189	if (!nrt) {
1190		fib6_info_release(f6i);
1191		goto fallback;
1192	}
1193
1194	ip6_rt_copy_init(nrt, res);
1195	return nrt;
1196
1197fallback:
1198	nrt = dev_net(dev)->ipv6.ip6_null_entry;
1199	dst_hold(&nrt->dst);
1200	return nrt;
1201}
1202
1203INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1204					     struct fib6_table *table,
1205					     struct flowi6 *fl6,
1206					     const struct sk_buff *skb,
1207					     int flags)
1208{
1209	struct fib6_result res = {};
1210	struct fib6_node *fn;
1211	struct rt6_info *rt;
1212
1213	rcu_read_lock();
1214	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1215restart:
1216	res.f6i = rcu_dereference(fn->leaf);
1217	if (!res.f6i)
1218		res.f6i = net->ipv6.fib6_null_entry;
1219	else
1220		rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1221				 flags);
1222
1223	if (res.f6i == net->ipv6.fib6_null_entry) {
1224		fn = fib6_backtrack(fn, &fl6->saddr);
1225		if (fn)
1226			goto restart;
1227
1228		rt = net->ipv6.ip6_null_entry;
1229		dst_hold(&rt->dst);
1230		goto out;
1231	} else if (res.fib6_flags & RTF_REJECT) {
1232		goto do_create;
1233	}
1234
1235	fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1236			 fl6->flowi6_oif != 0, skb, flags);
1237
1238	/* Search through exception table */
1239	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1240	if (rt) {
1241		if (ip6_hold_safe(net, &rt))
1242			dst_use_noref(&rt->dst, jiffies);
1243	} else {
1244do_create:
1245		rt = ip6_create_rt_rcu(&res);
1246	}
1247
1248out:
1249	trace_fib6_table_lookup(net, &res, table, fl6);
1250
1251	rcu_read_unlock();
1252
1253	return rt;
1254}
1255
1256struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1257				   const struct sk_buff *skb, int flags)
1258{
1259	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1260}
1261EXPORT_SYMBOL_GPL(ip6_route_lookup);
1262
1263struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1264			    const struct in6_addr *saddr, int oif,
1265			    const struct sk_buff *skb, int strict)
1266{
1267	struct flowi6 fl6 = {
1268		.flowi6_oif = oif,
1269		.daddr = *daddr,
1270	};
1271	struct dst_entry *dst;
1272	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1273
1274	if (saddr) {
1275		memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1276		flags |= RT6_LOOKUP_F_HAS_SADDR;
1277	}
1278
1279	dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1280	if (dst->error == 0)
1281		return (struct rt6_info *) dst;
1282
1283	dst_release(dst);
1284
1285	return NULL;
1286}
1287EXPORT_SYMBOL(rt6_lookup);
1288
1289/* ip6_ins_rt is called with FREE table->tb6_lock.
1290 * It takes new route entry, the addition fails by any reason the
1291 * route is released.
1292 * Caller must hold dst before calling it.
1293 */
1294
1295static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1296			struct netlink_ext_ack *extack)
1297{
1298	int err;
1299	struct fib6_table *table;
1300
1301	table = rt->fib6_table;
1302	spin_lock_bh(&table->tb6_lock);
1303	err = fib6_add(&table->tb6_root, rt, info, extack);
1304	spin_unlock_bh(&table->tb6_lock);
1305
1306	return err;
1307}
1308
1309int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1310{
1311	struct nl_info info = {	.nl_net = net, };
1312
1313	return __ip6_ins_rt(rt, &info, NULL);
1314}
1315
1316static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1317					   const struct in6_addr *daddr,
1318					   const struct in6_addr *saddr)
1319{
1320	struct fib6_info *f6i = res->f6i;
1321	struct net_device *dev;
1322	struct rt6_info *rt;
1323
1324	/*
1325	 *	Clone the route.
1326	 */
1327
1328	if (!fib6_info_hold_safe(f6i))
1329		return NULL;
1330
1331	dev = ip6_rt_get_dev_rcu(res);
1332	rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1333	if (!rt) {
1334		fib6_info_release(f6i);
1335		return NULL;
1336	}
1337
1338	ip6_rt_copy_init(rt, res);
1339	rt->rt6i_flags |= RTF_CACHE;
1340	rt->rt6i_dst.addr = *daddr;
1341	rt->rt6i_dst.plen = 128;
1342
1343	if (!rt6_is_gw_or_nonexthop(res)) {
1344		if (f6i->fib6_dst.plen != 128 &&
1345		    ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1346			rt->rt6i_flags |= RTF_ANYCAST;
1347#ifdef CONFIG_IPV6_SUBTREES
1348		if (rt->rt6i_src.plen && saddr) {
1349			rt->rt6i_src.addr = *saddr;
1350			rt->rt6i_src.plen = 128;
1351		}
1352#endif
1353	}
1354
1355	return rt;
1356}
1357
1358static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1359{
1360	struct fib6_info *f6i = res->f6i;
1361	unsigned short flags = fib6_info_dst_flags(f6i);
1362	struct net_device *dev;
1363	struct rt6_info *pcpu_rt;
1364
1365	if (!fib6_info_hold_safe(f6i))
1366		return NULL;
1367
1368	rcu_read_lock();
1369	dev = ip6_rt_get_dev_rcu(res);
1370	pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1371	rcu_read_unlock();
1372	if (!pcpu_rt) {
1373		fib6_info_release(f6i);
1374		return NULL;
1375	}
1376	ip6_rt_copy_init(pcpu_rt, res);
1377	pcpu_rt->rt6i_flags |= RTF_PCPU;
1378
1379	if (f6i->nh)
1380		pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1381
1382	return pcpu_rt;
1383}
1384
1385static bool rt6_is_valid(const struct rt6_info *rt6)
1386{
1387	return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1388}
1389
1390/* It should be called with rcu_read_lock() acquired */
1391static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1392{
1393	struct rt6_info *pcpu_rt;
1394
1395	pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1396
1397	if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1398		struct rt6_info *prev, **p;
1399
1400		p = this_cpu_ptr(res->nh->rt6i_pcpu);
1401		prev = xchg(p, NULL);
1402		if (prev) {
1403			dst_dev_put(&prev->dst);
1404			dst_release(&prev->dst);
1405		}
1406
1407		pcpu_rt = NULL;
1408	}
1409
1410	return pcpu_rt;
1411}
1412
1413static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1414					    const struct fib6_result *res)
1415{
1416	struct rt6_info *pcpu_rt, *prev, **p;
1417
1418	pcpu_rt = ip6_rt_pcpu_alloc(res);
1419	if (!pcpu_rt)
1420		return NULL;
1421
1422	p = this_cpu_ptr(res->nh->rt6i_pcpu);
1423	prev = cmpxchg(p, NULL, pcpu_rt);
1424	BUG_ON(prev);
1425
1426	if (res->f6i->fib6_destroying) {
1427		struct fib6_info *from;
1428
1429		from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1430		fib6_info_release(from);
1431	}
1432
1433	return pcpu_rt;
1434}
1435
1436/* exception hash table implementation
1437 */
1438static DEFINE_SPINLOCK(rt6_exception_lock);
1439
1440/* Remove rt6_ex from hash table and free the memory
1441 * Caller must hold rt6_exception_lock
1442 */
1443static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1444				 struct rt6_exception *rt6_ex)
1445{
1446	struct fib6_info *from;
1447	struct net *net;
1448
1449	if (!bucket || !rt6_ex)
1450		return;
1451
1452	net = dev_net(rt6_ex->rt6i->dst.dev);
1453	net->ipv6.rt6_stats->fib_rt_cache--;
1454
1455	/* purge completely the exception to allow releasing the held resources:
1456	 * some [sk] cache may keep the dst around for unlimited time
1457	 */
1458	from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1459	fib6_info_release(from);
1460	dst_dev_put(&rt6_ex->rt6i->dst);
1461
1462	hlist_del_rcu(&rt6_ex->hlist);
1463	dst_release(&rt6_ex->rt6i->dst);
1464	kfree_rcu(rt6_ex, rcu);
1465	WARN_ON_ONCE(!bucket->depth);
1466	bucket->depth--;
1467}
1468
1469/* Remove oldest rt6_ex in bucket and free the memory
1470 * Caller must hold rt6_exception_lock
1471 */
1472static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1473{
1474	struct rt6_exception *rt6_ex, *oldest = NULL;
1475
1476	if (!bucket)
1477		return;
1478
1479	hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1480		if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1481			oldest = rt6_ex;
1482	}
1483	rt6_remove_exception(bucket, oldest);
1484}
1485
1486static u32 rt6_exception_hash(const struct in6_addr *dst,
1487			      const struct in6_addr *src)
1488{
1489	static siphash_aligned_key_t rt6_exception_key;
1490	struct {
1491		struct in6_addr dst;
1492		struct in6_addr src;
1493	} __aligned(SIPHASH_ALIGNMENT) combined = {
1494		.dst = *dst,
1495	};
1496	u64 val;
1497
1498	net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1499
1500#ifdef CONFIG_IPV6_SUBTREES
1501	if (src)
1502		combined.src = *src;
1503#endif
1504	val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1505
1506	return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1507}
1508
1509/* Helper function to find the cached rt in the hash table
1510 * and update bucket pointer to point to the bucket for this
1511 * (daddr, saddr) pair
1512 * Caller must hold rt6_exception_lock
1513 */
1514static struct rt6_exception *
1515__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1516			      const struct in6_addr *daddr,
1517			      const struct in6_addr *saddr)
1518{
1519	struct rt6_exception *rt6_ex;
1520	u32 hval;
1521
1522	if (!(*bucket) || !daddr)
1523		return NULL;
1524
1525	hval = rt6_exception_hash(daddr, saddr);
1526	*bucket += hval;
1527
1528	hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1529		struct rt6_info *rt6 = rt6_ex->rt6i;
1530		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1531
1532#ifdef CONFIG_IPV6_SUBTREES
1533		if (matched && saddr)
1534			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1535#endif
1536		if (matched)
1537			return rt6_ex;
1538	}
1539	return NULL;
1540}
1541
1542/* Helper function to find the cached rt in the hash table
1543 * and update bucket pointer to point to the bucket for this
1544 * (daddr, saddr) pair
1545 * Caller must hold rcu_read_lock()
1546 */
1547static struct rt6_exception *
1548__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1549			 const struct in6_addr *daddr,
1550			 const struct in6_addr *saddr)
1551{
1552	struct rt6_exception *rt6_ex;
1553	u32 hval;
1554
1555	WARN_ON_ONCE(!rcu_read_lock_held());
1556
1557	if (!(*bucket) || !daddr)
1558		return NULL;
1559
1560	hval = rt6_exception_hash(daddr, saddr);
1561	*bucket += hval;
1562
1563	hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1564		struct rt6_info *rt6 = rt6_ex->rt6i;
1565		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1566
1567#ifdef CONFIG_IPV6_SUBTREES
1568		if (matched && saddr)
1569			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1570#endif
1571		if (matched)
1572			return rt6_ex;
1573	}
1574	return NULL;
1575}
1576
1577static unsigned int fib6_mtu(const struct fib6_result *res)
1578{
1579	const struct fib6_nh *nh = res->nh;
1580	unsigned int mtu;
1581
1582	if (res->f6i->fib6_pmtu) {
1583		mtu = res->f6i->fib6_pmtu;
1584	} else {
1585		struct net_device *dev = nh->fib_nh_dev;
1586		struct inet6_dev *idev;
1587
1588		rcu_read_lock();
1589		idev = __in6_dev_get(dev);
1590		mtu = idev->cnf.mtu6;
1591		rcu_read_unlock();
1592	}
1593
1594	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1595
1596	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1597}
1598
1599#define FIB6_EXCEPTION_BUCKET_FLUSHED  0x1UL
1600
1601/* used when the flushed bit is not relevant, only access to the bucket
1602 * (ie., all bucket users except rt6_insert_exception);
1603 *
1604 * called under rcu lock; sometimes called with rt6_exception_lock held
1605 */
1606static
1607struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1608						       spinlock_t *lock)
1609{
1610	struct rt6_exception_bucket *bucket;
1611
1612	if (lock)
1613		bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1614						   lockdep_is_held(lock));
1615	else
1616		bucket = rcu_dereference(nh->rt6i_exception_bucket);
1617
1618	/* remove bucket flushed bit if set */
1619	if (bucket) {
1620		unsigned long p = (unsigned long)bucket;
1621
1622		p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1623		bucket = (struct rt6_exception_bucket *)p;
1624	}
1625
1626	return bucket;
1627}
1628
1629static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1630{
1631	unsigned long p = (unsigned long)bucket;
1632
1633	return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1634}
1635
1636/* called with rt6_exception_lock held */
1637static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1638					      spinlock_t *lock)
1639{
1640	struct rt6_exception_bucket *bucket;
1641	unsigned long p;
1642
1643	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1644					   lockdep_is_held(lock));
1645
1646	p = (unsigned long)bucket;
1647	p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1648	bucket = (struct rt6_exception_bucket *)p;
1649	rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1650}
1651
1652static int rt6_insert_exception(struct rt6_info *nrt,
1653				const struct fib6_result *res)
1654{
1655	struct net *net = dev_net(nrt->dst.dev);
1656	struct rt6_exception_bucket *bucket;
1657	struct fib6_info *f6i = res->f6i;
1658	struct in6_addr *src_key = NULL;
1659	struct rt6_exception *rt6_ex;
1660	struct fib6_nh *nh = res->nh;
1661	int max_depth;
1662	int err = 0;
1663
1664	spin_lock_bh(&rt6_exception_lock);
1665
1666	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1667					  lockdep_is_held(&rt6_exception_lock));
1668	if (!bucket) {
1669		bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1670				 GFP_ATOMIC);
1671		if (!bucket) {
1672			err = -ENOMEM;
1673			goto out;
1674		}
1675		rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1676	} else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1677		err = -EINVAL;
1678		goto out;
1679	}
1680
1681#ifdef CONFIG_IPV6_SUBTREES
1682	/* fib6_src.plen != 0 indicates f6i is in subtree
1683	 * and exception table is indexed by a hash of
1684	 * both fib6_dst and fib6_src.
1685	 * Otherwise, the exception table is indexed by
1686	 * a hash of only fib6_dst.
1687	 */
1688	if (f6i->fib6_src.plen)
1689		src_key = &nrt->rt6i_src.addr;
1690#endif
1691	/* rt6_mtu_change() might lower mtu on f6i.
1692	 * Only insert this exception route if its mtu
1693	 * is less than f6i's mtu value.
1694	 */
1695	if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1696		err = -EINVAL;
1697		goto out;
1698	}
1699
1700	rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1701					       src_key);
1702	if (rt6_ex)
1703		rt6_remove_exception(bucket, rt6_ex);
1704
1705	rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1706	if (!rt6_ex) {
1707		err = -ENOMEM;
1708		goto out;
1709	}
1710	rt6_ex->rt6i = nrt;
1711	rt6_ex->stamp = jiffies;
1712	hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1713	bucket->depth++;
1714	net->ipv6.rt6_stats->fib_rt_cache++;
1715
1716	/* Randomize max depth to avoid some side channels attacks. */
1717	max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH);
1718	while (bucket->depth > max_depth)
1719		rt6_exception_remove_oldest(bucket);
1720
1721out:
1722	spin_unlock_bh(&rt6_exception_lock);
1723
1724	/* Update fn->fn_sernum to invalidate all cached dst */
1725	if (!err) {
1726		spin_lock_bh(&f6i->fib6_table->tb6_lock);
1727		fib6_update_sernum(net, f6i);
1728		spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1729		fib6_force_start_gc(net);
1730	}
1731
1732	return err;
1733}
1734
1735static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1736{
1737	struct rt6_exception_bucket *bucket;
1738	struct rt6_exception *rt6_ex;
1739	struct hlist_node *tmp;
1740	int i;
1741
1742	spin_lock_bh(&rt6_exception_lock);
1743
1744	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1745	if (!bucket)
1746		goto out;
1747
1748	/* Prevent rt6_insert_exception() to recreate the bucket list */
1749	if (!from)
1750		fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1751
1752	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1753		hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1754			if (!from ||
1755			    rcu_access_pointer(rt6_ex->rt6i->from) == from)
1756				rt6_remove_exception(bucket, rt6_ex);
1757		}
1758		WARN_ON_ONCE(!from && bucket->depth);
1759		bucket++;
1760	}
1761out:
1762	spin_unlock_bh(&rt6_exception_lock);
1763}
1764
1765static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1766{
1767	struct fib6_info *f6i = arg;
1768
1769	fib6_nh_flush_exceptions(nh, f6i);
1770
1771	return 0;
1772}
1773
1774void rt6_flush_exceptions(struct fib6_info *f6i)
1775{
1776	if (f6i->nh)
1777		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1778					 f6i);
1779	else
1780		fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1781}
1782
1783/* Find cached rt in the hash table inside passed in rt
1784 * Caller has to hold rcu_read_lock()
1785 */
1786static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1787					   const struct in6_addr *daddr,
1788					   const struct in6_addr *saddr)
1789{
1790	const struct in6_addr *src_key = NULL;
1791	struct rt6_exception_bucket *bucket;
1792	struct rt6_exception *rt6_ex;
1793	struct rt6_info *ret = NULL;
1794
1795#ifdef CONFIG_IPV6_SUBTREES
1796	/* fib6i_src.plen != 0 indicates f6i is in subtree
1797	 * and exception table is indexed by a hash of
1798	 * both fib6_dst and fib6_src.
