Linux Audio

Check our new training course

Loading...
v3.15
 
   1/*
   2 *	Linux INET6 implementation
   3 *	FIB front-end.
   4 *
   5 *	Authors:
   6 *	Pedro Roque		<roque@di.fc.ul.pt>
   7 *
   8 *	This program is free software; you can redistribute it and/or
   9 *      modify it under the terms of the GNU General Public License
  10 *      as published by the Free Software Foundation; either version
  11 *      2 of the License, or (at your option) any later version.
  12 */
  13
  14/*	Changes:
  15 *
  16 *	YOSHIFUJI Hideaki @USAGI
  17 *		reworked default router selection.
  18 *		- respect outgoing interface
  19 *		- select from (probably) reachable routers (i.e.
  20 *		routers in REACHABLE, STALE, DELAY or PROBE states).
  21 *		- always select the same router if it is (probably)
  22 *		reachable.  otherwise, round-robin the list.
  23 *	Ville Nuorvala
  24 *		Fixed routing subtrees.
  25 */
  26
  27#define pr_fmt(fmt) "IPv6: " fmt
  28
  29#include <linux/capability.h>
  30#include <linux/errno.h>
  31#include <linux/export.h>
  32#include <linux/types.h>
  33#include <linux/times.h>
  34#include <linux/socket.h>
  35#include <linux/sockios.h>
  36#include <linux/net.h>
  37#include <linux/route.h>
  38#include <linux/netdevice.h>
  39#include <linux/in6.h>
  40#include <linux/mroute6.h>
  41#include <linux/init.h>
  42#include <linux/if_arp.h>
  43#include <linux/proc_fs.h>
  44#include <linux/seq_file.h>
  45#include <linux/nsproxy.h>
  46#include <linux/slab.h>
 
 
  47#include <net/net_namespace.h>
  48#include <net/snmp.h>
  49#include <net/ipv6.h>
  50#include <net/ip6_fib.h>
  51#include <net/ip6_route.h>
  52#include <net/ndisc.h>
  53#include <net/addrconf.h>
  54#include <net/tcp.h>
  55#include <linux/rtnetlink.h>
  56#include <net/dst.h>
 
  57#include <net/xfrm.h>
  58#include <net/netevent.h>
  59#include <net/netlink.h>
  60#include <net/nexthop.h>
  61
  62#include <asm/uaccess.h>
 
 
 
 
  63
  64#ifdef CONFIG_SYSCTL
  65#include <linux/sysctl.h>
  66#endif
  67
 
 
 
 
 
 
 
  68enum rt6_nud_state {
  69	RT6_NUD_FAIL_HARD = -3,
  70	RT6_NUD_FAIL_PROBE = -2,
  71	RT6_NUD_FAIL_DO_RR = -1,
  72	RT6_NUD_SUCCEED = 1
  73};
  74
  75static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
  76				    const struct in6_addr *dest);
  77static struct dst_entry	*ip6_dst_check(struct dst_entry *dst, u32 cookie);
  78static unsigned int	 ip6_default_advmss(const struct dst_entry *dst);
  79static unsigned int	 ip6_mtu(const struct dst_entry *dst);
  80static struct dst_entry *ip6_negative_advice(struct dst_entry *);
 
 
  81static void		ip6_dst_destroy(struct dst_entry *);
  82static void		ip6_dst_ifdown(struct dst_entry *,
  83				       struct net_device *dev, int how);
  84static int		 ip6_dst_gc(struct dst_ops *ops);
  85
  86static int		ip6_pkt_discard(struct sk_buff *skb);
  87static int		ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
  88static int		ip6_pkt_prohibit(struct sk_buff *skb);
  89static int		ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
  90static void		ip6_link_failure(struct sk_buff *skb);
  91static void		ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  92					   struct sk_buff *skb, u32 mtu);
 
  93static void		rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  94					struct sk_buff *skb);
  95static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
 
 
 
 
 
 
 
 
 
 
  96
  97#ifdef CONFIG_IPV6_ROUTE_INFO
  98static struct rt6_info *rt6_add_route_info(struct net *net,
  99					   const struct in6_addr *prefix, int prefixlen,
 100					   const struct in6_addr *gwaddr, int ifindex,
 
 101					   unsigned int pref);
 102static struct rt6_info *rt6_get_route_info(struct net *net,
 103					   const struct in6_addr *prefix, int prefixlen,
 104					   const struct in6_addr *gwaddr, int ifindex);
 
 105#endif
 106
 107static void rt6_bind_peer(struct rt6_info *rt, int create)
 108{
 109	struct inet_peer_base *base;
 110	struct inet_peer *peer;
 111
 112	base = inetpeer_base_ptr(rt->_rt6i_peer);
 113	if (!base)
 114		return;
 115
 116	peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
 117	if (peer) {
 118		if (!rt6_set_peer(rt, peer))
 119			inet_putpeer(peer);
 120	}
 121}
 122
 123static struct inet_peer *__rt6_get_peer(struct rt6_info *rt, int create)
 124{
 125	if (rt6_has_peer(rt))
 126		return rt6_peer_ptr(rt);
 127
 128	rt6_bind_peer(rt, create);
 129	return (rt6_has_peer(rt) ? rt6_peer_ptr(rt) : NULL);
 
 
 
 130}
 131
 132static struct inet_peer *rt6_get_peer_create(struct rt6_info *rt)
 133{
 134	return __rt6_get_peer(rt, 1);
 
 
 
 
 