1799	 * However, the src addr used to create the hash
1800	 * might not be exactly the passed in saddr which
1801	 * is a /128 addr from the flow.
1802	 * So we need to use f6i->fib6_src to redo lookup
1803	 * if the passed in saddr does not find anything.
1804	 * (See the logic in ip6_rt_cache_alloc() on how
1805	 * rt->rt6i_src is updated.)
1806	 */
1807	if (res->f6i->fib6_src.plen)
1808		src_key = saddr;
1809find_ex:
1810#endif
1811	bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1812	rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1813
1814	if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1815		ret = rt6_ex->rt6i;
1816
1817#ifdef CONFIG_IPV6_SUBTREES
1818	/* Use fib6_src as src_key and redo lookup */
1819	if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1820		src_key = &res->f6i->fib6_src.addr;
1821		goto find_ex;
1822	}
1823#endif
1824
1825	return ret;
1826}
1827
1828/* Remove the passed in cached rt from the hash table that contains it */
1829static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1830				    const struct rt6_info *rt)
1831{
1832	const struct in6_addr *src_key = NULL;
1833	struct rt6_exception_bucket *bucket;
1834	struct rt6_exception *rt6_ex;
1835	int err;
1836
1837	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1838		return -ENOENT;
1839
1840	spin_lock_bh(&rt6_exception_lock);
1841	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1842
1843#ifdef CONFIG_IPV6_SUBTREES
1844	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1845	 * and exception table is indexed by a hash of
1846	 * both rt6i_dst and rt6i_src.
1847	 * Otherwise, the exception table is indexed by
1848	 * a hash of only rt6i_dst.
1849	 */
1850	if (plen)
1851		src_key = &rt->rt6i_src.addr;
1852#endif
1853	rt6_ex = __rt6_find_exception_spinlock(&bucket,
1854					       &rt->rt6i_dst.addr,
1855					       src_key);
1856	if (rt6_ex) {
1857		rt6_remove_exception(bucket, rt6_ex);
1858		err = 0;
1859	} else {
1860		err = -ENOENT;
1861	}
1862
1863	spin_unlock_bh(&rt6_exception_lock);
1864	return err;
1865}
1866
1867struct fib6_nh_excptn_arg {
1868	struct rt6_info	*rt;
1869	int		plen;
1870};
1871
1872static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1873{
1874	struct fib6_nh_excptn_arg *arg = _arg;
1875	int err;
1876
1877	err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1878	if (err == 0)
1879		return 1;
1880
1881	return 0;
1882}
1883
1884static int rt6_remove_exception_rt(struct rt6_info *rt)
1885{
1886	struct fib6_info *from;
1887
1888	from = rcu_dereference(rt->from);
1889	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1890		return -EINVAL;
1891
1892	if (from->nh) {
1893		struct fib6_nh_excptn_arg arg = {
1894			.rt = rt,
1895			.plen = from->fib6_src.plen
1896		};
1897		int rc;
1898
1899		/* rc = 1 means an entry was found */
1900		rc = nexthop_for_each_fib6_nh(from->nh,
1901					      rt6_nh_remove_exception_rt,
1902					      &arg);
1903		return rc ? 0 : -ENOENT;
1904	}
1905
1906	return fib6_nh_remove_exception(from->fib6_nh,
1907					from->fib6_src.plen, rt);
1908}
1909
1910/* Find rt6_ex which contains the passed in rt cache and
1911 * refresh its stamp
1912 */
1913static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1914				     const struct rt6_info *rt)
1915{
1916	const struct in6_addr *src_key = NULL;
1917	struct rt6_exception_bucket *bucket;
1918	struct rt6_exception *rt6_ex;
1919
1920	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1921#ifdef CONFIG_IPV6_SUBTREES
1922	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1923	 * and exception table is indexed by a hash of
1924	 * both rt6i_dst and rt6i_src.
1925	 * Otherwise, the exception table is indexed by
1926	 * a hash of only rt6i_dst.
1927	 */
1928	if (plen)
1929		src_key = &rt->rt6i_src.addr;
1930#endif
1931	rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1932	if (rt6_ex)
1933		rt6_ex->stamp = jiffies;
1934}
1935
1936struct fib6_nh_match_arg {
1937	const struct net_device *dev;
1938	const struct in6_addr	*gw;
1939	struct fib6_nh		*match;
1940};
1941
1942/* determine if fib6_nh has given device and gateway */
1943static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1944{
1945	struct fib6_nh_match_arg *arg = _arg;
1946
1947	if (arg->dev != nh->fib_nh_dev ||
1948	    (arg->gw && !nh->fib_nh_gw_family) ||
1949	    (!arg->gw && nh->fib_nh_gw_family) ||
1950	    (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1951		return 0;
1952
1953	arg->match = nh;
1954
1955	/* found a match, break the loop */
1956	return 1;
1957}
1958
1959static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1960{
1961	struct fib6_info *from;
1962	struct fib6_nh *fib6_nh;
1963
1964	rcu_read_lock();
1965
1966	from = rcu_dereference(rt->from);
1967	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1968		goto unlock;
1969
1970	if (from->nh) {
1971		struct fib6_nh_match_arg arg = {
1972			.dev = rt->dst.dev,
1973			.gw = &rt->rt6i_gateway,
1974		};
1975
1976		nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1977
1978		if (!arg.match)
1979			goto unlock;
1980		fib6_nh = arg.match;
1981	} else {
1982		fib6_nh = from->fib6_nh;
1983	}
1984	fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1985unlock:
1986	rcu_read_unlock();
1987}
1988
1989static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1990					 struct rt6_info *rt, int mtu)
1991{
1992	/* If the new MTU is lower than the route PMTU, this new MTU will be the
1993	 * lowest MTU in the path: always allow updating the route PMTU to
1994	 * reflect PMTU decreases.
1995	 *
1996	 * If the new MTU is higher, and the route PMTU is equal to the local
1997	 * MTU, this means the old MTU is the lowest in the path, so allow
1998	 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1999	 * handle this.
2000	 */
2001
2002	if (dst_mtu(&rt->dst) >= mtu)
2003		return true;
2004
2005	if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2006		return true;
2007
2008	return false;
2009}
2010
2011static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2012				       const struct fib6_nh *nh, int mtu)
2013{
2014	struct rt6_exception_bucket *bucket;
2015	struct rt6_exception *rt6_ex;
2016	int i;
2017
2018	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2019	if (!bucket)
2020		return;
2021
2022	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2023		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2024			struct rt6_info *entry = rt6_ex->rt6i;
2025
2026			/* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2027			 * route), the metrics of its rt->from have already
2028			 * been updated.
2029			 */
2030			if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2031			    rt6_mtu_change_route_allowed(idev, entry, mtu))
2032				dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2033		}
2034		bucket++;
2035	}
2036}
2037
2038#define RTF_CACHE_GATEWAY	(RTF_GATEWAY | RTF_CACHE)
2039
2040static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2041					    const struct in6_addr *gateway)
2042{
2043	struct rt6_exception_bucket *bucket;
2044	struct rt6_exception *rt6_ex;
2045	struct hlist_node *tmp;
2046	int i;
2047
2048	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2049		return;
2050
2051	spin_lock_bh(&rt6_exception_lock);
2052	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2053	if (bucket) {
2054		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2055			hlist_for_each_entry_safe(rt6_ex, tmp,
2056						  &bucket->chain, hlist) {
2057				struct rt6_info *entry = rt6_ex->rt6i;
2058
2059				if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2060				    RTF_CACHE_GATEWAY &&
2061				    ipv6_addr_equal(gateway,
2062						    &entry->rt6i_gateway)) {
2063					rt6_remove_exception(bucket, rt6_ex);
2064				}
2065			}
2066			bucket++;
2067		}
2068	}
2069
2070	spin_unlock_bh(&rt6_exception_lock);
2071}
2072
2073static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2074				      struct rt6_exception *rt6_ex,
2075				      struct fib6_gc_args *gc_args,
2076				      unsigned long now)
2077{
2078	struct rt6_info *rt = rt6_ex->rt6i;
2079
2080	/* we are pruning and obsoleting aged-out and non gateway exceptions
2081	 * even if others have still references to them, so that on next
2082	 * dst_check() such references can be dropped.
2083	 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2084	 * expired, independently from their aging, as per RFC 8201 section 4
2085	 */
2086	if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2087		if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2088			RT6_TRACE("aging clone %p\n", rt);
2089			rt6_remove_exception(bucket, rt6_ex);
2090			return;
2091		}
2092	} else if (time_after(jiffies, rt->dst.expires)) {
2093		RT6_TRACE("purging expired route %p\n", rt);
2094		rt6_remove_exception(bucket, rt6_ex);
2095		return;
2096	}
2097
2098	if (rt->rt6i_flags & RTF_GATEWAY) {
2099		struct neighbour *neigh;
2100
2101		neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2102
2103		if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2104			RT6_TRACE("purging route %p via non-router but gateway\n",
2105				  rt);
2106			rt6_remove_exception(bucket, rt6_ex);
2107			return;
2108		}
2109	}
2110
2111	gc_args->more++;
2112}
2113
2114static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2115				   struct fib6_gc_args *gc_args,
2116				   unsigned long now)
2117{
2118	struct rt6_exception_bucket *bucket;
2119	struct rt6_exception *rt6_ex;
2120	struct hlist_node *tmp;
2121	int i;
2122
2123	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2124		return;
2125
2126	rcu_read_lock_bh();
2127	spin_lock(&rt6_exception_lock);
2128	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2129	if (bucket) {
2130		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2131			hlist_for_each_entry_safe(rt6_ex, tmp,
2132						  &bucket->chain, hlist) {
2133				rt6_age_examine_exception(bucket, rt6_ex,
2134							  gc_args, now);
2135			}
2136			bucket++;
2137		}
2138	}
2139	spin_unlock(&rt6_exception_lock);
2140	rcu_read_unlock_bh();
2141}
2142
2143struct fib6_nh_age_excptn_arg {
2144	struct fib6_gc_args	*gc_args;
2145	unsigned long		now;
2146};
2147
2148static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2149{
2150	struct fib6_nh_age_excptn_arg *arg = _arg;
2151
2152	fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2153	return 0;
2154}
2155
2156void rt6_age_exceptions(struct fib6_info *f6i,
2157			struct fib6_gc_args *gc_args,
2158			unsigned long now)
2159{
2160	if (f6i->nh) {
2161		struct fib6_nh_age_excptn_arg arg = {
2162			.gc_args = gc_args,
2163			.now = now
2164		};
2165
2166		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2167					 &arg);
2168	} else {
2169		fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2170	}
2171}
2172
2173/* must be called with rcu lock held */
2174int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2175		      struct flowi6 *fl6, struct fib6_result *res, int strict)
2176{
2177	struct fib6_node *fn, *saved_fn;
2178
2179	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2180	saved_fn = fn;
2181
2182redo_rt6_select:
2183	rt6_select(net, fn, oif, res, strict);
2184	if (res->f6i == net->ipv6.fib6_null_entry) {
2185		fn = fib6_backtrack(fn, &fl6->saddr);
2186		if (fn)
2187			goto redo_rt6_select;
2188		else if (strict & RT6_LOOKUP_F_REACHABLE) {
2189			/* also consider unreachable route */
2190			strict &= ~RT6_LOOKUP_F_REACHABLE;
2191			fn = saved_fn;
2192			goto redo_rt6_select;
2193		}
2194	}
2195
2196	trace_fib6_table_lookup(net, res, table, fl6);
2197
2198	return 0;
2199}
2200
2201struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2202			       int oif, struct flowi6 *fl6,
2203			       const struct sk_buff *skb, int flags)
2204{
2205	struct fib6_result res = {};
2206	struct rt6_info *rt = NULL;
2207	int strict = 0;
2208
2209	WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2210		     !rcu_read_lock_held());
2211
2212	strict |= flags & RT6_LOOKUP_F_IFACE;
2213	strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2214	if (net->ipv6.devconf_all->forwarding == 0)
2215		strict |= RT6_LOOKUP_F_REACHABLE;
2216
2217	rcu_read_lock();
2218
2219	fib6_table_lookup(net, table, oif, fl6, &res, strict);
2220	if (res.f6i == net->ipv6.fib6_null_entry)
2221		goto out;
2222
2223	fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2224
2225	/*Search through exception table */
2226	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2227	if (rt) {
2228		goto out;
2229	} else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2230			    !res.nh->fib_nh_gw_family)) {
2231		/* Create a RTF_CACHE clone which will not be
2232		 * owned by the fib6 tree.  It is for the special case where
2233		 * the daddr in the skb during the neighbor look-up is different
2234		 * from the fl6->daddr used to look-up route here.
2235		 */
2236		rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2237
2238		if (rt) {
2239			/* 1 refcnt is taken during ip6_rt_cache_alloc().
2240			 * As rt6_uncached_list_add() does not consume refcnt,
2241			 * this refcnt is always returned to the caller even
2242			 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2243			 */
2244			rt6_uncached_list_add(rt);
2245			rcu_read_unlock();
2246
2247			return rt;
2248		}
2249	} else {
2250		/* Get a percpu copy */
2251		local_bh_disable();
2252		rt = rt6_get_pcpu_route(&res);
2253
2254		if (!rt)
2255			rt = rt6_make_pcpu_route(net, &res);
2256
2257		local_bh_enable();
2258	}
2259out:
2260	if (!rt)
2261		rt = net->ipv6.ip6_null_entry;
2262	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2263		ip6_hold_safe(net, &rt);
2264	rcu_read_unlock();
2265
2266	return rt;
2267}
2268EXPORT_SYMBOL_GPL(ip6_pol_route);
2269
2270INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2271					    struct fib6_table *table,
2272					    struct flowi6 *fl6,
2273					    const struct sk_buff *skb,
2274					    int flags)
2275{
2276	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2277}
2278
2279struct dst_entry *ip6_route_input_lookup(struct net *net,
2280					 struct net_device *dev,
2281					 struct flowi6 *fl6,
2282					 const struct sk_buff *skb,
2283					 int flags)
2284{
2285	if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2286		flags |= RT6_LOOKUP_F_IFACE;
2287
2288	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2289}
2290EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2291
2292static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2293				  struct flow_keys *keys,
2294				  struct flow_keys *flkeys)
2295{
2296	const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2297	const struct ipv6hdr *key_iph = outer_iph;
2298	struct flow_keys *_flkeys = flkeys;
2299	const struct ipv6hdr *inner_iph;
2300	const struct icmp6hdr *icmph;
2301	struct ipv6hdr _inner_iph;
2302	struct icmp6hdr _icmph;
2303
2304	if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2305		goto out;
2306
2307	icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2308				   sizeof(_icmph), &_icmph);
2309	if (!icmph)
2310		goto out;
2311
2312	if (!icmpv6_is_err(icmph->icmp6_type))
2313		goto out;
2314
2315	inner_iph = skb_header_pointer(skb,
2316				       skb_transport_offset(skb) + sizeof(*icmph),
2317				       sizeof(_inner_iph), &_inner_iph);
2318	if (!inner_iph)
2319		goto out;
2320
2321	key_iph = inner_iph;
2322	_flkeys = NULL;
2323out:
2324	if (_flkeys) {
2325		keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2326		keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2327		keys->tags.flow_label = _flkeys->tags.flow_label;
2328		keys->basic.ip_proto = _flkeys->basic.ip_proto;
2329	} else {
2330		keys->addrs.v6addrs.src = key_iph->saddr;
2331		keys->addrs.v6addrs.dst = key_iph->daddr;
2332		keys->tags.flow_label = ip6_flowlabel(key_iph);
2333		keys->basic.ip_proto = key_iph->nexthdr;
2334	}
2335}
2336
2337static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2338					   const struct sk_buff *skb,
2339					   bool *p_has_inner)
2340{
2341	u32 hash_fields = ip6_multipath_hash_fields(net);
2342	struct flow_keys keys, hash_keys;
2343
2344	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2345		return 0;
2346
2347	memset(&hash_keys, 0, sizeof(hash_keys));
2348	skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2349
2350	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2351	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2352		hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2353	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2354		hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2355	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2356		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2357	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2358		hash_keys.tags.flow_label = keys.tags.flow_label;
2359	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2360		hash_keys.ports.src = keys.ports.src;
2361	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2362		hash_keys.ports.dst = keys.ports.dst;
2363
2364	*p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2365	return flow_hash_from_keys(&hash_keys);
2366}
2367
2368static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2369					   const struct sk_buff *skb,
2370					   bool has_inner)
2371{
2372	u32 hash_fields = ip6_multipath_hash_fields(net);
2373	struct flow_keys keys, hash_keys;
2374
2375	/* We assume the packet carries an encapsulation, but if none was
2376	 * encountered during dissection of the outer flow, then there is no
2377	 * point in calling the flow dissector again.