 
 135}
 136
 137static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 138{
 139	struct rt6_info *rt = (struct rt6_info *) dst;
 140	struct inet_peer *peer;
 141	u32 *p = NULL;
 142
 143	if (!(rt->dst.flags & DST_HOST))
 144		return NULL;
 145
 146	peer = rt6_get_peer_create(rt);
 147	if (peer) {
 148		u32 *old_p = __DST_METRICS_PTR(old);
 149		unsigned long prev, new;
 150
 151		p = peer->metrics;
 152		if (inet_metrics_new(peer) ||
 153		    (old & DST_METRICS_FORCE_OVERWRITE))
 154			memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
 155
 156		new = (unsigned long) p;
 157		prev = cmpxchg(&dst->_metrics, old, new);
 
 
 
 
 
 
 
 
 
 158
 159		if (prev != old) {
 160			p = __DST_METRICS_PTR(prev);
 161			if (prev & DST_METRICS_READ_ONLY)
 162				p = NULL;
 
 
 
 
 
 163		}
 
 164	}
 165	return p;
 166}
 167
 168static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 169					     struct sk_buff *skb,
 170					     const void *daddr)
 171{
 172	struct in6_addr *p = &rt->rt6i_gateway;
 173
 174	if (!ipv6_addr_any(p))
 175		return (const void *) p;
 176	else if (skb)
 177		return &ipv6_hdr(skb)->daddr;
 178	return daddr;
 179}
 180
 181static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 182					  struct sk_buff *skb,
 183					  const void *daddr)
 
 184{
 185	struct rt6_info *rt = (struct rt6_info *) dst;
 186	struct neighbour *n;
 187
 188	daddr = choose_neigh_daddr(rt, skb, daddr);
 189	n = __ipv6_neigh_lookup(dst->dev, daddr);
 190	if (n)
 191		return n;
 192	return neigh_create(&nd_tbl, daddr, dst->dev);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 193}
 194
 195static struct dst_ops ip6_dst_ops_template = {
 196	.family			=	AF_INET6,
 197	.protocol		=	cpu_to_be16(ETH_P_IPV6),
 198	.gc			=	ip6_dst_gc,
 199	.gc_thresh		=	1024,
 200	.check			=	ip6_dst_check,
 201	.default_advmss		=	ip6_default_advmss,
 202	.mtu			=	ip6_mtu,
 203	.cow_metrics		=	ipv6_cow_metrics,
 204	.destroy		=	ip6_dst_destroy,
 205	.ifdown			=	ip6_dst_ifdown,
 206	.negative_advice	=	ip6_negative_advice,
 207	.link_failure		=	ip6_link_failure,
 208	.update_pmtu		=	ip6_rt_update_pmtu,
 209	.redirect		=	rt6_do_redirect,
 210	.local_out		=	__ip6_local_out,
 211	.neigh_lookup		=	ip6_neigh_lookup,
 
 212};
 213
 214static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 215{
 216	unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 217
 218	return mtu ? : dst->dev->mtu;
 219}
 220
 221static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 222					 struct sk_buff *skb, u32 mtu)
 223{
 224}
 225
 226static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 227				      struct sk_buff *skb)
 228{
 229}
 230
 231static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
 232					 unsigned long old)
 233{
 234	return NULL;
 235}
 236
 237static struct dst_ops ip6_dst_blackhole_ops = {
 238	.family			=	AF_INET6,
 239	.protocol		=	cpu_to_be16(ETH_P_IPV6),
 240	.destroy		=	ip6_dst_destroy,
 241	.check			=	ip6_dst_check,
 242	.mtu			=	ip6_blackhole_mtu,
 243	.default_advmss		=	ip6_default_advmss,
 244	.update_pmtu		=	ip6_rt_blackhole_update_pmtu,
 245	.redirect		=	ip6_rt_blackhole_redirect,
 246	.cow_metrics		=	ip6_rt_blackhole_cow_metrics,
 247	.neigh_lookup		=	ip6_neigh_lookup,
 248};
 249
 250static const u32 ip6_template_metrics[RTAX_MAX] = {
 251	[RTAX_HOPLIMIT - 1] = 0,
 252};
 253
 
 
 
 
 
 
 
 
 
 254static const struct rt6_info ip6_null_entry_template = {
 255	.dst = {
 256		.__refcnt	= ATOMIC_INIT(1),
 257		.__use		= 1,
 258		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 259		.error		= -ENETUNREACH,
 260		.input		= ip6_pkt_discard,
 261		.output		= ip6_pkt_discard_out,
 262	},
 263	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 264	.rt6i_protocol  = RTPROT_KERNEL,
 265	.rt6i_metric	= ~(u32) 0,
 266	.rt6i_ref	= ATOMIC_INIT(1),
 267};
 268
 269#ifdef CONFIG_IPV6_MULTIPLE_TABLES
 270
 271static const struct rt6_info ip6_prohibit_entry_template = {
 272	.dst = {
 273		.__refcnt	= ATOMIC_INIT(1),
 274		.__use		= 1,
 275		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 276		.error		= -EACCES,
 277		.input		= ip6_pkt_prohibit,
 278		.output		= ip6_pkt_prohibit_out,
 279	},
 280	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 281	.rt6i_protocol  = RTPROT_KERNEL,
 282	.rt6i_metric	= ~(u32) 0,
 283	.rt6i_ref	= ATOMIC_INIT(1),
 284};
 285
 286static const struct rt6_info ip6_blk_hole_entry_template = {
 287	.dst = {
 288		.__refcnt	= ATOMIC_INIT(1),
 289		.__use		= 1,
 290		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 291		.error		= -EINVAL,
 292		.input		= dst_discard,
 293		.output		= dst_discard_sk,
 294	},
 295	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 296	.rt6i_protocol  = RTPROT_KERNEL,
 297	.rt6i_metric	= ~(u32) 0,
 298	.rt6i_ref	= ATOMIC_INIT(1),
 299};
 300
 301#endif
 302
 