2378	 */
2379	if (!has_inner)
2380		return 0;
2381
2382	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2383		return 0;
2384
2385	memset(&hash_keys, 0, sizeof(hash_keys));
2386	skb_flow_dissect_flow_keys(skb, &keys, 0);
2387
2388	if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2389		return 0;
2390
2391	if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2392		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2393		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2394			hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2395		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2396			hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2397	} else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2398		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2399		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2400			hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2401		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2402			hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2403		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2404			hash_keys.tags.flow_label = keys.tags.flow_label;
2405	}
2406
2407	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2408		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2409	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2410		hash_keys.ports.src = keys.ports.src;
2411	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2412		hash_keys.ports.dst = keys.ports.dst;
2413
2414	return flow_hash_from_keys(&hash_keys);
2415}
2416
2417static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2418					 const struct sk_buff *skb)
2419{
2420	u32 mhash, mhash_inner;
2421	bool has_inner = true;
2422
2423	mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2424	mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2425
2426	return jhash_2words(mhash, mhash_inner, 0);
2427}
2428
2429static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2430					 const struct flowi6 *fl6)
2431{
2432	u32 hash_fields = ip6_multipath_hash_fields(net);
2433	struct flow_keys hash_keys;
2434
2435	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2436		return 0;
2437
2438	memset(&hash_keys, 0, sizeof(hash_keys));
2439	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2440	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2441		hash_keys.addrs.v6addrs.src = fl6->saddr;
2442	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2443		hash_keys.addrs.v6addrs.dst = fl6->daddr;
2444	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2445		hash_keys.basic.ip_proto = fl6->flowi6_proto;
2446	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2447		hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2448	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2449		hash_keys.ports.src = fl6->fl6_sport;
2450	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2451		hash_keys.ports.dst = fl6->fl6_dport;
2452
2453	return flow_hash_from_keys(&hash_keys);
2454}
2455
2456/* if skb is set it will be used and fl6 can be NULL */
2457u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2458		       const struct sk_buff *skb, struct flow_keys *flkeys)
2459{
2460	struct flow_keys hash_keys;
2461	u32 mhash = 0;
2462
2463	switch (ip6_multipath_hash_policy(net)) {
2464	case 0:
2465		memset(&hash_keys, 0, sizeof(hash_keys));
2466		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2467		if (skb) {
2468			ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2469		} else {
2470			hash_keys.addrs.v6addrs.src = fl6->saddr;
2471			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2472			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2473			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2474		}
2475		mhash = flow_hash_from_keys(&hash_keys);
2476		break;
2477	case 1:
2478		if (skb) {
2479			unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2480			struct flow_keys keys;
2481
2482			/* short-circuit if we already have L4 hash present */
2483			if (skb->l4_hash)
2484				return skb_get_hash_raw(skb) >> 1;
2485
2486			memset(&hash_keys, 0, sizeof(hash_keys));
2487
2488			if (!flkeys) {
2489				skb_flow_dissect_flow_keys(skb, &keys, flag);
2490				flkeys = &keys;
2491			}
2492			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2493			hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2494			hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2495			hash_keys.ports.src = flkeys->ports.src;
2496			hash_keys.ports.dst = flkeys->ports.dst;
2497			hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2498		} else {
2499			memset(&hash_keys, 0, sizeof(hash_keys));
2500			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2501			hash_keys.addrs.v6addrs.src = fl6->saddr;
2502			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2503			hash_keys.ports.src = fl6->fl6_sport;
2504			hash_keys.ports.dst = fl6->fl6_dport;
2505			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2506		}
2507		mhash = flow_hash_from_keys(&hash_keys);
2508		break;
2509	case 2:
2510		memset(&hash_keys, 0, sizeof(hash_keys));
2511		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2512		if (skb) {
2513			struct flow_keys keys;
2514
2515			if (!flkeys) {
2516				skb_flow_dissect_flow_keys(skb, &keys, 0);
2517				flkeys = &keys;
2518			}
2519
2520			/* Inner can be v4 or v6 */
2521			if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2522				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2523				hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2524				hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2525			} else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2526				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2527				hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2528				hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2529				hash_keys.tags.flow_label = flkeys->tags.flow_label;
2530				hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2531			} else {
2532				/* Same as case 0 */
2533				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2534				ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2535			}
2536		} else {
2537			/* Same as case 0 */
2538			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2539			hash_keys.addrs.v6addrs.src = fl6->saddr;
2540			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2541			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2542			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2543		}
2544		mhash = flow_hash_from_keys(&hash_keys);
2545		break;
2546	case 3:
2547		if (skb)
2548			mhash = rt6_multipath_custom_hash_skb(net, skb);
2549		else
2550			mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2551		break;
2552	}
2553
2554	return mhash >> 1;
2555}
2556
2557/* Called with rcu held */
2558void ip6_route_input(struct sk_buff *skb)
2559{
2560	const struct ipv6hdr *iph = ipv6_hdr(skb);
2561	struct net *net = dev_net(skb->dev);
2562	int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2563	struct ip_tunnel_info *tun_info;
2564	struct flowi6 fl6 = {
2565		.flowi6_iif = skb->dev->ifindex,
2566		.daddr = iph->daddr,
2567		.saddr = iph->saddr,
2568		.flowlabel = ip6_flowinfo(iph),
2569		.flowi6_mark = skb->mark,
2570		.flowi6_proto = iph->nexthdr,
2571	};
2572	struct flow_keys *flkeys = NULL, _flkeys;
2573
2574	tun_info = skb_tunnel_info(skb);
2575	if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2576		fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2577
2578	if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2579		flkeys = &_flkeys;
2580
2581	if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2582		fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2583	skb_dst_drop(skb);
2584	skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2585						      &fl6, skb, flags));
2586}
2587
2588INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2589					     struct fib6_table *table,
2590					     struct flowi6 *fl6,
2591					     const struct sk_buff *skb,
2592					     int flags)
2593{
2594	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2595}
2596
2597static struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2598						      const struct sock *sk,
2599						      struct flowi6 *fl6,
2600						      int flags)
2601{
2602	bool any_src;
2603
2604	if (ipv6_addr_type(&fl6->daddr) &
2605	    (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2606		struct dst_entry *dst;
2607
2608		/* This function does not take refcnt on the dst */
2609		dst = l3mdev_link_scope_lookup(net, fl6);
2610		if (dst)
2611			return dst;
2612	}
2613
2614	fl6->flowi6_iif = LOOPBACK_IFINDEX;
2615
2616	flags |= RT6_LOOKUP_F_DST_NOREF;
2617	any_src = ipv6_addr_any(&fl6->saddr);
2618	if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2619	    (fl6->flowi6_oif && any_src))
2620		flags |= RT6_LOOKUP_F_IFACE;
2621
2622	if (!any_src)
2623		flags |= RT6_LOOKUP_F_HAS_SADDR;
2624	else if (sk)
2625		flags |= rt6_srcprefs2flags(READ_ONCE(inet6_sk(sk)->srcprefs));
2626
2627	return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2628}
2629
2630struct dst_entry *ip6_route_output_flags(struct net *net,
2631					 const struct sock *sk,
2632					 struct flowi6 *fl6,
2633					 int flags)
2634{
2635	struct dst_entry *dst;
2636	struct rt6_info *rt6;
2637
2638	rcu_read_lock();
2639	dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2640	rt6 = (struct rt6_info *)dst;
2641	/* For dst cached in uncached_list, refcnt is already taken. */
2642	if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) {
2643		dst = &net->ipv6.ip6_null_entry->dst;
2644		dst_hold(dst);
2645	}
2646	rcu_read_unlock();
2647
2648	return dst;
2649}
2650EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2651
2652struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2653{
2654	struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2655	struct net_device *loopback_dev = net->loopback_dev;
2656	struct dst_entry *new = NULL;
2657
2658	rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev,
2659		       DST_OBSOLETE_DEAD, 0);
2660	if (rt) {
2661		rt6_info_init(rt);
2662		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2663
2664		new = &rt->dst;
2665		new->__use = 1;
2666		new->input = dst_discard;
2667		new->output = dst_discard_out;
2668
2669		dst_copy_metrics(new, &ort->dst);
2670
2671		rt->rt6i_idev = in6_dev_get(loopback_dev);
2672		rt->rt6i_gateway = ort->rt6i_gateway;
2673		rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2674
2675		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2676#ifdef CONFIG_IPV6_SUBTREES
2677		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2678#endif
2679	}
2680
2681	dst_release(dst_orig);
2682	return new ? new : ERR_PTR(-ENOMEM);
2683}
2684
2685/*
2686 *	Destination cache support functions
2687 */
2688
2689static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2690{
2691	u32 rt_cookie = 0;
2692
2693	if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2694		return false;
2695
2696	if (fib6_check_expired(f6i))
2697		return false;
2698
2699	return true;
2700}
2701
2702static struct dst_entry *rt6_check(struct rt6_info *rt,
2703				   struct fib6_info *from,
2704				   u32 cookie)
2705{
2706	u32 rt_cookie = 0;
2707
2708	if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2709	    rt_cookie != cookie)
2710		return NULL;
2711
2712	if (rt6_check_expired(rt))
2713		return NULL;
2714
2715	return &rt->dst;
2716}
2717
2718static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2719					    struct fib6_info *from,
2720					    u32 cookie)
2721{
2722	if (!__rt6_check_expired(rt) &&
2723	    rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2724	    fib6_check(from, cookie))
2725		return &rt->dst;
2726	else
2727		return NULL;
2728}
2729
2730INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2731							u32 cookie)
2732{
2733	struct dst_entry *dst_ret;
2734	struct fib6_info *from;
2735	struct rt6_info *rt;
2736
2737	rt = container_of(dst, struct rt6_info, dst);
2738
2739	if (rt->sernum)
2740		return rt6_is_valid(rt) ? dst : NULL;
2741
2742	rcu_read_lock();
2743
2744	/* All IPV6 dsts are created with ->obsolete set to the value
2745	 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2746	 * into this function always.
2747	 */
2748
2749	from = rcu_dereference(rt->from);
2750
2751	if (from && (rt->rt6i_flags & RTF_PCPU ||
2752	    unlikely(!list_empty(&rt->dst.rt_uncached))))
2753		dst_ret = rt6_dst_from_check(rt, from, cookie);
2754	else
2755		dst_ret = rt6_check(rt, from, cookie);
2756
2757	rcu_read_unlock();
2758
2759	return dst_ret;
2760}
2761EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2762
2763static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2764{
2765	struct rt6_info *rt = (struct rt6_info *) dst;
2766
2767	if (rt) {
2768		if (rt->rt6i_flags & RTF_CACHE) {
2769			rcu_read_lock();
2770			if (rt6_check_expired(rt)) {
2771				rt6_remove_exception_rt(rt);
2772				dst = NULL;
2773			}
2774			rcu_read_unlock();
2775		} else {
2776			dst_release(dst);
2777			dst = NULL;
2778		}
2779	}
2780	return dst;
2781}
2782
2783static void ip6_link_failure(struct sk_buff *skb)
2784{
2785	struct rt6_info *rt;
2786
2787	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2788
2789	rt = (struct rt6_info *) skb_dst(skb);
2790	if (rt) {
2791		rcu_read_lock();
2792		if (rt->rt6i_flags & RTF_CACHE) {
2793			rt6_remove_exception_rt(rt);
2794		} else {
2795			struct fib6_info *from;
2796			struct fib6_node *fn;
2797
2798			from = rcu_dereference(rt->from);
2799			if (from) {
2800				fn = rcu_dereference(from->fib6_node);
2801				if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2802					WRITE_ONCE(fn->fn_sernum, -1);
2803			}
2804		}
2805		rcu_read_unlock();
2806	}
2807}
2808
2809static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2810{
2811	if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2812		struct fib6_info *from;
2813
2814		rcu_read_lock();
2815		from = rcu_dereference(rt0->from);
2816		if (from)
2817			rt0->dst.expires = from->expires;
2818		rcu_read_unlock();
2819	}
2820
2821	dst_set_expires(&rt0->dst, timeout);
2822	rt0->rt6i_flags |= RTF_EXPIRES;
2823}
2824
2825static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2826{
2827	struct net *net = dev_net(rt->dst.dev);
2828
2829	dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2830	rt->rt6i_flags |= RTF_MODIFIED;
2831	rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2832}
2833
2834static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2835{
2836	return !(rt->rt6i_flags & RTF_CACHE) &&
2837		(rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2838}
2839
2840static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2841				 const struct ipv6hdr *iph, u32 mtu,
2842				 bool confirm_neigh)
2843{
2844	const struct in6_addr *daddr, *saddr;
2845	struct rt6_info *rt6 = (struct rt6_info *)dst;
2846
2847	/* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2848	 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2849	 * [see also comment in rt6_mtu_change_route()]
2850	 */
2851
2852	if (iph) {
2853		daddr = &iph->daddr;
2854		saddr = &iph->saddr;
2855	} else if (sk) {
2856		daddr = &sk->sk_v6_daddr;
2857		saddr = &inet6_sk(sk)->saddr;
2858	} else {
2859		daddr = NULL;
2860		saddr = NULL;
2861	}
2862
2863	if (confirm_neigh)
2864		dst_confirm_neigh(dst, daddr);
2865
2866	if (mtu < IPV6_MIN_MTU)
2867		return;
2868	if (mtu >= dst_mtu(dst))
2869		return;
2870
2871	if (!rt6_cache_allowed_for_pmtu(rt6)) {
2872		rt6_do_update_pmtu(rt6, mtu);
2873		/* update rt6_ex->stamp for cache */
2874		if (rt6->rt6i_flags & RTF_CACHE)
2875			rt6_update_exception_stamp_rt(rt6);
2876	} else if (daddr) {
2877		struct fib6_result res = {};
2878		struct rt6_info *nrt6;
2879
2880		rcu_read_lock();
2881		res.f6i = rcu_dereference(rt6->from);
2882		if (!res.f6i)
2883			goto out_unlock;
2884
2885		res.fib6_flags = res.f6i->fib6_flags;
2886		res.fib6_type = res.f6i->fib6_type;
2887
2888		if (res.f6i->nh) {
2889			struct fib6_nh_match_arg arg = {
2890				.dev = dst->dev,
2891				.gw = &rt6->rt6i_gateway,
2892			};
2893
2894			nexthop_for_each_fib6_nh(res.f6i->nh,
2895						 fib6_nh_find_match, &arg);
2896
2897			/* fib6_info uses a nexthop that does not have fib6_nh
2898			 * using the dst->dev + gw. Should be impossible.
2899			 */
2900			if (!arg.match)
2901				goto out_unlock;
2902
2903			res.nh = arg.match;
2904		} else {
2905			res.nh = res.f6i->fib6_nh;
2906		}
2907
2908		nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2909		if (nrt6) {
2910			rt6_do_update_pmtu(nrt6, mtu);
2911			if (rt6_insert_exception(nrt6, &res))
2912				dst_release_immediate(&nrt6->dst);
2913		}
2914out_unlock:
2915		rcu_read_unlock();
2916	}
2917}
2918
2919static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2920			       struct sk_buff *skb, u32 mtu,
2921			       bool confirm_neigh)
2922{
2923	__ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2924			     confirm_neigh);
2925}
2926
2927void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2928		     int oif, u32 mark, kuid_t uid)
2929{
2930	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2931	struct dst_entry *dst;
2932	struct flowi6 fl6 = {
2933		.flowi6_oif = oif,
2934		.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2935		.daddr = iph->daddr,
2936		.saddr = iph->saddr,
2937		.flowlabel = ip6_flowinfo(iph),
2938		.flowi6_uid = uid,
2939	};
2940
2941	dst = ip6_route_output(net, NULL, &fl6);
2942	if (!dst->error)
2943		__ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2944	dst_release(dst);
2945}
2946EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2947
2948void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2949{
2950	int oif = sk->sk_bound_dev_if;
2951	struct dst_entry *dst;
2952
2953	if (!oif && skb->dev)
2954		oif = l3mdev_master_ifindex(skb->dev);
2955
2956	ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark),
2957			sk->sk_uid);
2958
2959	dst = __sk_dst_get(sk);
2960	if (!dst || !dst->obsolete ||
2961	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2962		return;
2963
2964	bh_lock_sock(sk);
2965	if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2966		ip6_datagram_dst_update(sk, false);
2967	bh_unlock_sock(sk);
2968}
2969EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2970
2971void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2972			   const struct flowi6 *fl6)
2973{
2974#ifdef CONFIG_IPV6_SUBTREES
2975	struct ipv6_pinfo *np = inet6_sk(sk);
2976#endif
2977
2978	ip6_dst_store(sk, dst,
2979		      ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2980		      &sk->sk_v6_daddr : NULL,
2981#ifdef CONFIG_IPV6_SUBTREES
2982		      ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2983		      &np->saddr :
2984#endif
2985		      NULL);
2986}
2987
2988static bool ip6_redirect_nh_match(const struct fib6_result *res,
2989				  struct flowi6 *fl6,
2990				  const struct in6_addr *gw,
2991				  struct rt6_info **ret)
2992{
2993	const struct fib6_nh *nh = res->nh;
2994
2995	if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2996	    fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2997		return false;
2998
2999	/* rt_cache's gateway might be different from its 'parent'
3000	 * in the case of an ip redirect.