 
 
 
 
 303/* allocate dst with ip6_dst_ops */
 304static inline struct rt6_info *ip6_dst_alloc(struct net *net,
 305					     struct net_device *dev,
 306					     int flags,
 307					     struct fib6_table *table)
 308{
 309	struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 310					0, DST_OBSOLETE_FORCE_CHK, flags);
 311
 312	if (rt) {
 313		struct dst_entry *dst = &rt->dst;
 314
 315		memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 316		rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
 317		rt->rt6i_genid = rt_genid_ipv6(net);
 318		INIT_LIST_HEAD(&rt->rt6i_siblings);
 319	}
 
 320	return rt;
 321}
 
 322
 323static void ip6_dst_destroy(struct dst_entry *dst)
 324{
 325	struct rt6_info *rt = (struct rt6_info *)dst;
 326	struct inet6_dev *idev = rt->rt6i_idev;
 327	struct dst_entry *from = dst->from;
 328
 329	if (!(rt->dst.flags & DST_HOST))
 330		dst_destroy_metrics_generic(dst);
 331
 
 332	if (idev) {
 333		rt->rt6i_idev = NULL;
 334		in6_dev_put(idev);
 335	}
 336
 337	dst->from = NULL;
 338	dst_release(from);
 339
 340	if (rt6_has_peer(rt)) {
 341		struct inet_peer *peer = rt6_peer_ptr(rt);
 342		inet_putpeer(peer);
 343	}
 344}
 345
 346static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 347			   int how)
 348{
 349	struct rt6_info *rt = (struct rt6_info *)dst;
 350	struct inet6_dev *idev = rt->rt6i_idev;
 351	struct net_device *loopback_dev =
 352		dev_net(dev)->loopback_dev;
 353
 354	if (dev != loopback_dev) {
 355		if (idev && idev->dev == dev) {
 356			struct inet6_dev *loopback_idev =
 357				in6_dev_get(loopback_dev);
 358			if (loopback_idev) {
 359				rt->rt6i_idev = loopback_idev;
 360				in6_dev_put(idev);
 361			}
 362		}
 363	}
 
 
 
 
 
 
 
 
 
 
 364}
 365
 366static bool rt6_check_expired(const struct rt6_info *rt)
 367{
 
 
 
 
 368	if (rt->rt6i_flags & RTF_EXPIRES) {
 369		if (time_after(jiffies, rt->dst.expires))
 370			return true;
 371	} else if (rt->dst.from) {
 372		return rt6_check_expired((struct rt6_info *) rt->dst.from);
 
 373	}
 374	return false;
 375}
 376
 377/* Multipath route selection:
 378 *   Hash based function using packet header and flowlabel.
 379 * Adapted from fib_info_hashfn()
 380 */
 381static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 382			       const struct flowi6 *fl6)
 383{
 384	unsigned int val = fl6->flowi6_proto;
 
 
 
 
 
 
 
 385
 386	val ^= ipv6_addr_hash(&fl6->daddr);
 387	val ^= ipv6_addr_hash(&fl6->saddr);
 388
 389	/* Work only if this not encapsulated */
 390	switch (fl6->flowi6_proto) {
 391	case IPPROTO_UDP:
 392	case IPPROTO_TCP:
 393	case IPPROTO_SCTP:
 394		val ^= (__force u16)fl6->fl6_sport;
 395		val ^= (__force u16)fl6->fl6_dport;
 396		break;
 
 
 
 397
 398	case IPPROTO_ICMPV6:
 399		val ^= (__force u16)fl6->fl6_icmp_type;
 400		val ^= (__force u16)fl6->fl6_icmp_code;
 
 
 
 
 
 
 
 
 
 
 
 401		break;
 402	}
 403	/* RFC6438 recommands to use flowlabel */
 404	val ^= (__force u32)fl6->flowlabel;
 405
 406	/* Perhaps, we need to tune, this function? */
 407	val = val ^ (val >> 7) ^ (val >> 12);
 408	return val % candidate_count;
 409}
 410
 411static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 412					     struct flowi6 *fl6, int oif,
 413					     int strict)
 
 
 
 414{
 415	struct rt6_info *sibling, *next_sibling;
 416	int route_choosen;
 417
 418	route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 419	/* Don't change the route, if route_choosen == 0
 420	 * (siblings does not include ourself)
 421	 */
 422	if (route_choosen)
 423		list_for_each_entry_safe(sibling, next_sibling,
 424				&match->rt6i_siblings, rt6i_siblings) {
 425			route_choosen--;
 426			if (route_choosen == 0) {
 427				if (rt6_score_route(sibling, oif, strict) < 0)
 428					break;
 429				match = sibling;
 430				break;
 431			}
 432		}
 433	return match;
 434}
 435
 436/*
 437 *	Route lookup. Any table->tb6_lock is implied.
 438 */
 
 
 
 
 439
 440static inline struct rt6_info *rt6_device_match(struct net *net,
 441						    struct rt6_info *rt,
 442						    const struct in6_addr *saddr,
 443						    int oif,
 444						    int flags)
 445{
 446	struct rt6_info *local = NULL;
 447	struct rt6_info *sprt;
 448
 449	if (!oif && ipv6_addr_any(saddr))
 450		goto out;
 451
 452	for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 453		struct net_device *dev = sprt->dst.dev;
 454
 455		if (oif) {
 456			if (dev->ifindex == oif)
 457				return sprt;
 458			if (dev->flags & IFF_LOOPBACK) {
 459				if (!sprt->rt6i_idev ||
 460				    sprt->rt6i_idev->dev->ifindex != oif) {
 461					if (flags & RT6_LOOKUP_F_IFACE && oif)
 462						continue;
 463					if (local && (!oif ||
 464						      local->rt6i_idev->dev->ifindex == oif))
 465						continue;
 466				}
 467				local = sprt;
 468			}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 469		} else {
 470			if (ipv6_chk_addr(net, saddr, dev,
 471					  flags & RT6_LOOKUP_F_IFACE))
 472				return sprt;
 473		}
 
 
 474	}
 475
 476	if (oif) {
 477		if (local)
 478			return local;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 479
 480		if (flags & RT6_LOOKUP_F_IFACE)
 481			return net->ipv6.ip6_null_entry;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 482	}
 483out:
 484	return rt;
 
 
 
 
 
 
 
 
 485}
 486
 487#ifdef CONFIG_IPV6_ROUTER_PREF
 488struct __rt6_probe_work {
 489	struct work_struct work;
 490	struct in6_addr target;
 491	struct net_device *dev;
 
 492};
 493
 494static void rt6_probe_deferred(struct work_struct *w)
 495{
 496	struct in6_addr mcaddr;
 497	struct __rt6_probe_work *work =
 498		container_of(w, struct __rt6_probe_work, work);
 499
 500	addrconf_addr_solict_mult(&work->target, &mcaddr);
 501	ndisc_send_ns(work->dev, NULL, &work->target, &mcaddr, NULL);
 502	dev_put(work->dev);
 503	kfree(w);
 504}
 505
 506static void rt6_probe(struct rt6_info *rt)
 507{
 
 
 
 508	struct neighbour *neigh;
 
 
 
 509	/*
 510	 * Okay, this does not seem to be appropriate
 511	 * for now, however, we need to check if it
 512	 * is really so; aka Router Reachability Probing.
 513	 *
 514	 * Router Reachability Probe MUST be rate-limited
 515	 * to no more than one per minute.
 516	 */
 517	if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 518		return;
 519	rcu_read_lock_bh();
 520	neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 
 
 
 
 
 
 
 521	if (neigh) {
 522		write_lock(&neigh->lock);
 523		if (neigh->nud_state & NUD_VALID)
 524			goto out;
 525	}
 526
 527	if (!neigh ||
 528	    time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
 529		struct __rt6_probe_work *work;
 530
 
 
 
 
 
 
 
 
 
 
 
 
 531		work = kmalloc(sizeof(*work), GFP_ATOMIC);
 
 532
 533		if (neigh && work)
 534			__neigh_set_probe_once(neigh);
 535
 536		if (neigh)
 537			write_unlock(&neigh->lock);
 538
 539		if (work) {
 540			INIT_WORK(&work->work, rt6_probe_deferred);
 541			work->target = rt->rt6i_gateway;
 542			dev_hold(rt->dst.dev);
 543			work->dev = rt->dst.dev;
 544			schedule_work(&work->work);
 545		}
 546	} else {
 547out:
 548		write_unlock(&neigh->lock);
 
 
 
 549	}
 550	rcu_read_unlock_bh();
 
 
 551}
 552#else
 553static inline void rt6_probe(struct rt6_info *rt)
 554{
 555}
 556#endif
 557
 558/*
 559 * Default Router Selection (RFC 2461 6.3.6)
 560 */
 561static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 562{
 563	struct net_device *dev = rt->dst.dev;
 564	if (!oif || dev->ifindex == oif)
 565		return 2;
 566	if ((dev->flags & IFF_LOOPBACK) &&
 567	    rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 568		return 1;
 569	return 0;
 570}
 571
 572static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 573{
 574	struct neighbour *neigh;
 575	enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 
 576
 577	if (rt->rt6i_flags & RTF_NONEXTHOP ||
 578	    !(rt->rt6i_flags & RTF_GATEWAY))
 579		return RT6_NUD_SUCCEED;
 580
 581	rcu_read_lock_bh();
 582	neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 583	if (neigh) {
 584		read_lock(&neigh->lock);
 585		if (neigh->nud_state & NUD_VALID)
 
 586			ret = RT6_NUD_SUCCEED;
 587#ifdef CONFIG_IPV6_ROUTER_PREF
 588		else if (!(neigh->nud_state & NUD_FAILED))
 589			ret = RT6_NUD_SUCCEED;
 590		else
 591			ret = RT6_NUD_FAIL_PROBE;
 592#endif
 593		read_unlock(&neigh->lock);
 594	} else {
 595		ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 596		      RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 597	}
 598	rcu_read_unlock_bh();
 599
 600	return ret;
 601}
 602
 603static int rt6_score_route(struct rt6_info *rt, int oif,
 604			   int strict)
 605{
 606	int m;
 
 
 
 607
 608	m = rt6_check_dev(rt, oif);
 609	if (!m && (strict & RT6_LOOKUP_F_IFACE))
 610		return RT6_NUD_FAIL_HARD;
 611#ifdef CONFIG_IPV6_ROUTER_PREF
 612	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 613#endif
 614	if (strict & RT6_LOOKUP_F_REACHABLE) {
 615		int n = rt6_check_neigh(rt);
 