3001	 * So we keep searching in the exception table if the gateway
3002	 * is different.
3003	 */
3004	if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3005		struct rt6_info *rt_cache;
3006
3007		rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3008		if (rt_cache &&
3009		    ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3010			*ret = rt_cache;
3011			return true;
3012		}
3013		return false;
3014	}
3015	return true;
3016}
3017
3018struct fib6_nh_rd_arg {
3019	struct fib6_result	*res;
3020	struct flowi6		*fl6;
3021	const struct in6_addr	*gw;
3022	struct rt6_info		**ret;
3023};
3024
3025static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3026{
3027	struct fib6_nh_rd_arg *arg = _arg;
3028
3029	arg->res->nh = nh;
3030	return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3031}
3032
3033/* Handle redirects */
3034struct ip6rd_flowi {
3035	struct flowi6 fl6;
3036	struct in6_addr gateway;
3037};
3038
3039INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3040					     struct fib6_table *table,
3041					     struct flowi6 *fl6,
3042					     const struct sk_buff *skb,
3043					     int flags)
3044{
3045	struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3046	struct rt6_info *ret = NULL;
3047	struct fib6_result res = {};
3048	struct fib6_nh_rd_arg arg = {
3049		.res = &res,
3050		.fl6 = fl6,
3051		.gw  = &rdfl->gateway,
3052		.ret = &ret
3053	};
3054	struct fib6_info *rt;
3055	struct fib6_node *fn;
3056
3057	/* Get the "current" route for this destination and
3058	 * check if the redirect has come from appropriate router.
3059	 *
3060	 * RFC 4861 specifies that redirects should only be
3061	 * accepted if they come from the nexthop to the target.
3062	 * Due to the way the routes are chosen, this notion
3063	 * is a bit fuzzy and one might need to check all possible
3064	 * routes.
3065	 */
3066
3067	rcu_read_lock();
3068	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3069restart:
3070	for_each_fib6_node_rt_rcu(fn) {
3071		res.f6i = rt;
3072		if (fib6_check_expired(rt))
3073			continue;
3074		if (rt->fib6_flags & RTF_REJECT)
3075			break;
3076		if (unlikely(rt->nh)) {
3077			if (nexthop_is_blackhole(rt->nh))
3078				continue;
3079			/* on match, res->nh is filled in and potentially ret */
3080			if (nexthop_for_each_fib6_nh(rt->nh,
3081						     fib6_nh_redirect_match,
3082						     &arg))
3083				goto out;
3084		} else {
3085			res.nh = rt->fib6_nh;
3086			if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3087						  &ret))
3088				goto out;
3089		}
3090	}
3091
3092	if (!rt)
3093		rt = net->ipv6.fib6_null_entry;
3094	else if (rt->fib6_flags & RTF_REJECT) {
3095		ret = net->ipv6.ip6_null_entry;
3096		goto out;
3097	}
3098
3099	if (rt == net->ipv6.fib6_null_entry) {
3100		fn = fib6_backtrack(fn, &fl6->saddr);
3101		if (fn)
3102			goto restart;
3103	}
3104
3105	res.f6i = rt;
3106	res.nh = rt->fib6_nh;
3107out:
3108	if (ret) {
3109		ip6_hold_safe(net, &ret);
3110	} else {
3111		res.fib6_flags = res.f6i->fib6_flags;
3112		res.fib6_type = res.f6i->fib6_type;
3113		ret = ip6_create_rt_rcu(&res);
3114	}
3115
3116	rcu_read_unlock();
3117
3118	trace_fib6_table_lookup(net, &res, table, fl6);
3119	return ret;
3120};
3121
3122static struct dst_entry *ip6_route_redirect(struct net *net,
3123					    const struct flowi6 *fl6,
3124					    const struct sk_buff *skb,
3125					    const struct in6_addr *gateway)
3126{
3127	int flags = RT6_LOOKUP_F_HAS_SADDR;
3128	struct ip6rd_flowi rdfl;
3129
3130	rdfl.fl6 = *fl6;
3131	rdfl.gateway = *gateway;
3132
3133	return fib6_rule_lookup(net, &rdfl.fl6, skb,
3134				flags, __ip6_route_redirect);
3135}
3136
3137void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3138		  kuid_t uid)
3139{
3140	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3141	struct dst_entry *dst;
3142	struct flowi6 fl6 = {
3143		.flowi6_iif = LOOPBACK_IFINDEX,
3144		.flowi6_oif = oif,
3145		.flowi6_mark = mark,
3146		.daddr = iph->daddr,
3147		.saddr = iph->saddr,
3148		.flowlabel = ip6_flowinfo(iph),
3149		.flowi6_uid = uid,
3150	};
3151
3152	dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3153	rt6_do_redirect(dst, NULL, skb);
3154	dst_release(dst);
3155}
3156EXPORT_SYMBOL_GPL(ip6_redirect);
3157
3158void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3159{
3160	const struct ipv6hdr *iph = ipv6_hdr(skb);
3161	const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3162	struct dst_entry *dst;
3163	struct flowi6 fl6 = {
3164		.flowi6_iif = LOOPBACK_IFINDEX,
3165		.flowi6_oif = oif,
3166		.daddr = msg->dest,
3167		.saddr = iph->daddr,
3168		.flowi6_uid = sock_net_uid(net, NULL),
3169	};
3170
3171	dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3172	rt6_do_redirect(dst, NULL, skb);
3173	dst_release(dst);
3174}
3175
3176void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3177{
3178	ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if,
3179		     READ_ONCE(sk->sk_mark), sk->sk_uid);
3180}
3181EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3182
3183static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3184{
3185	struct net_device *dev = dst->dev;
3186	unsigned int mtu = dst_mtu(dst);
3187	struct net *net = dev_net(dev);
3188
3189	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3190
3191	if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3192		mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3193
3194	/*
3195	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3196	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3197	 * IPV6_MAXPLEN is also valid and means: "any MSS,
3198	 * rely only on pmtu discovery"
3199	 */
3200	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3201		mtu = IPV6_MAXPLEN;
3202	return mtu;
3203}
3204
3205INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3206{
3207	return ip6_dst_mtu_maybe_forward(dst, false);
3208}
3209EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3210
3211/* MTU selection:
3212 * 1. mtu on route is locked - use it
3213 * 2. mtu from nexthop exception
3214 * 3. mtu from egress device
3215 *
3216 * based on ip6_dst_mtu_forward and exception logic of
3217 * rt6_find_cached_rt; called with rcu_read_lock
3218 */
3219u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3220		      const struct in6_addr *daddr,
3221		      const struct in6_addr *saddr)
3222{
3223	const struct fib6_nh *nh = res->nh;
3224	struct fib6_info *f6i = res->f6i;
3225	struct inet6_dev *idev;
3226	struct rt6_info *rt;
3227	u32 mtu = 0;
3228
3229	if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3230		mtu = f6i->fib6_pmtu;
3231		if (mtu)
3232			goto out;
3233	}
3234
3235	rt = rt6_find_cached_rt(res, daddr, saddr);
3236	if (unlikely(rt)) {
3237		mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3238	} else {
3239		struct net_device *dev = nh->fib_nh_dev;
3240
3241		mtu = IPV6_MIN_MTU;
3242		idev = __in6_dev_get(dev);
3243		if (idev && idev->cnf.mtu6 > mtu)
3244			mtu = idev->cnf.mtu6;
3245	}
3246
3247	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3248out:
3249	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3250}
3251
3252struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3253				  struct flowi6 *fl6)
3254{
3255	struct dst_entry *dst;
3256	struct rt6_info *rt;
3257	struct inet6_dev *idev = in6_dev_get(dev);
3258	struct net *net = dev_net(dev);
3259
3260	if (unlikely(!idev))
3261		return ERR_PTR(-ENODEV);
3262
3263	rt = ip6_dst_alloc(net, dev, 0);
3264	if (unlikely(!rt)) {
3265		in6_dev_put(idev);
3266		dst = ERR_PTR(-ENOMEM);
3267		goto out;
3268	}
3269
3270	rt->dst.input = ip6_input;
3271	rt->dst.output  = ip6_output;
3272	rt->rt6i_gateway  = fl6->daddr;
3273	rt->rt6i_dst.addr = fl6->daddr;
3274	rt->rt6i_dst.plen = 128;
3275	rt->rt6i_idev     = idev;
3276	dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3277
3278	/* Add this dst into uncached_list so that rt6_disable_ip() can
3279	 * do proper release of the net_device
3280	 */
3281	rt6_uncached_list_add(rt);
3282
3283	dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3284
3285out:
3286	return dst;
3287}
3288
3289static void ip6_dst_gc(struct dst_ops *ops)
3290{
3291	struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3292	int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3293	int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3294	int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3295	unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3296	unsigned int val;
3297	int entries;
3298
3299	if (time_after(rt_last_gc + rt_min_interval, jiffies))
3300		goto out;
3301
3302	fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3303	entries = dst_entries_get_slow(ops);
3304	if (entries < ops->gc_thresh)
3305		atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3306out:
3307	val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3308	atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3309}
3310
3311static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3312			       const struct in6_addr *gw_addr, u32 tbid,
3313			       int flags, struct fib6_result *res)
3314{
3315	struct flowi6 fl6 = {
3316		.flowi6_oif = cfg->fc_ifindex,
3317		.daddr = *gw_addr,
3318		.saddr = cfg->fc_prefsrc,
3319	};
3320	struct fib6_table *table;
3321	int err;
3322
3323	table = fib6_get_table(net, tbid);
3324	if (!table)
3325		return -EINVAL;
3326
3327	if (!ipv6_addr_any(&cfg->fc_prefsrc))
3328		flags |= RT6_LOOKUP_F_HAS_SADDR;
3329
3330	flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3331
3332	err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3333	if (!err && res->f6i != net->ipv6.fib6_null_entry)
3334		fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3335				 cfg->fc_ifindex != 0, NULL, flags);
3336
3337	return err;
3338}
3339
3340static int ip6_route_check_nh_onlink(struct net *net,
3341				     struct fib6_config *cfg,
3342				     const struct net_device *dev,
3343				     struct netlink_ext_ack *extack)
3344{
3345	u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3346	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3347	struct fib6_result res = {};
3348	int err;
3349
3350	err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3351	if (!err && !(res.fib6_flags & RTF_REJECT) &&
3352	    /* ignore match if it is the default route */
3353	    !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3354	    (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3355		NL_SET_ERR_MSG(extack,
3356			       "Nexthop has invalid gateway or device mismatch");
3357		err = -EINVAL;
3358	}
3359
3360	return err;
3361}
3362
3363static int ip6_route_check_nh(struct net *net,
3364			      struct fib6_config *cfg,
3365			      struct net_device **_dev,
3366			      netdevice_tracker *dev_tracker,
3367			      struct inet6_dev **idev)
3368{
3369	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3370	struct net_device *dev = _dev ? *_dev : NULL;
3371	int flags = RT6_LOOKUP_F_IFACE;
3372	struct fib6_result res = {};
3373	int err = -EHOSTUNREACH;
3374
3375	if (cfg->fc_table) {
3376		err = ip6_nh_lookup_table(net, cfg, gw_addr,
3377					  cfg->fc_table, flags, &res);
3378		/* gw_addr can not require a gateway or resolve to a reject
3379		 * route. If a device is given, it must match the result.
3380		 */
3381		if (err || res.fib6_flags & RTF_REJECT ||
3382		    res.nh->fib_nh_gw_family ||
3383		    (dev && dev != res.nh->fib_nh_dev))
3384			err = -EHOSTUNREACH;
3385	}
3386
3387	if (err < 0) {
3388		struct flowi6 fl6 = {
3389			.flowi6_oif = cfg->fc_ifindex,
3390			.daddr = *gw_addr,
3391		};
3392
3393		err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3394		if (err || res.fib6_flags & RTF_REJECT ||
3395		    res.nh->fib_nh_gw_family)
3396			err = -EHOSTUNREACH;
3397
3398		if (err)
3399			return err;
3400
3401		fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3402				 cfg->fc_ifindex != 0, NULL, flags);
3403	}
3404
3405	err = 0;
3406	if (dev) {
3407		if (dev != res.nh->fib_nh_dev)
3408			err = -EHOSTUNREACH;
3409	} else {
3410		*_dev = dev = res.nh->fib_nh_dev;
3411		netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3412		*idev = in6_dev_get(dev);
3413	}
3414
3415	return err;
3416}
3417
3418static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3419			   struct net_device **_dev,
3420			   netdevice_tracker *dev_tracker,
3421			   struct inet6_dev **idev,
3422			   struct netlink_ext_ack *extack)
3423{
3424	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3425	int gwa_type = ipv6_addr_type(gw_addr);
3426	bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3427	const struct net_device *dev = *_dev;
3428	bool need_addr_check = !dev;
3429	int err = -EINVAL;
3430
3431	/* if gw_addr is local we will fail to detect this in case
3432	 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3433	 * will return already-added prefix route via interface that
3434	 * prefix route was assigned to, which might be non-loopback.
3435	 */
3436	if (dev &&
3437	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3438		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3439		goto out;
3440	}
3441
3442	if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3443		/* IPv6 strictly inhibits using not link-local
3444		 * addresses as nexthop address.
3445		 * Otherwise, router will not able to send redirects.
3446		 * It is very good, but in some (rare!) circumstances
3447		 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3448		 * some exceptions. --ANK
3449		 * We allow IPv4-mapped nexthops to support RFC4798-type
3450		 * addressing
3451		 */
3452		if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3453			NL_SET_ERR_MSG(extack, "Invalid gateway address");
3454			goto out;
3455		}
3456
3457		rcu_read_lock();
3458
3459		if (cfg->fc_flags & RTNH_F_ONLINK)
3460			err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3461		else
3462			err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3463						 idev);
3464
3465		rcu_read_unlock();
3466
3467		if (err)
3468			goto out;
3469	}
3470
3471	/* reload in case device was changed */
3472	dev = *_dev;
3473
3474	err = -EINVAL;
3475	if (!dev) {
3476		NL_SET_ERR_MSG(extack, "Egress device not specified");
3477		goto out;
3478	} else if (dev->flags & IFF_LOOPBACK) {
3479		NL_SET_ERR_MSG(extack,
3480			       "Egress device can not be loopback device for this route");
3481		goto out;
3482	}
3483
3484	/* if we did not check gw_addr above, do so now that the
3485	 * egress device has been resolved.
3486	 */
3487	if (need_addr_check &&
3488	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3489		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3490		goto out;
3491	}
3492
3493	err = 0;
3494out:
3495	return err;
3496}
3497
3498static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3499{
3500	if ((flags & RTF_REJECT) ||
3501	    (dev && (dev->flags & IFF_LOOPBACK) &&
3502	     !(addr_type & IPV6_ADDR_LOOPBACK) &&
3503	     !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3504		return true;
3505
3506	return false;
3507}
3508
3509int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3510		 struct fib6_config *cfg, gfp_t gfp_flags,
3511		 struct netlink_ext_ack *extack)
3512{
3513	netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
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 = netdev_get_by_index(net, cfg->fc_ifindex,
3532					  dev_tracker, gfp_flags);
3533		if (!dev)
3534			goto out;
3535		idev = in6_dev_get(dev);
3536		if (!idev)
3537			goto out;
3538	}
3539
3540	if (cfg->fc_flags & RTNH_F_ONLINK) {
3541		if (!dev) {
3542			NL_SET_ERR_MSG(extack,
3543				       "Nexthop device required for onlink");
3544			goto out;
3545		}
3546
3547		if (!(dev->flags & IFF_UP)) {
3548			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3549			err = -ENETDOWN;
3550			goto out;
3551		}
3552
3553		fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3554	}
3555
3556	fib6_nh->fib_nh_weight = 1;
3557
3558	/* We cannot add true routes via loopback here,
3559	 * they would result in kernel looping; promote them to reject routes
3560	 */
3561	addr_type = ipv6_addr_type(&cfg->fc_dst);
3562	if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3563		/* hold loopback dev/idev if we haven't done so. */
3564		if (dev != net->loopback_dev) {
3565			if (dev) {
3566				netdev_put(dev, dev_tracker);
3567				in6_dev_put(idev);
3568			}
3569			dev = net->loopback_dev;
3570			netdev_hold(dev, dev_tracker, gfp_flags);
3571			idev = in6_dev_get(dev);
3572			if (!idev) {
3573				err = -ENODEV;
3574				goto out;
3575			}
3576		}
3577		goto pcpu_alloc;
3578	}
3579
3580	if (cfg->fc_flags & RTF_GATEWAY) {
3581		err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3582				      &idev, extack);
3583		if (err)
3584			goto out;
3585
3586		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3587		fib6_nh->fib_nh_gw_family = AF_INET6;
3588	}
3589
3590	err = -ENODEV;
3591	if (!dev)
3592		goto out;
3593
3594	if (idev->cnf.disable_ipv6) {
3595		NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3596		err = -EACCES;
3597		goto out;
3598	}
3599
3600	if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3601		NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3602		err = -ENETDOWN;
3603		goto out;
3604	}
3605
3606	if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3607	    !netif_carrier_ok(dev))
3608		fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3609
3610	err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3611				 cfg->fc_encap_type, cfg, gfp_flags, extack);
3612	if (err)
3613		goto out;
3614
3615pcpu_alloc:
3616	fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3617	if (!fib6_nh->rt6i_pcpu) {
3618		err = -ENOMEM;
3619		goto out;
3620	}
3621
3622	fib6_nh->fib_nh_dev = dev;
3623	fib6_nh->fib_nh_oif = dev->ifindex;
3624	err = 0;
3625out:
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		netdev_put(dev, dev_tracker);
3633	}
3634
3635	return err;
3636}
3637
3638void 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
3659void 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
3678static 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;
3828out:
3829	fib6_info_release(rt);
3830	return ERR_PTR(err);
3831out_free:
3832	ip_fib_metrics_put(rt->fib6_metrics);
3833	kfree(rt);
3834	return ERR_PTR(err);
3835}
3836
3837int 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
3853static 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
3869out:
3870	fib6_info_release(rt);
3871	return err;
3872}
3873
3874int 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
3884static 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);
3951out_unlock:
3952	spin_unlock_bh(&table->tb6_lock);
3953out_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
3963static 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);
3975out:
3976	return rc;
3977}
3978
3979static 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
3995struct fib6_nh_del_cached_rt_arg {
3996	struct fib6_config *cfg;
3997	struct fib6_info *f6i;
3998};
3999
4000static 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
4009static 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
4019static 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
4106static 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
4243out:
4244	rcu_read_unlock();
4245	neigh_release(neigh);
4246}
4247
4248#ifdef CONFIG_IPV6_ROUTE_INFO
4249static 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	}
4284out:
4285	rcu_read_unlock();
4286	return rt;
4287}
4288
4289static 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
4322struct 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
4354struct 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
4386static void __rt6_purge_dflt_routers(struct net *net,
4387				     struct fib6_table *table)
4388{
4389	struct fib6_info *rt;
4390
4391restart:
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
4410void 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
4429static 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
4452int 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
4481static 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
4520static int ip6_pkt_discard(struct sk_buff *skb)
4521{
4522	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4523}
4524
4525static 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
4531static int ip6_pkt_prohibit(struct sk_buff *skb)
4532{
4533	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4534}
4535
4536static 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
4546struct 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				     struct netlink_ext_ack *extack)
4551{
4552	struct fib6_config cfg = {
4553		.fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4554		.fc_ifindex = idev->dev->ifindex,
4555		.fc_flags = RTF_UP | RTF_NONEXTHOP,
4556		.fc_dst = *addr,
4557		.fc_dst_len = 128,
4558		.fc_protocol = RTPROT_KERNEL,
4559		.fc_nlinfo.nl_net = net,
4560		.fc_ignore_dev_down = true,
4561	};
4562	struct fib6_info *f6i;
4563
4564	if (anycast) {
4565		cfg.fc_type = RTN_ANYCAST;
4566		cfg.fc_flags |= RTF_ANYCAST;
4567	} else {
4568		cfg.fc_type = RTN_LOCAL;
4569		cfg.fc_flags |= RTF_LOCAL;
4570	}
4571
4572	f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4573	if (!IS_ERR(f6i)) {
4574		f6i->dst_nocount = true;
4575
4576		if (!anycast &&
4577		    (net->ipv6.devconf_all->disable_policy ||
4578		     idev->cnf.disable_policy))
4579			f6i->dst_nopolicy = true;
4580	}
4581
4582	return f6i;
4583}
4584
4585/* remove deleted ip from prefsrc entries */
4586struct arg_dev_net_ip {
4587	struct net *net;
4588	struct in6_addr *addr;
4589};
4590
4591static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4592{
4593	struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4594	struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4595
4596	if (!rt->nh &&
4597	    rt != net->ipv6.fib6_null_entry &&
4598	    ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4599	    !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4600		spin_lock_bh(&rt6_exception_lock);
4601		/* remove prefsrc entry */
4602		rt->fib6_prefsrc.plen = 0;
4603		spin_unlock_bh(&rt6_exception_lock);
4604	}
4605	return 0;
4606}
4607
4608void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4609{
4610	struct net *net = dev_net(ifp->idev->dev);
4611	struct arg_dev_net_ip adni = {
4612		.net = net,
4613		.addr = &ifp->addr,
4614	};
4615	fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4616}
4617
4618#define RTF_RA_ROUTER		(RTF_ADDRCONF | RTF_DEFAULT)
4619
4620/* Remove routers and update dst entries when gateway turn into host. */
4621static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4622{
4623	struct in6_addr *gateway = (struct in6_addr *)arg;
4624	struct fib6_nh *nh;
4625
4626	/* RA routes do not use nexthops */
4627	if (rt->nh)
4628		return 0;
4629
4630	nh = rt->fib6_nh;
4631	if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4632	    nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4633		return -1;
4634
4635	/* Further clean up cached routes in exception table.
4636	 * This is needed because cached route may have a different
4637	 * gateway than its 'parent' in the case of an ip redirect.
4638	 */
4639	fib6_nh_exceptions_clean_tohost(nh, gateway);
4640
4641	return 0;
4642}
4643
4644void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4645{
4646	fib6_clean_all(net, fib6_clean_tohost, gateway);
4647}
4648
4649struct arg_netdev_event {
4650	const struct net_device *dev;
4651	union {
4652		unsigned char nh_flags;
4653		unsigned long event;
4654	};
4655};
4656
4657static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4658{
4659	struct fib6_info *iter;
4660	struct fib6_node *fn;
4661
4662	fn = rcu_dereference_protected(rt->fib6_node,
4663			lockdep_is_held(&rt->fib6_table->tb6_lock));
4664	iter = rcu_dereference_protected(fn->leaf,
4665			lockdep_is_held(&rt->fib6_table->tb6_lock));
4666	while (iter) {
4667		if (iter->fib6_metric == rt->fib6_metric &&
4668		    rt6_qualify_for_ecmp(iter))
4669			return iter;
4670		iter = rcu_dereference_protected(iter->fib6_next,
4671				lockdep_is_held(&rt->fib6_table->tb6_lock));
4672	}
4673
4674	return NULL;
4675}
4676
4677/* only called for fib entries with builtin fib6_nh */
4678static bool rt6_is_dead(const struct fib6_info *rt)
4679{
4680	if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4681	    (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4682	     ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4683		return true;
4684
4685	return false;
4686}
4687
4688static int rt6_multipath_total_weight(const struct fib6_info *rt)
4689{
4690	struct fib6_info *iter;
4691	int total = 0;
4692
4693	if (!rt6_is_dead(rt))
4694		total += rt->fib6_nh->fib_nh_weight;
4695
4696	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4697		if (!rt6_is_dead(iter))
4698			total += iter->fib6_nh->fib_nh_weight;
4699	}
4700
4701	return total;
4702}
4703
4704static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4705{
4706	int upper_bound = -1;
4707
4708	if (!rt6_is_dead(rt)) {
4709		*weight += rt->fib6_nh->fib_nh_weight;
4710		upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4711						    total) - 1;
4712	}
4713	atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4714}
4715
4716static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4717{
4718	struct fib6_info *iter;
4719	int weight = 0;
4720
4721	rt6_upper_bound_set(rt, &weight, total);
4722
4723	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4724		rt6_upper_bound_set(iter, &weight, total);
4725}
4726
4727void rt6_multipath_rebalance(struct fib6_info *rt)
4728{
4729	struct fib6_info *first;
4730	int total;
4731
4732	/* In case the entire multipath route was marked for flushing,
4733	 * then there is no need to rebalance upon the removal of every
4734	 * sibling route.
4735	 */
4736	if (!rt->fib6_nsiblings || rt->should_flush)
4737		return;
4738
4739	/* During lookup routes are evaluated in order, so we need to
4740	 * make sure upper bounds are assigned from the first sibling
4741	 * onwards.
4742	 */
4743	first = rt6_multipath_first_sibling(rt);
4744	if (WARN_ON_ONCE(!first))
4745		return;
4746
4747	total = rt6_multipath_total_weight(first);
4748	rt6_multipath_upper_bound_set(first, total);
4749}
4750
4751static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4752{
4753	const struct arg_netdev_event *arg = p_arg;
4754	struct net *net = dev_net(arg->dev);
4755
4756	if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4757	    rt->fib6_nh->fib_nh_dev == arg->dev) {
4758		rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4759		fib6_update_sernum_upto_root(net, rt);
4760		rt6_multipath_rebalance(rt);
4761	}
4762
4763	return 0;
4764}
4765
4766void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4767{
4768	struct arg_netdev_event arg = {
4769		.dev = dev,
4770		{
4771			.nh_flags = nh_flags,
4772		},
4773	};
4774
4775	if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4776		arg.nh_flags |= RTNH_F_LINKDOWN;
4777
4778	fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4779}
4780
4781/* only called for fib entries with inline fib6_nh */
4782static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4783				   const struct net_device *dev)
4784{
4785	struct fib6_info *iter;
4786
4787	if (rt->fib6_nh->fib_nh_dev == dev)
4788		return true;
4789	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4790		if (iter->fib6_nh->fib_nh_dev == dev)
4791			return true;
4792
4793	return false;
4794}
4795
4796static void rt6_multipath_flush(struct fib6_info *rt)
4797{
4798	struct fib6_info *iter;
4799
4800	rt->should_flush = 1;
4801	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4802		iter->should_flush = 1;
4803}
4804
4805static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4806					     const struct net_device *down_dev)
4807{
4808	struct fib6_info *iter;
4809	unsigned int dead = 0;
4810
4811	if (rt->fib6_nh->fib_nh_dev == down_dev ||
4812	    rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4813		dead++;
4814	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4815		if (iter->fib6_nh->fib_nh_dev == down_dev ||
4816		    iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4817			dead++;
4818
4819	return dead;
4820}
4821
4822static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4823				       const struct net_device *dev,
4824				       unsigned char nh_flags)
4825{
4826	struct fib6_info *iter;
4827
4828	if (rt->fib6_nh->fib_nh_dev == dev)
4829		rt->fib6_nh->fib_nh_flags |= nh_flags;
4830	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4831		if (iter->fib6_nh->fib_nh_dev == dev)
4832			iter->fib6_nh->fib_nh_flags |= nh_flags;
4833}
4834
4835/* called with write lock held for table with rt */
4836static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4837{
4838	const struct arg_netdev_event *arg = p_arg;
4839	const struct net_device *dev = arg->dev;
4840	struct net *net = dev_net(dev);
4841
4842	if (rt == net->ipv6.fib6_null_entry || rt->nh)
4843		return 0;
4844
4845	switch (arg->event) {
4846	case NETDEV_UNREGISTER:
4847		return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4848	case NETDEV_DOWN:
4849		if (rt->should_flush)
4850			return -1;
4851		if (!rt->fib6_nsiblings)
4852			return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4853		if (rt6_multipath_uses_dev(rt, dev)) {
4854			unsigned int count;
4855
4856			count = rt6_multipath_dead_count(rt, dev);
4857			if (rt->fib6_nsiblings + 1 == count) {
4858				rt6_multipath_flush(rt);
4859				return -1;
4860			}
4861			rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4862						   RTNH_F_LINKDOWN);
4863			fib6_update_sernum(net, rt);
4864			rt6_multipath_rebalance(rt);
4865		}
4866		return -2;
4867	case NETDEV_CHANGE:
4868		if (rt->fib6_nh->fib_nh_dev != dev ||
4869		    rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4870			break;
4871		rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4872		rt6_multipath_rebalance(rt);
4873		break;
4874	}
4875
4876	return 0;
4877}
4878
4879void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4880{
4881	struct arg_netdev_event arg = {
4882		.dev = dev,
4883		{
4884			.event = event,
4885		},
4886	};
4887	struct net *net = dev_net(dev);
4888
4889	if (net->ipv6.sysctl.skip_notify_on_dev_down)
4890		fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4891	else
4892		fib6_clean_all(net, fib6_ifdown, &arg);
4893}
4894
4895void rt6_disable_ip(struct net_device *dev, unsigned long event)
4896{
4897	rt6_sync_down_dev(dev, event);
4898	rt6_uncached_list_flush_dev(dev);
4899	neigh_ifdown(&nd_tbl, dev);
4900}
4901
4902struct rt6_mtu_change_arg {
4903	struct net_device *dev;
4904	unsigned int mtu;
4905	struct fib6_info *f6i;
4906};
4907
4908static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4909{
4910	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4911	struct fib6_info *f6i = arg->f6i;
4912
4913	/* For administrative MTU increase, there is no way to discover
4914	 * IPv6 PMTU increase, so PMTU increase should be updated here.
4915	 * Since RFC 1981 doesn't include administrative MTU increase
4916	 * update PMTU increase is a MUST. (i.e. jumbo frame)
4917	 */
4918	if (nh->fib_nh_dev == arg->dev) {
4919		struct inet6_dev *idev = __in6_dev_get(arg->dev);
4920		u32 mtu = f6i->fib6_pmtu;
4921
4922		if (mtu >= arg->mtu ||
4923		    (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4924			fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4925
4926		spin_lock_bh(&rt6_exception_lock);
4927		rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4928		spin_unlock_bh(&rt6_exception_lock);
4929	}
4930
4931	return 0;
4932}
4933
4934static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4935{
4936	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4937	struct inet6_dev *idev;
4938
4939	/* In IPv6 pmtu discovery is not optional,
4940	   so that RTAX_MTU lock cannot disable it.
4941	   We still use this lock to block changes
4942	   caused by addrconf/ndisc.