 616		if (n < 0)
 617			return n;
 618	}
 619	return m;
 620}
 621
 622static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 623				   int *mpri, struct rt6_info *match,
 624				   bool *do_rr)
 625{
 626	int m;
 627	bool match_do_rr = false;
 
 
 628
 629	if (rt6_check_expired(rt))
 630		goto out;
 631
 632	m = rt6_score_route(rt, oif, strict);
 
 
 
 
 
 633	if (m == RT6_NUD_FAIL_DO_RR) {
 634		match_do_rr = true;
 635		m = 0; /* lowest valid score */
 636	} else if (m == RT6_NUD_FAIL_HARD) {
 637		goto out;
 638	}
 639
 640	if (strict & RT6_LOOKUP_F_REACHABLE)
 641		rt6_probe(rt);
 642
 643	/* note that m can be RT6_NUD_FAIL_PROBE at this point */
 644	if (m > *mpri) {
 645		*do_rr = match_do_rr;
 646		*mpri = m;
 647		match = rt;
 648	}
 649out:
 650	return match;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 651}
 652
 653static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 654				     struct rt6_info *rr_head,
 655				     u32 metric, int oif, int strict,
 656				     bool *do_rr)
 657{
 658	struct rt6_info *rt, *match;
 
 659	int mpri = -1;
 660
 661	match = NULL;
 662	for (rt = rr_head; rt && rt->rt6i_metric == metric;
 663	     rt = rt->dst.rt6_next)
 664		match = find_match(rt, oif, strict, &mpri, match, do_rr);
 665	for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
 666	     rt = rt->dst.rt6_next)
 667		match = find_match(rt, oif, strict, &mpri, match, do_rr);
 668
 669	return match;
 
 
 
 
 
 
 
 670}
 671
 672static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 
 673{
 674	struct rt6_info *match, *rt0;
 675	struct net *net;
 676	bool do_rr = false;
 
 677
 678	rt0 = fn->rr_ptr;
 
 
 
 
 
 
 679	if (!rt0)
 680		fn->rr_ptr = rt0 = fn->leaf;
 681
 682	match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 683			     &do_rr);
 
 
 
 
 
 
 
 
 
 
 684
 
 685	if (do_rr) {
 686		struct rt6_info *next = rt0->dst.rt6_next;
 687
 688		/* no entries matched; do round-robin */
 689		if (!next || next->rt6i_metric != rt0->rt6i_metric)
 690			next = fn->leaf;
 
 
 
 
 
 
 
 
 
 691
 692		if (next != rt0)
 693			fn->rr_ptr = next;
 
 
 
 
 694	}
 
 695
 696	net = dev_net(rt0->dst.dev);
 697	return match ? match : net->ipv6.ip6_null_entry;
 
 
 698}
 699
 700#ifdef CONFIG_IPV6_ROUTE_INFO
 701int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 702		  const struct in6_addr *gwaddr)
 703{
 704	struct net *net = dev_net(dev);
 705	struct route_info *rinfo = (struct route_info *) opt;
 706	struct in6_addr prefix_buf, *prefix;
 
 707	unsigned int pref;
 708	unsigned long lifetime;
 709	struct rt6_info *rt;
 710
 711	if (len < sizeof(struct route_info)) {
 712		return -EINVAL;
 713	}
 714
 715	/* Sanity check for prefix_len and length */
 716	if (rinfo->length > 3) {
 717		return -EINVAL;
 718	} else if (rinfo->prefix_len > 128) {
 719		return -EINVAL;
 720	} else if (rinfo->prefix_len > 64) {
 721		if (rinfo->length < 2) {
 722			return -EINVAL;
 723		}
 724	} else if (rinfo->prefix_len > 0) {
 725		if (rinfo->length < 1) {
 726			return -EINVAL;
 727		}
 728	}
 729
 730	pref = rinfo->route_pref;
 731	if (pref == ICMPV6_ROUTER_PREF_INVALID)
 732		return -EINVAL;
 733
 734	lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 735
 736	if (rinfo->length == 3)
 737		prefix = (struct in6_addr *)rinfo->prefix;
 738	else {
 739		/* this function is safe */
 740		ipv6_addr_prefix(&prefix_buf,
 741				 (struct in6_addr *)rinfo->prefix,
 742				 rinfo->prefix_len);
 743		prefix = &prefix_buf;
 744	}
 745
 746	if (rinfo->prefix_len == 0)
 747		rt = rt6_get_dflt_router(gwaddr, dev);
 748	else
 749		rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 750					gwaddr, dev->ifindex);
 751
 752	if (rt && !lifetime) {
 753		ip6_del_rt(rt);
 754		rt = NULL;
 755	}
 756
 757	if (!rt && lifetime)
 758		rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 759					pref);
 760	else if (rt)
 761		rt->rt6i_flags = RTF_ROUTEINFO |
 762				 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 763
 764	if (rt) {
 765		if (!addrconf_finite_timeout(lifetime))
 766			rt6_clean_expires(rt);
 767		else
 768			rt6_set_expires(rt, jiffies + HZ * lifetime);
 769
 770		ip6_rt_put(rt);
 
 
 
 
 
 
 
 
 