4943	*/
4944
4945	idev = __in6_dev_get(arg->dev);
4946	if (!idev)
4947		return 0;
4948
4949	if (fib6_metric_locked(f6i, RTAX_MTU))
4950		return 0;
4951
4952	arg->f6i = f6i;
4953	if (f6i->nh) {
4954		/* fib6_nh_mtu_change only returns 0, so this is safe */
4955		return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4956						arg);
4957	}
4958
4959	return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4960}
4961
4962void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4963{
4964	struct rt6_mtu_change_arg arg = {
4965		.dev = dev,
4966		.mtu = mtu,
4967	};
4968
4969	fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4970}
4971
4972static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4973	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
4974	[RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4975	[RTA_PREFSRC]		= { .len = sizeof(struct in6_addr) },
4976	[RTA_OIF]               = { .type = NLA_U32 },
4977	[RTA_IIF]		= { .type = NLA_U32 },
4978	[RTA_PRIORITY]          = { .type = NLA_U32 },
4979	[RTA_METRICS]           = { .type = NLA_NESTED },
4980	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
4981	[RTA_PREF]              = { .type = NLA_U8 },
4982	[RTA_ENCAP_TYPE]	= { .type = NLA_U16 },
4983	[RTA_ENCAP]		= { .type = NLA_NESTED },
4984	[RTA_EXPIRES]		= { .type = NLA_U32 },
4985	[RTA_UID]		= { .type = NLA_U32 },
4986	[RTA_MARK]		= { .type = NLA_U32 },
4987	[RTA_TABLE]		= { .type = NLA_U32 },
4988	[RTA_IP_PROTO]		= { .type = NLA_U8 },
4989	[RTA_SPORT]		= { .type = NLA_U16 },
4990	[RTA_DPORT]		= { .type = NLA_U16 },
4991	[RTA_NH_ID]		= { .type = NLA_U32 },
4992};
4993
4994static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4995			      struct fib6_config *cfg,
4996			      struct netlink_ext_ack *extack)
4997{
4998	struct rtmsg *rtm;
4999	struct nlattr *tb[RTA_MAX+1];
5000	unsigned int pref;
5001	int err;
5002
5003	err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5004				     rtm_ipv6_policy, extack);
5005	if (err < 0)
5006		goto errout;
5007
5008	err = -EINVAL;
5009	rtm = nlmsg_data(nlh);
5010
5011	if (rtm->rtm_tos) {
5012		NL_SET_ERR_MSG(extack,
5013			       "Invalid dsfield (tos): option not available for IPv6");
5014		goto errout;
5015	}
5016
5017	*cfg = (struct fib6_config){
5018		.fc_table = rtm->rtm_table,
5019		.fc_dst_len = rtm->rtm_dst_len,
5020		.fc_src_len = rtm->rtm_src_len,
5021		.fc_flags = RTF_UP,
5022		.fc_protocol = rtm->rtm_protocol,
5023		.fc_type = rtm->rtm_type,
5024
5025		.fc_nlinfo.portid = NETLINK_CB(skb).portid,
5026		.fc_nlinfo.nlh = nlh,
5027		.fc_nlinfo.nl_net = sock_net(skb->sk),
5028	};
5029
5030	if (rtm->rtm_type == RTN_UNREACHABLE ||
5031	    rtm->rtm_type == RTN_BLACKHOLE ||
5032	    rtm->rtm_type == RTN_PROHIBIT ||
5033	    rtm->rtm_type == RTN_THROW)
5034		cfg->fc_flags |= RTF_REJECT;
5035
5036	if (rtm->rtm_type == RTN_LOCAL)
5037		cfg->fc_flags |= RTF_LOCAL;
5038
5039	if (rtm->rtm_flags & RTM_F_CLONED)
5040		cfg->fc_flags |= RTF_CACHE;
5041
5042	cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5043
5044	if (tb[RTA_NH_ID]) {
5045		if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
5046		    tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5047			NL_SET_ERR_MSG(extack,
5048				       "Nexthop specification and nexthop id are mutually exclusive");
5049			goto errout;
5050		}
5051		cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5052	}
5053
5054	if (tb[RTA_GATEWAY]) {
5055		cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5056		cfg->fc_flags |= RTF_GATEWAY;
5057	}
5058	if (tb[RTA_VIA]) {
5059		NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5060		goto errout;
5061	}
5062
5063	if (tb[RTA_DST]) {
5064		int plen = (rtm->rtm_dst_len + 7) >> 3;
5065
5066		if (nla_len(tb[RTA_DST]) < plen)
5067			goto errout;
5068
5069		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5070	}
5071
5072	if (tb[RTA_SRC]) {
5073		int plen = (rtm->rtm_src_len + 7) >> 3;
5074
5075		if (nla_len(tb[RTA_SRC]) < plen)
5076			goto errout;
5077
5078		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5079	}
5080
5081	if (tb[RTA_PREFSRC])
5082		cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5083
5084	if (tb[RTA_OIF])
5085		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5086
5087	if (tb[RTA_PRIORITY])
5088		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5089
5090	if (tb[RTA_METRICS]) {
5091		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5092		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5093	}
5094
5095	if (tb[RTA_TABLE])
5096		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5097
5098	if (tb[RTA_MULTIPATH]) {
5099		cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5100		cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5101
5102		err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5103						     cfg->fc_mp_len, extack);
5104		if (err < 0)
5105			goto errout;
5106	}
5107
5108	if (tb[RTA_PREF]) {
5109		pref = nla_get_u8(tb[RTA_PREF]);
5110		if (pref != ICMPV6_ROUTER_PREF_LOW &&
5111		    pref != ICMPV6_ROUTER_PREF_HIGH)
5112			pref = ICMPV6_ROUTER_PREF_MEDIUM;
5113		cfg->fc_flags |= RTF_PREF(pref);
5114	}
5115
5116	if (tb[RTA_ENCAP])
5117		cfg->fc_encap = tb[RTA_ENCAP];
5118
5119	if (tb[RTA_ENCAP_TYPE]) {
5120		cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5121
5122		err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5123		if (err < 0)
5124			goto errout;
5125	}
5126
5127	if (tb[RTA_EXPIRES]) {
5128		unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5129
5130		if (addrconf_finite_timeout(timeout)) {
5131			cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5132			cfg->fc_flags |= RTF_EXPIRES;
5133		}
5134	}
5135
5136	err = 0;
5137errout:
5138	return err;
5139}
5140
5141struct rt6_nh {
5142	struct fib6_info *fib6_info;
5143	struct fib6_config r_cfg;
5144	struct list_head next;
5145};
5146
5147static int ip6_route_info_append(struct net *net,
5148				 struct list_head *rt6_nh_list,
5149				 struct fib6_info *rt,
5150				 struct fib6_config *r_cfg)
5151{
5152	struct rt6_nh *nh;
5153	int err = -EEXIST;
5154
5155	list_for_each_entry(nh, rt6_nh_list, next) {
5156		/* check if fib6_info already exists */
5157		if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5158			return err;
5159	}
5160
5161	nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5162	if (!nh)
5163		return -ENOMEM;
5164	nh->fib6_info = rt;
5165	memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5166	list_add_tail(&nh->next, rt6_nh_list);
5167
5168	return 0;
5169}
5170
5171static void ip6_route_mpath_notify(struct fib6_info *rt,
5172				   struct fib6_info *rt_last,
5173				   struct nl_info *info,
5174				   __u16 nlflags)
5175{
5176	/* if this is an APPEND route, then rt points to the first route
5177	 * inserted and rt_last points to last route inserted. Userspace
5178	 * wants a consistent dump of the route which starts at the first
5179	 * nexthop. Since sibling routes are always added at the end of
5180	 * the list, find the first sibling of the last route appended
5181	 */
5182	if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5183		rt = list_first_entry(&rt_last->fib6_siblings,
5184				      struct fib6_info,
5185				      fib6_siblings);
5186	}
5187
5188	if (rt)
5189		inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5190}
5191
5192static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5193{
5194	bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5195	bool should_notify = false;
5196	struct fib6_info *leaf;
5197	struct fib6_node *fn;
5198
5199	rcu_read_lock();
5200	fn = rcu_dereference(rt->fib6_node);
5201	if (!fn)
5202		goto out;
5203
5204	leaf = rcu_dereference(fn->leaf);
5205	if (!leaf)
5206		goto out;
5207
5208	if (rt == leaf ||
5209	    (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5210	     rt6_qualify_for_ecmp(leaf)))
5211		should_notify = true;
5212out:
5213	rcu_read_unlock();
5214
5215	return should_notify;
5216}
5217
5218static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5219			     struct netlink_ext_ack *extack)
5220{
5221	if (nla_len(nla) < sizeof(*gw)) {
5222		NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5223		return -EINVAL;
5224	}
5225
5226	*gw = nla_get_in6_addr(nla);
5227
5228	return 0;
5229}
5230
5231static int ip6_route_multipath_add(struct fib6_config *cfg,
5232				   struct netlink_ext_ack *extack)
5233{
5234	struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5235	struct nl_info *info = &cfg->fc_nlinfo;
5236	struct fib6_config r_cfg;
5237	struct rtnexthop *rtnh;
5238	struct fib6_info *rt;
5239	struct rt6_nh *err_nh;
5240	struct rt6_nh *nh, *nh_safe;
5241	__u16 nlflags;
5242	int remaining;
5243	int attrlen;
5244	int err = 1;
5245	int nhn = 0;
5246	int replace = (cfg->fc_nlinfo.nlh &&
5247		       (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5248	LIST_HEAD(rt6_nh_list);
5249
5250	nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5251	if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5252		nlflags |= NLM_F_APPEND;
5253
5254	remaining = cfg->fc_mp_len;
5255	rtnh = (struct rtnexthop *)cfg->fc_mp;
5256
5257	/* Parse a Multipath Entry and build a list (rt6_nh_list) of
5258	 * fib6_info structs per nexthop
5259	 */
5260	while (rtnh_ok(rtnh, remaining)) {
5261		memcpy(&r_cfg, cfg, sizeof(*cfg));
5262		if (rtnh->rtnh_ifindex)
5263			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5264
5265		attrlen = rtnh_attrlen(rtnh);
5266		if (attrlen > 0) {
5267			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5268
5269			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5270			if (nla) {
5271				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5272							extack);
5273				if (err)
5274					goto cleanup;
5275
5276				r_cfg.fc_flags |= RTF_GATEWAY;
5277			}
5278			r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5279
5280			/* RTA_ENCAP_TYPE length checked in
5281			 * lwtunnel_valid_encap_type_attr
5282			 */
5283			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5284			if (nla)
5285				r_cfg.fc_encap_type = nla_get_u16(nla);
5286		}
5287
5288		r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5289		rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5290		if (IS_ERR(rt)) {
5291			err = PTR_ERR(rt);
5292			rt = NULL;
5293			goto cleanup;
5294		}
5295		if (!rt6_qualify_for_ecmp(rt)) {
5296			err = -EINVAL;
5297			NL_SET_ERR_MSG(extack,
5298				       "Device only routes can not be added for IPv6 using the multipath API.");
5299			fib6_info_release(rt);
5300			goto cleanup;
5301		}
5302
5303		rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5304
5305		err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5306					    rt, &r_cfg);
5307		if (err) {
5308			fib6_info_release(rt);
5309			goto cleanup;
5310		}
5311
5312		rtnh = rtnh_next(rtnh, &remaining);
5313	}
5314
5315	if (list_empty(&rt6_nh_list)) {
5316		NL_SET_ERR_MSG(extack,
5317			       "Invalid nexthop configuration - no valid nexthops");
5318		return -EINVAL;
5319	}
5320
5321	/* for add and replace send one notification with all nexthops.
5322	 * Skip the notification in fib6_add_rt2node and send one with
5323	 * the full route when done
5324	 */
5325	info->skip_notify = 1;
5326
5327	/* For add and replace, send one notification with all nexthops. For
5328	 * append, send one notification with all appended nexthops.
5329	 */
5330	info->skip_notify_kernel = 1;
5331
5332	err_nh = NULL;
5333	list_for_each_entry(nh, &rt6_nh_list, next) {
5334		err = __ip6_ins_rt(nh->fib6_info, info, extack);
5335
5336		if (err) {
5337			if (replace && nhn)
5338				NL_SET_ERR_MSG_MOD(extack,
5339						   "multipath route replace failed (check consistency of installed routes)");
5340			err_nh = nh;
5341			goto add_errout;
5342		}
5343		/* save reference to last route successfully inserted */
5344		rt_last = nh->fib6_info;
5345
5346		/* save reference to first route for notification */
5347		if (!rt_notif)
5348			rt_notif = nh->fib6_info;
5349
5350		/* Because each route is added like a single route we remove
5351		 * these flags after the first nexthop: if there is a collision,
5352		 * we have already failed to add the first nexthop:
5353		 * fib6_add_rt2node() has rejected it; when replacing, old
5354		 * nexthops have been replaced by first new, the rest should
5355		 * be added to it.
5356		 */
5357		if (cfg->fc_nlinfo.nlh) {
5358			cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5359							     NLM_F_REPLACE);
5360			cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5361		}
5362		nhn++;
5363	}
5364
5365	/* An in-kernel notification should only be sent in case the new
5366	 * multipath route is added as the first route in the node, or if
5367	 * it was appended to it. We pass 'rt_notif' since it is the first
5368	 * sibling and might allow us to skip some checks in the replace case.
5369	 */
5370	if (ip6_route_mpath_should_notify(rt_notif)) {
5371		enum fib_event_type fib_event;
5372
5373		if (rt_notif->fib6_nsiblings != nhn - 1)
5374			fib_event = FIB_EVENT_ENTRY_APPEND;
5375		else
5376			fib_event = FIB_EVENT_ENTRY_REPLACE;
5377
5378		err = call_fib6_multipath_entry_notifiers(info->nl_net,
5379							  fib_event, rt_notif,
5380							  nhn - 1, extack);
5381		if (err) {
5382			/* Delete all the siblings that were just added */
5383			err_nh = NULL;
5384			goto add_errout;
5385		}
5386	}
5387
5388	/* success ... tell user about new route */
5389	ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5390	goto cleanup;
5391
5392add_errout:
5393	/* send notification for routes that were added so that
5394	 * the delete notifications sent by ip6_route_del are
5395	 * coherent
5396	 */
5397	if (rt_notif)
5398		ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5399
5400	/* Delete routes that were already added */
5401	list_for_each_entry(nh, &rt6_nh_list, next) {
5402		if (err_nh == nh)
5403			break;
5404		ip6_route_del(&nh->r_cfg, extack);
5405	}
5406
5407cleanup:
5408	list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5409		fib6_info_release(nh->fib6_info);
5410		list_del(&nh->next);
5411		kfree(nh);
5412	}
5413
5414	return err;
5415}
5416
5417static int ip6_route_multipath_del(struct fib6_config *cfg,
5418				   struct netlink_ext_ack *extack)
5419{
5420	struct fib6_config r_cfg;
5421	struct rtnexthop *rtnh;
5422	int last_err = 0;
5423	int remaining;
5424	int attrlen;
5425	int err;
5426
5427	remaining = cfg->fc_mp_len;
5428	rtnh = (struct rtnexthop *)cfg->fc_mp;
5429
5430	/* Parse a Multipath Entry */
5431	while (rtnh_ok(rtnh, remaining)) {
5432		memcpy(&r_cfg, cfg, sizeof(*cfg));
5433		if (rtnh->rtnh_ifindex)
5434			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5435
5436		attrlen = rtnh_attrlen(rtnh);
5437		if (attrlen > 0) {
5438			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5439
5440			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5441			if (nla) {
5442				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5443							extack);
5444				if (err) {
5445					last_err = err;
5446					goto next_rtnh;
5447				}
5448
5449				r_cfg.fc_flags |= RTF_GATEWAY;
5450			}
5451		}
5452		err = ip6_route_del(&r_cfg, extack);
5453		if (err)
5454			last_err = err;
5455
5456next_rtnh:
5457		rtnh = rtnh_next(rtnh, &remaining);
5458	}
5459
5460	return last_err;
5461}
5462
5463static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5464			      struct netlink_ext_ack *extack)
5465{
5466	struct fib6_config cfg;
5467	int err;
5468
5469	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5470	if (err < 0)
5471		return err;
5472
5473	if (cfg.fc_nh_id &&
5474	    !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5475		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5476		return -EINVAL;
5477	}
5478
5479	if (cfg.fc_mp)
5480		return ip6_route_multipath_del(&cfg, extack);
5481	else {
5482		cfg.fc_delete_all_nh = 1;
5483		return ip6_route_del(&cfg, extack);
5484	}
5485}
5486
5487static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5488			      struct netlink_ext_ack *extack)
5489{
5490	struct fib6_config cfg;
5491	int err;
5492
5493	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5494	if (err < 0)
5495		return err;
5496
5497	if (cfg.fc_metric == 0)
5498		cfg.fc_metric = IP6_RT_PRIO_USER;
5499
5500	if (cfg.fc_mp)
5501		return ip6_route_multipath_add(&cfg, extack);
5502	else
5503		return ip6_route_add(&cfg, GFP_KERNEL, extack);
5504}
5505
5506/* add the overhead of this fib6_nh to nexthop_len */
5507static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5508{
5509	int *nexthop_len = arg;
5510
5511	*nexthop_len += nla_total_size(0)	 /* RTA_MULTIPATH */
5512		     + NLA_ALIGN(sizeof(struct rtnexthop))
5513		     + nla_total_size(16); /* RTA_GATEWAY */
5514
5515	if (nh->fib_nh_lws) {
5516		/* RTA_ENCAP_TYPE */
5517		*nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5518		/* RTA_ENCAP */
5519		*nexthop_len += nla_total_size(2);
5520	}
5521
5522	return 0;
5523}
5524
5525static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5526{
5527	int nexthop_len;
5528
5529	if (f6i->nh) {
5530		nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5531		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5532					 &nexthop_len);
5533	} else {
5534		struct fib6_info *sibling, *next_sibling;
5535		struct fib6_nh *nh = f6i->fib6_nh;
5536
5537		nexthop_len = 0;
5538		if (f6i->fib6_nsiblings) {
5539			rt6_nh_nlmsg_size(nh, &nexthop_len);
5540
5541			list_for_each_entry_safe(sibling, next_sibling,
5542						 &f6i->fib6_siblings, fib6_siblings) {
5543				rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5544			}
5545		}
5546		nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5547	}
5548
5549	return NLMSG_ALIGN(sizeof(struct rtmsg))
5550	       + nla_total_size(16) /* RTA_SRC */
5551	       + nla_total_size(16) /* RTA_DST */
5552	       + nla_total_size(16) /* RTA_GATEWAY */
5553	       + nla_total_size(16) /* RTA_PREFSRC */
5554	       + nla_total_size(4) /* RTA_TABLE */
5555	       + nla_total_size(4) /* RTA_IIF */
5556	       + nla_total_size(4) /* RTA_OIF */
5557	       + nla_total_size(4) /* RTA_PRIORITY */
5558	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5559	       + nla_total_size(sizeof(struct rta_cacheinfo))
5560	       + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5561	       + nla_total_size(1) /* RTA_PREF */
5562	       + nexthop_len;
5563}
5564
5565static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5566				 unsigned char *flags)
5567{
5568	if (nexthop_is_multipath(nh)) {
5569		struct nlattr *mp;
5570
5571		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5572		if (!mp)
5573			goto nla_put_failure;
5574
5575		if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5576			goto nla_put_failure;
5577
5578		nla_nest_end(skb, mp);
5579	} else {
5580		struct fib6_nh *fib6_nh;
5581
5582		fib6_nh = nexthop_fib6_nh(nh);
5583		if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5584				     flags, false) < 0)
5585			goto nla_put_failure;
5586	}
5587
5588	return 0;
5589
5590nla_put_failure:
5591	return -EMSGSIZE;
5592}
5593
5594static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5595			 struct fib6_info *rt, struct dst_entry *dst,
5596			 struct in6_addr *dest, struct in6_addr *src,
5597			 int iif, int type, u32 portid, u32 seq,
5598			 unsigned int flags)
5599{
5600	struct rt6_info *rt6 = (struct rt6_info *)dst;
5601	struct rt6key *rt6_dst, *rt6_src;
5602	u32 *pmetrics, table, rt6_flags;
5603	unsigned char nh_flags = 0;
5604	struct nlmsghdr *nlh;
5605	struct rtmsg *rtm;
5606	long expires = 0;
5607
5608	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5609	if (!nlh)
5610		return -EMSGSIZE;
5611
5612	if (rt6) {
5613		rt6_dst = &rt6->rt6i_dst;
5614		rt6_src = &rt6->rt6i_src;
5615		rt6_flags = rt6->rt6i_flags;
5616	} else {
5617		rt6_dst = &rt->fib6_dst;
5618		rt6_src = &rt->fib6_src;
5619		rt6_flags = rt->fib6_flags;
5620	}
5621
5622	rtm = nlmsg_data(nlh);
5623	rtm->rtm_family = AF_INET6;
5624	rtm->rtm_dst_len = rt6_dst->plen;
5625	rtm->rtm_src_len = rt6_src->plen;
5626	rtm->rtm_tos = 0;
5627	if (rt->fib6_table)
5628		table = rt->fib6_table->tb6_id;
5629	else
5630		table = RT6_TABLE_UNSPEC;
5631	rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5632	if (nla_put_u32(skb, RTA_TABLE, table))
5633		goto nla_put_failure;
5634
5635	rtm->rtm_type = rt->fib6_type;
5636	rtm->rtm_flags = 0;
5637	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5638	rtm->rtm_protocol = rt->fib6_protocol;
5639
5640	if (rt6_flags & RTF_CACHE)
5641		rtm->rtm_flags |= RTM_F_CLONED;
5642
5643	if (dest) {
5644		if (nla_put_in6_addr(skb, RTA_DST, dest))
5645			goto nla_put_failure;
5646		rtm->rtm_dst_len = 128;
5647	} else if (rtm->rtm_dst_len)
5648		if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5649			goto nla_put_failure;
5650#ifdef CONFIG_IPV6_SUBTREES
5651	if (src) {
5652		if (nla_put_in6_addr(skb, RTA_SRC, src))
5653			goto nla_put_failure;
5654		rtm->rtm_src_len = 128;
5655	} else if (rtm->rtm_src_len &&
5656		   nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5657		goto nla_put_failure;
5658#endif
5659	if (iif) {
5660#ifdef CONFIG_IPV6_MROUTE
5661		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5662			int err = ip6mr_get_route(net, skb, rtm, portid);
5663
5664			if (err == 0)
5665				return 0;
5666			if (err < 0)
5667				goto nla_put_failure;
5668		} else
5669#endif
5670			if (nla_put_u32(skb, RTA_IIF, iif))
5671				goto nla_put_failure;
5672	} else if (dest) {
5673		struct in6_addr saddr_buf;
5674		if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5675		    nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5676			goto nla_put_failure;
5677	}
5678
5679	if (rt->fib6_prefsrc.plen) {
5680		struct in6_addr saddr_buf;
5681		saddr_buf = rt->fib6_prefsrc.addr;
5682		if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5683			goto nla_put_failure;
5684	}
5685
5686	pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5687	if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5688		goto nla_put_failure;
5689
5690	if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5691		goto nla_put_failure;
5692
5693	/* For multipath routes, walk the siblings list and add
5694	 * each as a nexthop within RTA_MULTIPATH.
5695	 */
5696	if (rt6) {
5697		if (rt6_flags & RTF_GATEWAY &&
5698		    nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5699			goto nla_put_failure;
5700
5701		if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5702			goto nla_put_failure;
5703
5704		if (dst->lwtstate &&
5705		    lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5706			goto nla_put_failure;
5707	} else if (rt->fib6_nsiblings) {
5708		struct fib6_info *sibling, *next_sibling;
5709		struct nlattr *mp;
5710
5711		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5712		if (!mp)
5713			goto nla_put_failure;
5714
5715		if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5716				    rt->fib6_nh->fib_nh_weight, AF_INET6,
5717				    0) < 0)
5718			goto nla_put_failure;
5719
5720		list_for_each_entry_safe(sibling, next_sibling,
5721					 &rt->fib6_siblings, fib6_siblings) {
5722			if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5723					    sibling->fib6_nh->fib_nh_weight,
5724					    AF_INET6, 0) < 0)
5725				goto nla_put_failure;
5726		}
5727
5728		nla_nest_end(skb, mp);
5729	} else if (rt->nh) {
5730		if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5731			goto nla_put_failure;
5732
5733		if (nexthop_is_blackhole(rt->nh))
5734			rtm->rtm_type = RTN_BLACKHOLE;
5735
5736		if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5737		    rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5738			goto nla_put_failure;
5739
5740		rtm->rtm_flags |= nh_flags;
5741	} else {
5742		if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5743				     &nh_flags, false) < 0)
5744			goto nla_put_failure;
5745
5746		rtm->rtm_flags |= nh_flags;
5747	}
5748
5749	if (rt6_flags & RTF_EXPIRES) {
5750		expires = dst ? dst->expires : rt->expires;
5751		expires -= jiffies;
5752	}
5753
5754	if (!dst) {
5755		if (READ_ONCE(rt->offload))
5756			rtm->rtm_flags |= RTM_F_OFFLOAD;
5757		if (READ_ONCE(rt->trap))
5758			rtm->rtm_flags |= RTM_F_TRAP;
5759		if (READ_ONCE(rt->offload_failed))
5760			rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5761	}
5762
5763	if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5764		goto nla_put_failure;
5765
5766	if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5767		goto nla_put_failure;
5768
5769
5770	nlmsg_end(skb, nlh);
5771	return 0;
5772
5773nla_put_failure:
5774	nlmsg_cancel(skb, nlh);
5775	return -EMSGSIZE;
5776}
5777
5778static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5779{
5780	const struct net_device *dev = arg;
5781
5782	if (nh->fib_nh_dev == dev)
5783		return 1;
5784
5785	return 0;
5786}
5787
5788static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5789			       const struct net_device *dev)
5790{
5791	if (f6i->nh) {
5792		struct net_device *_dev = (struct net_device *)dev;
5793
5794		return !!nexthop_for_each_fib6_nh(f6i->nh,
5795						  fib6_info_nh_uses_dev,
5796						  _dev);
5797	}
5798
5799	if (f6i->fib6_nh->fib_nh_dev == dev)
5800		return true;
5801
5802	if (f6i->fib6_nsiblings) {
5803		struct fib6_info *sibling, *next_sibling;
5804
5805		list_for_each_entry_safe(sibling, next_sibling,
5806					 &f6i->fib6_siblings, fib6_siblings) {
5807			if (sibling->fib6_nh->fib_nh_dev == dev)
5808				return true;
5809		}
5810	}
5811
5812	return false;
5813}
5814
5815struct fib6_nh_exception_dump_walker {
5816	struct rt6_rtnl_dump_arg *dump;
5817	struct fib6_info *rt;
5818	unsigned int flags;
5819	unsigned int skip;
5820	unsigned int count;
5821};
5822
5823static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5824{
5825	struct fib6_nh_exception_dump_walker *w = arg;
5826	struct rt6_rtnl_dump_arg *dump = w->dump;
5827	struct rt6_exception_bucket *bucket;
5828	struct rt6_exception *rt6_ex;
5829	int i, err;
5830
5831	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5832	if (!bucket)
5833		return 0;
5834
5835	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5836		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5837			if (w->skip) {
5838				w->skip--;
5839				continue;
5840			}
5841
5842			/* Expiration of entries doesn't bump sernum, insertion
5843			 * does. Removal is triggered by insertion, so we can
5844			 * rely on the fact that if entries change between two
5845			 * partial dumps, this node is scanned again completely,
5846			 * see rt6_insert_exception() and fib6_dump_table().
5847			 *
5848			 * Count expired entries we go through as handled
5849			 * entries that we'll skip next time, in case of partial
5850			 * node dump. Otherwise, if entries expire meanwhile,
5851			 * we'll skip the wrong amount.
5852			 */
5853			if (rt6_check_expired(rt6_ex->rt6i)) {
5854				w->count++;
5855				continue;
5856			}
5857
5858			err = rt6_fill_node(dump->net, dump->skb, w->rt,
5859					    &rt6_ex->rt6i->dst, NULL, NULL, 0,
5860					    RTM_NEWROUTE,
5861					    NETLINK_CB(dump->cb->skb).portid,
5862					    dump->cb->nlh->nlmsg_seq, w->flags);
5863			if (err)
5864				return err;
5865
5866			w->count++;
5867		}
5868		bucket++;
5869	}
5870
5871	return 0;
5872}
5873
5874/* Return -1 if done with node, number of handled routes on partial dump */
5875int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5876{
5877	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5878	struct fib_dump_filter *filter = &arg->filter;
5879	unsigned int flags = NLM_F_MULTI;
5880	struct net *net = arg->net;
5881	int count = 0;
5882
5883	if (rt == net->ipv6.fib6_null_entry)
5884		return -1;
5885
5886	if ((filter->flags & RTM_F_PREFIX) &&
5887	    !(rt->fib6_flags & RTF_PREFIX_RT)) {
5888		/* success since this is not a prefix route */
5889		return -1;
5890	}
5891	if (filter->filter_set &&
5892	    ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5893	     (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5894	     (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5895		return -1;
5896	}
5897
5898	if (filter->filter_set ||
5899	    !filter->dump_routes || !filter->dump_exceptions) {
5900		flags |= NLM_F_DUMP_FILTERED;
5901	}
5902
5903	if (filter->dump_routes) {
5904		if (skip) {
5905			skip--;
5906		} else {
5907			if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5908					  0, RTM_NEWROUTE,
5909					  NETLINK_CB(arg->cb->skb).portid,
5910					  arg->cb->nlh->nlmsg_seq, flags)) {
5911				return 0;
5912			}
5913			count++;
5914		}
5915	}
5916
5917	if (filter->dump_exceptions) {
5918		struct fib6_nh_exception_dump_walker w = { .dump = arg,
5919							   .rt = rt,
5920							   .flags = flags,
5921							   .skip = skip,
5922							   .count = 0 };
5923		int err;
5924
5925		rcu_read_lock();
5926		if (rt->nh) {
5927			err = nexthop_for_each_fib6_nh(rt->nh,
5928						       rt6_nh_dump_exceptions,
5929						       &w);
5930		} else {
5931			err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5932		}
5933		rcu_read_unlock();
5934
5935		if (err)
5936			return count + w.count;
5937	}
5938
5939	return -1;
5940}
5941
5942static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5943					const struct nlmsghdr *nlh,
5944					struct nlattr **tb,
5945					struct netlink_ext_ack *extack)
5946{
5947	struct rtmsg *rtm;
5948	int i, err;
5949
5950	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5951		NL_SET_ERR_MSG_MOD(extack,
5952				   "Invalid header for get route request");
5953		return -EINVAL;
5954	}
5955
5956	if (!netlink_strict_get_check(skb))
5957		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5958					      rtm_ipv6_policy, extack);
5959
5960	rtm = nlmsg_data(nlh);
5961	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5962	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5963	    rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5964	    rtm->rtm_type) {
5965		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5966		return -EINVAL;
5967	}
5968	if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5969		NL_SET_ERR_MSG_MOD(extack,
5970				   "Invalid flags for get route request");
5971		return -EINVAL;
5972	}
5973
5974	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5975					    rtm_ipv6_policy, extack);
5976	if (err)
5977		return err;
5978
5979	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5980	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5981		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5982		return -EINVAL;
5983	}
5984
5985	for (i = 0; i <= RTA_MAX; i++) {
5986		if (!tb[i])
5987			continue;
5988
5989		switch (i) {
5990		case RTA_SRC:
5991		case RTA_DST:
5992		case RTA_IIF:
5993		case RTA_OIF:
5994		case RTA_MARK:
5995		case RTA_UID:
5996		case RTA_SPORT:
5997		case RTA_DPORT:
5998		case RTA_IP_PROTO:
5999			break;
6000		default:
6001			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6002			return -EINVAL;
6003		}
6004	}
6005
6006	return 0;
6007}
6008
6009static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6010			      struct netlink_ext_ack *extack)
6011{
6012	struct net *net = sock_net(in_skb->sk);
6013	struct nlattr *tb[RTA_MAX+1];
6014	int err, iif = 0, oif = 0;
6015	struct fib6_info *from;
6016	struct dst_entry *dst;
6017	struct rt6_info *rt;
6018	struct sk_buff *skb;
6019	struct rtmsg *rtm;
6020	struct flowi6 fl6 = {};
6021	bool fibmatch;
6022
6023	err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6024	if (err < 0)
6025		goto errout;
6026
6027	err = -EINVAL;
6028	rtm = nlmsg_data(nlh);
6029	fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6030	fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6031
6032	if (tb[RTA_SRC]) {
6033		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6034			goto errout;
6035
6036		fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6037	}
6038
6039	if (tb[RTA_DST]) {
6040		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6041			goto errout;
6042
6043		fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6044	}
6045
6046	if (tb[RTA_IIF])
6047		iif = nla_get_u32(tb[RTA_IIF]);
6048
6049	if (tb[RTA_OIF])
6050		oif = nla_get_u32(tb[RTA_OIF]);
6051
6052	if (tb[RTA_MARK])
6053		fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6054
6055	if (tb[RTA_UID])
6056		fl6.flowi6_uid = make_kuid(current_user_ns(),
6057					   nla_get_u32(tb[RTA_UID]));
6058	else
6059		fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6060
6061	if (tb[RTA_SPORT])
6062		fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6063
6064	if (tb[RTA_DPORT])
6065		fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6066
6067	if (tb[RTA_IP_PROTO]) {
6068		err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6069						  &fl6.flowi6_proto, AF_INET6,
6070						  extack);
6071		if (err)
6072			goto errout;
6073	}
6074
6075	if (iif) {
6076		struct net_device *dev;
6077		int flags = 0;
6078
6079		rcu_read_lock();
6080
6081		dev = dev_get_by_index_rcu(net, iif);
6082		if (!dev) {
6083			rcu_read_unlock();
6084			err = -ENODEV;
6085			goto errout;
6086		}
6087
6088		fl6.flowi6_iif = iif;
6089
6090		if (!ipv6_addr_any(&fl6.saddr))
6091			flags |= RT6_LOOKUP_F_HAS_SADDR;
6092
6093		dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6094
6095		rcu_read_unlock();
6096	} else {
6097		fl6.flowi6_oif = oif;
6098
6099		dst = ip6_route_output(net, NULL, &fl6);
6100	}
6101
6102
6103	rt = container_of(dst, struct rt6_info, dst);
6104	if (rt->dst.error) {
6105		err = rt->dst.error;
6106		ip6_rt_put(rt);
6107		goto errout;
6108	}
6109
6110	if (rt == net->ipv6.ip6_null_entry) {
6111		err = rt->dst.error;
6112		ip6_rt_put(rt);
6113		goto errout;
6114	}
6115
6116	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6117	if (!skb) {
6118		ip6_rt_put(rt);
6119		err = -ENOBUFS;
6120		goto errout;
6121	}
6122
6123	skb_dst_set(skb, &rt->dst);
6124
6125	rcu_read_lock();
6126	from = rcu_dereference(rt->from);
6127	if (from) {
6128		if (fibmatch)
6129			err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6130					    iif, RTM_NEWROUTE,
6131					    NETLINK_CB(in_skb).portid,
6132					    nlh->nlmsg_seq, 0);
6133		else
6134			err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6135					    &fl6.saddr, iif, RTM_NEWROUTE,
6136					    NETLINK_CB(in_skb).portid,
6137					    nlh->nlmsg_seq, 0);
6138	} else {
6139		err = -ENETUNREACH;
6140	}
6141	rcu_read_unlock();
6142
6143	if (err < 0) {
6144		kfree_skb(skb);
6145		goto errout;
6146	}
6147
6148	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6149errout:
6150	return err;
6151}
6152
6153void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6154		     unsigned int nlm_flags)
6155{
6156	struct sk_buff *skb;
6157	struct net *net = info->nl_net;
6158	u32 seq;
6159	int err;
6160
6161	err = -ENOBUFS;
6162	seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6163
6164	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6165	if (!skb)
6166		goto errout;
6167
6168	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6169			    event, info->portid, seq, nlm_flags);
6170	if (err < 0) {
6171		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6172		WARN_ON(err == -EMSGSIZE);
6173		kfree_skb(skb);
6174		goto errout;
6175	}
6176	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6177		    info->nlh, gfp_any());
6178	return;
6179errout:
6180	if (err < 0)
6181		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6182}
6183
6184void fib6_rt_update(struct net *net, struct fib6_info *rt,
6185		    struct nl_info *info)
6186{
6187	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6188	struct sk_buff *skb;
6189	int err = -ENOBUFS;
6190
6191	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6192	if (!skb)
6193		goto errout;
6194
6195	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6196			    RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6197	if (err < 0) {
6198		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6199		WARN_ON(err == -EMSGSIZE);
6200		kfree_skb(skb);
6201		goto errout;
6202	}
6203	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6204		    info->nlh, gfp_any());
6205	return;
6206errout:
6207	if (err < 0)
6208		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6209}
6210
6211void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6212			    bool offload, bool trap, bool offload_failed)
6213{
6214	struct sk_buff *skb;
6215	int err;
6216
6217	if (READ_ONCE(f6i->offload) == offload &&
6218	    READ_ONCE(f6i->trap) == trap &&
6219	    READ_ONCE(f6i->offload_failed) == offload_failed)
6220		return;
6221
6222	WRITE_ONCE(f6i->offload, offload);
6223	WRITE_ONCE(f6i->trap, trap);
6224
6225	/* 2 means send notifications only if offload_failed was changed. */
6226	if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6227	    READ_ONCE(f6i->offload_failed) == offload_failed)
6228		return;
6229
6230	WRITE_ONCE(f6i->offload_failed, offload_failed);
6231
6232	if (!rcu_access_pointer(f6i->fib6_node))
6233		/* The route was removed from the tree, do not send
6234		 * notification.