 
 771	}
 772	return 0;
 773}
 774#endif
 775
 776#define BACKTRACK(__net, saddr)			\
 777do { \
 778	if (rt == __net->ipv6.ip6_null_entry) {	\
 779		struct fib6_node *pn; \
 780		while (1) { \
 781			if (fn->fn_flags & RTN_TL_ROOT) \
 782				goto out; \
 783			pn = fn->parent; \
 784			if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
 785				fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
 786			else \
 787				fn = pn; \
 788			if (fn->fn_flags & RTN_RTINFO) \
 789				goto restart; \
 790		} \
 791	} \
 792} while (0)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 793
 794static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 795					     struct fib6_table *table,
 796					     struct flowi6 *fl6, int flags)
 
 
 797{
 
 798	struct fib6_node *fn;
 799	struct rt6_info *rt;
 800
 801	read_lock_bh(&table->tb6_lock);
 802	fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 803restart:
 804	rt = fn->leaf;
 805	rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 806	if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 807		rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 808	BACKTRACK(net, &fl6->saddr);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 809out:
 810	dst_use(&rt->dst, jiffies);
 811	read_unlock_bh(&table->tb6_lock);
 812	return rt;
 813
 
 
 
 814}
 815
 816struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 817				    int flags)
 818{
 819	return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 820}
 821EXPORT_SYMBOL_GPL(ip6_route_lookup);
 822
 823struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 824			    const struct in6_addr *saddr, int oif, int strict)
 
 825{
 826	struct flowi6 fl6 = {
 827		.flowi6_oif = oif,
 828		.daddr = *daddr,
 829	};
 830	struct dst_entry *dst;
 831	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 832
 833	if (saddr) {
 834		memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 835		flags |= RT6_LOOKUP_F_HAS_SADDR;
 836	}
 837
 838	dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 839	if (dst->error == 0)
 840		return (struct rt6_info *) dst;
 841
 842	dst_release(dst);
 843
 844	return NULL;
 845}
 846
 847EXPORT_SYMBOL(rt6_lookup);
 848
 849/* ip6_ins_rt is called with FREE table->tb6_lock.
 850   It takes new route entry, the addition fails by any reason the
 851   route is freed. In any case, if caller does not hold it, it may
 852   be destroyed.
 853 */
 854
 855static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 856			struct nlattr *mx, int mx_len)
 857{
 858	int err;
 859	struct fib6_table *table;
 860
 861	table = rt->rt6i_table;
 862	write_lock_bh(&table->tb6_lock);
 863	err = fib6_add(&table->tb6_root, rt, info, mx, mx_len);
 864	write_unlock_bh(&table->tb6_lock);
 865
 866	return err;
 867}
 868
 869int ip6_ins_rt(struct rt6_info *rt)
 870{
 871	struct nl_info info = {
 872		.nl_net = dev_net(rt->dst.dev),
 873	};
 874	return __ip6_ins_rt(rt, &info, NULL, 0);
 875}
 876
 877static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
 878				      const struct in6_addr *daddr,
 879				      const struct in6_addr *saddr)
 880{
 
 
 881	struct rt6_info *rt;
 882
 883	/*
 884	 *	Clone the route.
 885	 */
 886
 887	rt = ip6_rt_copy(ort, daddr);
 
 888
 889	if (rt) {
 890		if (ort->rt6i_dst.plen != 128 &&
 891		    ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 892			rt->rt6i_flags |= RTF_ANYCAST;
 
 
 893
 894		rt->rt6i_flags |= RTF_CACHE;
 
 
 
 895
 
 
 
 
 896#ifdef CONFIG_IPV6_SUBTREES
 897		if (rt->rt6i_src.plen && saddr) {
 898			rt->rt6i_src.addr = *saddr;
 899			rt->rt6i_src.plen = 128;
 900		}
 901#endif
 902	}
 903
 904	return rt;
 905}
 906
 907static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
 908					const struct in6_addr *daddr)
 909{
 910	struct rt6_info *rt = ip6_rt_copy(ort, daddr);
 
 
 
 911
 912	if (rt)
 913		rt->rt6i_flags |= RTF_CACHE;
 914	return rt;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 915}
 916
 917static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
 918				      struct flowi6 *fl6, int flags)
 919{
 920	struct fib6_node *fn;
 921	struct rt6_info *rt, *nrt;
 922	int strict = 0;
 923	int attempts = 3;
 924	int err;
 925	int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 926
 927	strict |= flags & RT6_LOOKUP_F_IFACE;
 
 
 
 928
 929relookup:
 930	read_lock_bh(&table->tb6_lock);
 931
 932restart_2:
 933	fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 934
 935restart:
 936	rt = rt6_select(fn, oif, strict | reachable);
 937	if (rt->rt6i_nsiblings)
 938		rt = rt6_multipath_select(rt, fl6, oif, strict | reachable);
 939	BACKTRACK(net, &fl6->saddr);
 940	if (rt == net->ipv6.ip6_null_entry ||
 941	    rt->rt6i_flags & RTF_CACHE)
 942		goto out;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 943
 944	dst_hold(&rt->dst);
 945	read_unlock_bh(&table->tb6_lock);
 946
 947	if (!(rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY)))
 948		nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
 949	else if (!(rt->dst.flags & DST_HOST))
 950		nrt = rt6_alloc_clone(rt, &fl6->daddr);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 951	else
 952		goto out2;
 953
 954	ip6_rt_put(rt);
 955	rt = nrt ? : net->ipv6.ip6_null_entry;
 