6235		 */
6236		return;
6237
6238	if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6239		return;
6240
6241	skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6242	if (!skb) {
6243		err = -ENOBUFS;
6244		goto errout;
6245	}
6246
6247	err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6248			    0, 0);
6249	if (err < 0) {
6250		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6251		WARN_ON(err == -EMSGSIZE);
6252		kfree_skb(skb);
6253		goto errout;
6254	}
6255
6256	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6257	return;
6258
6259errout:
6260	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6261}
6262EXPORT_SYMBOL(fib6_info_hw_flags_set);
6263
6264static int ip6_route_dev_notify(struct notifier_block *this,
6265				unsigned long event, void *ptr)
6266{
6267	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6268	struct net *net = dev_net(dev);
6269
6270	if (!(dev->flags & IFF_LOOPBACK))
6271		return NOTIFY_OK;
6272
6273	if (event == NETDEV_REGISTER) {
6274		net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6275		net->ipv6.ip6_null_entry->dst.dev = dev;
6276		net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6277#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6278		net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6279		net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6280		net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6281		net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6282#endif
6283	 } else if (event == NETDEV_UNREGISTER &&
6284		    dev->reg_state != NETREG_UNREGISTERED) {
6285		/* NETDEV_UNREGISTER could be fired for multiple times by
6286		 * netdev_wait_allrefs(). Make sure we only call this once.
6287		 */
6288		in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6289#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6290		in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6291		in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6292#endif
6293	}
6294
6295	return NOTIFY_OK;
6296}
6297
6298/*
6299 *	/proc
6300 */
6301
6302#ifdef CONFIG_PROC_FS
6303static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6304{
6305	struct net *net = (struct net *)seq->private;
6306	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6307		   net->ipv6.rt6_stats->fib_nodes,
6308		   net->ipv6.rt6_stats->fib_route_nodes,
6309		   atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6310		   net->ipv6.rt6_stats->fib_rt_entries,
6311		   net->ipv6.rt6_stats->fib_rt_cache,
6312		   dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6313		   net->ipv6.rt6_stats->fib_discarded_routes);
6314
6315	return 0;
6316}
6317#endif	/* CONFIG_PROC_FS */
6318
6319#ifdef CONFIG_SYSCTL
6320
6321static int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6322			      void *buffer, size_t *lenp, loff_t *ppos)
6323{
6324	struct net *net;
6325	int delay;
6326	int ret;
6327	if (!write)
6328		return -EINVAL;
6329
6330	net = (struct net *)ctl->extra1;
6331	delay = net->ipv6.sysctl.flush_delay;
6332	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6333	if (ret)
6334		return ret;
6335
6336	fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6337	return 0;
6338}
6339
6340static struct ctl_table ipv6_route_table_template[] = {
6341	{
6342		.procname	=	"max_size",
6343		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
6344		.maxlen		=	sizeof(int),
6345		.mode		=	0644,
6346		.proc_handler	=	proc_dointvec,
6347	},
6348	{
6349		.procname	=	"gc_thresh",
6350		.data		=	&ip6_dst_ops_template.gc_thresh,
6351		.maxlen		=	sizeof(int),
6352		.mode		=	0644,
6353		.proc_handler	=	proc_dointvec,
6354	},
6355	{
6356		.procname	=	"flush",
6357		.data		=	&init_net.ipv6.sysctl.flush_delay,
6358		.maxlen		=	sizeof(int),
6359		.mode		=	0200,
6360		.proc_handler	=	ipv6_sysctl_rtcache_flush
6361	},
6362	{
6363		.procname	=	"gc_min_interval",
6364		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6365		.maxlen		=	sizeof(int),
6366		.mode		=	0644,
6367		.proc_handler	=	proc_dointvec_jiffies,
6368	},
6369	{
6370		.procname	=	"gc_timeout",
6371		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6372		.maxlen		=	sizeof(int),
6373		.mode		=	0644,
6374		.proc_handler	=	proc_dointvec_jiffies,
6375	},
6376	{
6377		.procname	=	"gc_interval",
6378		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
6379		.maxlen		=	sizeof(int),
6380		.mode		=	0644,
6381		.proc_handler	=	proc_dointvec_jiffies,
6382	},
6383	{
6384		.procname	=	"gc_elasticity",
6385		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6386		.maxlen		=	sizeof(int),
6387		.mode		=	0644,
6388		.proc_handler	=	proc_dointvec,
6389	},
6390	{
6391		.procname	=	"mtu_expires",
6392		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6393		.maxlen		=	sizeof(int),
6394		.mode		=	0644,
6395		.proc_handler	=	proc_dointvec_jiffies,
6396	},
6397	{
6398		.procname	=	"min_adv_mss",
6399		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
6400		.maxlen		=	sizeof(int),
6401		.mode		=	0644,
6402		.proc_handler	=	proc_dointvec,
6403	},
6404	{
6405		.procname	=	"gc_min_interval_ms",
6406		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6407		.maxlen		=	sizeof(int),
6408		.mode		=	0644,
6409		.proc_handler	=	proc_dointvec_ms_jiffies,
6410	},
6411	{
6412		.procname	=	"skip_notify_on_dev_down",
6413		.data		=	&init_net.ipv6.sysctl.skip_notify_on_dev_down,
6414		.maxlen		=	sizeof(u8),
6415		.mode		=	0644,
6416		.proc_handler	=	proc_dou8vec_minmax,
6417		.extra1		=	SYSCTL_ZERO,
6418		.extra2		=	SYSCTL_ONE,
6419	},
6420	{ }
6421};
6422
6423struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6424{
6425	struct ctl_table *table;
6426
6427	table = kmemdup(ipv6_route_table_template,
6428			sizeof(ipv6_route_table_template),
6429			GFP_KERNEL);
6430
6431	if (table) {
6432		table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
6433		table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6434		table[2].data = &net->ipv6.sysctl.flush_delay;
6435		table[2].extra1 = net;
6436		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6437		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6438		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6439		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6440		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6441		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6442		table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6443		table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6444
6445		/* Don't export sysctls to unprivileged users */
6446		if (net->user_ns != &init_user_ns)
6447			table[1].procname = NULL;
6448	}
6449
6450	return table;
6451}
6452
6453size_t ipv6_route_sysctl_table_size(struct net *net)
6454{
6455	/* Don't export sysctls to unprivileged users */
6456	if (net->user_ns != &init_user_ns)
6457		return 1;
6458
6459	return ARRAY_SIZE(ipv6_route_table_template);
6460}
6461#endif
6462
6463static int __net_init ip6_route_net_init(struct net *net)
6464{
6465	int ret = -ENOMEM;
6466
6467	memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6468	       sizeof(net->ipv6.ip6_dst_ops));
6469
6470	if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6471		goto out_ip6_dst_ops;
6472
6473	net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6474	if (!net->ipv6.fib6_null_entry)
6475		goto out_ip6_dst_entries;
6476	memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6477	       sizeof(*net->ipv6.fib6_null_entry));
6478
6479	net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6480					   sizeof(*net->ipv6.ip6_null_entry),
6481					   GFP_KERNEL);
6482	if (!net->ipv6.ip6_null_entry)
6483		goto out_fib6_null_entry;
6484	net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6485	dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6486			 ip6_template_metrics, true);
6487	INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6488
6489#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6490	net->ipv6.fib6_has_custom_rules = false;
6491	net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6492					       sizeof(*net->ipv6.ip6_prohibit_entry),
6493					       GFP_KERNEL);
6494	if (!net->ipv6.ip6_prohibit_entry)
6495		goto out_ip6_null_entry;
6496	net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6497	dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6498			 ip6_template_metrics, true);
6499	INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6500
6501	net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6502					       sizeof(*net->ipv6.ip6_blk_hole_entry),
6503					       GFP_KERNEL);
6504	if (!net->ipv6.ip6_blk_hole_entry)
6505		goto out_ip6_prohibit_entry;
6506	net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6507	dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6508			 ip6_template_metrics, true);
6509	INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6510#ifdef CONFIG_IPV6_SUBTREES
6511	net->ipv6.fib6_routes_require_src = 0;
6512#endif
6513#endif
6514
6515	net->ipv6.sysctl.flush_delay = 0;
6516	net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6517	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6518	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6519	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6520	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6521	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6522	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6523	net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6524
6525	atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6526
6527	ret = 0;
6528out:
6529	return ret;
6530
6531#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6532out_ip6_prohibit_entry:
6533	kfree(net->ipv6.ip6_prohibit_entry);
6534out_ip6_null_entry:
6535	kfree(net->ipv6.ip6_null_entry);
6536#endif
6537out_fib6_null_entry:
6538	kfree(net->ipv6.fib6_null_entry);
6539out_ip6_dst_entries:
6540	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6541out_ip6_dst_ops:
6542	goto out;
6543}
6544
6545static void __net_exit ip6_route_net_exit(struct net *net)
6546{
6547	kfree(net->ipv6.fib6_null_entry);
6548	kfree(net->ipv6.ip6_null_entry);
6549#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6550	kfree(net->ipv6.ip6_prohibit_entry);
6551	kfree(net->ipv6.ip6_blk_hole_entry);
6552#endif
6553	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6554}
6555
6556static int __net_init ip6_route_net_init_late(struct net *net)
6557{
6558#ifdef CONFIG_PROC_FS
6559	if (!proc_create_net("ipv6_route", 0, net->proc_net,
6560			     &ipv6_route_seq_ops,
6561			     sizeof(struct ipv6_route_iter)))
6562		return -ENOMEM;
6563
6564	if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6565				    rt6_stats_seq_show, NULL)) {
6566		remove_proc_entry("ipv6_route", net->proc_net);
6567		return -ENOMEM;
6568	}
6569#endif
6570	return 0;
6571}
6572
6573static void __net_exit ip6_route_net_exit_late(struct net *net)
6574{
6575#ifdef CONFIG_PROC_FS
6576	remove_proc_entry("ipv6_route", net->proc_net);
6577	remove_proc_entry("rt6_stats", net->proc_net);
6578#endif
6579}
6580
6581static struct pernet_operations ip6_route_net_ops = {
6582	.init = ip6_route_net_init,
6583	.exit = ip6_route_net_exit,
6584};
6585
6586static int __net_init ipv6_inetpeer_init(struct net *net)
6587{
6588	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6589
6590	if (!bp)
6591		return -ENOMEM;
6592	inet_peer_base_init(bp);
6593	net->ipv6.peers = bp;
6594	return 0;
6595}
6596
6597static void __net_exit ipv6_inetpeer_exit(struct net *net)
6598{
6599	struct inet_peer_base *bp = net->ipv6.peers;
6600
6601	net->ipv6.peers = NULL;
6602	inetpeer_invalidate_tree(bp);
6603	kfree(bp);
6604}
6605
6606static struct pernet_operations ipv6_inetpeer_ops = {
6607	.init	=	ipv6_inetpeer_init,
6608	.exit	=	ipv6_inetpeer_exit,
6609};
6610
6611static struct pernet_operations ip6_route_net_late_ops = {
6612	.init = ip6_route_net_init_late,
6613	.exit = ip6_route_net_exit_late,
6614};
6615
6616static struct notifier_block ip6_route_dev_notifier = {
6617	.notifier_call = ip6_route_dev_notify,
6618	.priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6619};
6620
6621void __init ip6_route_init_special_entries(void)
6622{
6623	/* Registering of the loopback is done before this portion of code,
6624	 * the loopback reference in rt6_info will not be taken, do it
6625	 * manually for init_net */
6626	init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6627	init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6628	init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6629  #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6630	init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6631	init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6632	init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6633	init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6634  #endif
6635}
6636
6637#if IS_BUILTIN(CONFIG_IPV6)
6638#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6639DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6640
6641BTF_ID_LIST(btf_fib6_info_id)
6642BTF_ID(struct, fib6_info)
6643
6644static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6645	.seq_ops		= &ipv6_route_seq_ops,
6646	.init_seq_private	= bpf_iter_init_seq_net,
6647	.fini_seq_private	= bpf_iter_fini_seq_net,
6648	.seq_priv_size		= sizeof(struct ipv6_route_iter),
6649};
6650
6651static struct bpf_iter_reg ipv6_route_reg_info = {
6652	.target			= "ipv6_route",
6653	.ctx_arg_info_size	= 1,
6654	.ctx_arg_info		= {
6655		{ offsetof(struct bpf_iter__ipv6_route, rt),
6656		  PTR_TO_BTF_ID_OR_NULL },
6657	},
6658	.seq_info		= &ipv6_route_seq_info,
6659};
6660
6661static int __init bpf_iter_register(void)
6662{
6663	ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6664	return bpf_iter_reg_target(&ipv6_route_reg_info);
6665}
6666
6667static void bpf_iter_unregister(void)
6668{
6669	bpf_iter_unreg_target(&ipv6_route_reg_info);
6670}
6671#endif
6672#endif
6673
6674int __init ip6_route_init(void)
6675{
6676	int ret;
6677	int cpu;
6678
6679	ret = -ENOMEM;
6680	ip6_dst_ops_template.kmem_cachep =
6681		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6682				  SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6683	if (!ip6_dst_ops_template.kmem_cachep)
6684		goto out;
6685
6686	ret = dst_entries_init(&ip6_dst_blackhole_ops);
6687	if (ret)
6688		goto out_kmem_cache;
6689
6690	ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6691	if (ret)
6692		goto out_dst_entries;
6693
6694	ret = register_pernet_subsys(&ip6_route_net_ops);
6695	if (ret)
6696		goto out_register_inetpeer;
6697
6698	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6699
6700	ret = fib6_init();
6701	if (ret)
6702		goto out_register_subsys;
6703
6704	ret = xfrm6_init();
6705	if (ret)
6706		goto out_fib6_init;
6707
6708	ret = fib6_rules_init();
6709	if (ret)
6710		goto xfrm6_init;
6711
6712	ret = register_pernet_subsys(&ip6_route_net_late_ops);
6713	if (ret)
6714		goto fib6_rules_init;
6715
6716	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6717				   inet6_rtm_newroute, NULL, 0);
6718	if (ret < 0)
6719		goto out_register_late_subsys;
6720
6721	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6722				   inet6_rtm_delroute, NULL, 0);
6723	if (ret < 0)
6724		goto out_register_late_subsys;
6725
6726	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6727				   inet6_rtm_getroute, NULL,
6728				   RTNL_FLAG_DOIT_UNLOCKED);
6729	if (ret < 0)
6730		goto out_register_late_subsys;
6731
6732	ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6733	if (ret)
6734		goto out_register_late_subsys;
6735
6736#if IS_BUILTIN(CONFIG_IPV6)
6737#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6738	ret = bpf_iter_register();
6739	if (ret)
6740		goto out_register_late_subsys;
6741#endif
6742#endif
6743
6744	for_each_possible_cpu(cpu) {
6745		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6746
6747		INIT_LIST_HEAD(&ul->head);
6748		INIT_LIST_HEAD(&ul->quarantine);
6749		spin_lock_init(&ul->lock);
6750	}
6751
6752out:
6753	return ret;
6754
6755out_register_late_subsys:
6756	rtnl_unregister_all(PF_INET6);
6757	unregister_pernet_subsys(&ip6_route_net_late_ops);
6758fib6_rules_init:
6759	fib6_rules_cleanup();
6760xfrm6_init:
6761	xfrm6_fini();
6762out_fib6_init:
6763	fib6_gc_cleanup();
6764out_register_subsys:
6765	unregister_pernet_subsys(&ip6_route_net_ops);
6766out_register_inetpeer:
6767	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6768out_dst_entries:
6769	dst_entries_destroy(&ip6_dst_blackhole_ops);
6770out_kmem_cache:
6771	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6772	goto out;
6773}
6774
6775void ip6_route_cleanup(void)
6776{
6777#if IS_BUILTIN(CONFIG_IPV6)
6778#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6779	bpf_iter_unregister();
6780#endif
6781#endif
6782	unregister_netdevice_notifier(&ip6_route_dev_notifier);
6783	unregister_pernet_subsys(&ip6_route_net_late_ops);
6784	fib6_rules_cleanup();
6785	xfrm6_fini();
6786	fib6_gc_cleanup();
6787	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6788	unregister_pernet_subsys(&ip6_route_net_ops);
6789	dst_entries_destroy(&ip6_dst_blackhole_ops);
6790	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6791}