 956
 957	dst_hold(&rt->dst);
 958	if (nrt) {
 959		err = ip6_ins_rt(nrt);
 960		if (!err)
 961			goto out2;
 962	}
 963
 964	if (--attempts <= 0)
 965		goto out2;
 966
 967	/*
 968	 * Race condition! In the gap, when table->tb6_lock was
 969	 * released someone could insert this route.  Relookup.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 970	 */
 971	ip6_rt_put(rt);
 972	goto relookup;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 973
 974out:
 975	if (reachable) {
 976		reachable = 0;
 977		goto restart_2;
 
 
 
 
 
 978	}
 979	dst_hold(&rt->dst);
 980	read_unlock_bh(&table->tb6_lock);
 981out2:
 982	rt->dst.lastuse = jiffies;
 983	rt->dst.__use++;
 984
 985	return rt;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 986}
 987
 988static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
 989					    struct flowi6 *fl6, int flags)
 990{
 991	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
 
 
 
 
 992}
 993
 994static struct dst_entry *ip6_route_input_lookup(struct net *net,
 995						struct net_device *dev,
 996						struct flowi6 *fl6, int flags)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 997{
 998	if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
 999		flags |= RT6_LOOKUP_F_IFACE;
1000
1001	return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1002}
1003
 
1004void ip6_route_input(struct sk_buff *skb)
1005{
1006	const struct ipv6hdr *iph = ipv6_hdr(skb);
1007	struct net *net = dev_net(skb->dev);
1008	int flags = RT6_LOOKUP_F_HAS_SADDR;
 
1009	struct flowi6 fl6 = {
1010		.flowi6_iif = skb->dev->ifindex,
1011		.daddr = iph->daddr,
1012		.saddr = iph->saddr,
1013		.flowlabel = ip6_flowinfo(iph),
1014		.flowi6_mark = skb->mark,
1015		.flowi6_proto = iph->nexthdr,
1016	};
 
 
 
 
 
1017
1018	skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
 
 
 
 
 
 
 
1019}
1020
1021static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1022					     struct flowi6 *fl6, int flags)
 
 
 
1023{
1024	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1025}
1026
1027struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
1028				    struct flowi6 *fl6)
 
 
1029{
1030	int flags = 0;
 
 
 
 
 
 
 
 
 
 
1031
1032	fl6->flowi6_iif = LOOPBACK_IFINDEX;
1033
1034	if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
 
 
 
1035		flags |= RT6_LOOKUP_F_IFACE;
1036
1037	if (!ipv6_addr_any(&fl6->saddr))
1038		flags |= RT6_LOOKUP_F_HAS_SADDR;
1039	else if (sk)
1040		flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1041
1042	return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1043}
1044
1045EXPORT_SYMBOL(ip6_route_output);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1046
1047struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1048{
1049	struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
 
1050	struct dst_entry *new = NULL;
1051
1052	rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
 
1053	if (rt) {
1054		new = &rt->dst;
1055
1056		memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1057		rt6_init_peer(rt, net->ipv6.peers);
1058
 
1059		new->__use = 1;
1060		new->input = dst_discard;
1061		new->output = dst_discard_sk;
1062
1063		if (dst_metrics_read_only(&ort->dst))
1064			new->_metrics = ort->dst._metrics;
1065		else
1066			dst_copy_metrics(new, &ort->dst);
1067		rt->rt6i_idev = ort->rt6i_idev;
1068		if (rt->rt6i_idev)
1069			in6_dev_hold(rt->rt6i_idev);
1070
 
1071		rt->rt6i_gateway = ort->rt6i_gateway;
1072		rt->rt6i_flags = ort->rt6i_flags;
1073		rt->rt6i_metric = 0;
1074
1075		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1076#ifdef CONFIG_IPV6_SUBTREES
1077		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1078#endif
1079
1080		dst_free(new);
1081	}
1082
1083	dst_release(dst_orig);
1084	return new ? new : ERR_PTR(-ENOMEM);
1085}
1086
1087/*
1088 *	Destination cache support functions
1089 */
1090
1091static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1092{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1093	struct rt6_info *rt;
1094
1095	rt = (struct rt6_info *) dst;
 
 
 
 
 
1096
1097	/* All IPV6 dsts are created with ->obsolete set to the value
1098	 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1099	 * into this function always.
1100	 */
1101	if (rt->rt6i_genid != rt_genid_ipv6(dev_net(rt->dst.dev)))
1102		return NULL;
1103
1104	if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1105		return NULL;
1106
1107	if (rt6_check_expired(rt))
1108		return NULL;
 
 
 
1109
1110	return dst;
 
 
1111}
 
1112
1113static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
 
1114{
1115	struct rt6_info *rt = (struct rt6_info *) dst;
1116
1117	if (rt) {
1118		if (rt->rt6i_flags & RTF_CACHE) {
1119			if (rt6_check_expired(rt)) {
1120				ip6_del_rt(rt);
1121				dst = NULL;
1122			}
1123		} else {
1124			dst_release(dst);
1125			dst = NULL;
1126		}
 
 
1127	}
1128	return dst;
1129}
1130
1131static void ip6_link_failure(struct sk_buff *skb)
1132{
1133	struct rt6_info *rt;
1134
1135	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1136
1137	rt = (struct rt6_info *) skb_dst(skb);
1138	if (rt) {
 
1139		if (rt->rt6i_flags & RTF_CACHE) {
1140			dst_hold(&rt->dst);
1141			if (ip6_del_rt(rt))
1142				dst_free(&rt->dst);
1143		} else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1144			rt->rt6i_node->fn_sernum = -1;
 
 
 
 
 
 
1145		}
 
1146	}
1147}
1148
1149static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1150			       struct sk_buff *skb, u32 mtu)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1151{
1152	struct rt6_info *rt6 = (struct rt6_info*)dst;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1153
1154	dst_confirm(dst);
1155	if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1156		struct net *net = dev_net(dst->dev);
1157
1158		rt6->rt6i_flags |= RTF_MODIFIED;
1159		if (mtu < IPV6_MIN_MTU) {
1160			u32 features = dst_metric(dst, RTAX_FEATURES);
1161			mtu = IPV6_MIN_MTU;
1162			features |= RTAX_FEATURE_ALLFRAG;
1163			dst_metric_set(dst, RTAX_FEATURES, features);
1164		}
1165		dst_metric_set(dst, RTAX_MTU, mtu);
1166		rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1167	}
1168}
1169
 
 
 
 
 
 
 
 
1170void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1171		     int oif, u32 mark)
1172{
1173	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1174	struct dst_entry *dst;
1175	struct flowi6 fl6;
1176
1177	memset(&fl6, 0, sizeof(fl6));
1178	fl6.flowi6_oif = oif;
1179	fl6.flowi6_mark = mark;
1180	fl6.daddr = iph->daddr;
1181	fl6.saddr = iph->saddr;
1182	fl6.flowlabel = ip6_flowinfo(iph);
1183
1184	dst = ip6_route_output(net, NULL, &fl6);
1185	if (!dst->error)
1186		ip6_rt_update_pmtu(dst, NULL, skb, ntohl(mtu));
1187	dst_release(dst);
1188}
1189EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1190
1191void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1192{
1193	ip6_update_pmtu(skb, sock_net(sk), mtu,
1194			sk->sk_bound_dev_if, sk->sk_mark);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1195}
1196EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1197
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1198/* Handle redirects */
1199struct ip6rd_flowi {
1200	struct flowi6 fl6;
1201	struct in6_addr gateway;
1202};
1203
1204static struct rt6_info *__ip6_route_redirect(struct net *net,
1205					     struct fib6_table *table,
1206					     struct flowi6 *fl6,
 
1207					     int flags)
1208{
1209	struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1210	struct rt6_info *rt;
 
 
 
 
 
 
 
 
1211	struct fib6_node *fn;
1212
1213	/* Get the "current" route for this destination and
1214	 * check if the redirect has come from approriate router.
1215	 *
1216	 * RFC 4861 specifies that redirects should only be
1217	 * accepted if they come from the nexthop to the target.
1218	 * Due to the way the routes are chosen, this notion
1219	 * is a bit fuzzy and one might need to check all possible
1220	 * routes.
1221	 */
1222
1223	read_lock_bh(&table->tb6_lock);
1224	fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1225restart:
1226	for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1227		if (rt6_check_expired(rt))
 
1228			continue;
1229		if (rt->dst.error)
1230			break;
1231		if (!(rt->rt6i_flags & RTF_GATEWAY))
1232			continue;
1233		if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1234			continue;
1235		if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1236			continue;
1237		break;
 
 
 
 
 
 
 
1238	}
1239
1240	if (!rt)
1241		rt = net->ipv6.ip6_null_entry;
1242	else if (rt->dst.error) {
1243		rt = net->ipv6.ip6_null_entry;
1244		goto out;
1245	}
1246	BACKTRACK(net, &fl6->saddr);
 
 
 
 
 
 
 
 
1247out:
1248	dst_hold(&rt->dst);
 
 
 
 
 
 
1249
1250	read_unlock_bh(&table->tb6_lock);
1251
1252	return rt;
 
1253};
1254
1255static struct dst_entry *ip6_route_redirect(struct net *net,
1256					const struct flowi6 *fl6,
1257					const struct in6_addr *gateway)
 
1258{
1259	int flags = RT6_LOOKUP_F_HAS_SADDR;
1260	struct ip6rd_flowi rdfl;
1261
1262	rdfl.fl6 = *fl6;
1263	rdfl.gateway = *gateway;
1264
1265	return fib6_rule_lookup(net, &rdfl.fl6,
1266				flags, __ip6_route_redirect);
1267}
1268
1269void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
 
1270{
1271	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1272	struct dst_entry *dst;
1273	struct flowi6 fl6;
1274
1275	memset(&fl6, 0, sizeof(fl6));
1276	fl6.flowi6_iif = LOOPBACK_IFINDEX;
1277	fl6.flowi6_oif = oif;
1278	fl6.flowi6_mark = mark;
1279	fl6.daddr = iph->daddr;
1280	fl6.saddr = iph->saddr;
1281	fl6.flowlabel = ip6_flowinfo(iph);
1282
1283	dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1284	rt6_do_redirect(dst, NULL, skb);
1285	dst_release(dst);
1286}
1287EXPORT_SYMBOL_GPL(ip6_redirect);
1288
1289void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1290			    u32 mark)
1291{
1292	const struct ipv6hdr *iph = ipv6_hdr(skb);
1293	const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1294	struct dst_entry *dst;
1295	struct flowi6 fl6;
1296
1297	memset(&fl6, 0, sizeof(fl6));
1298	fl6.flowi6_iif = LOOPBACK_IFINDEX;
1299	fl6.flowi6_oif = oif;
1300	fl6.flowi6_mark = mark;
1301	fl6.daddr = msg->dest;
1302	fl6.saddr = iph->daddr;
1303
1304	dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1305	rt6_do_redirect(dst, NULL, skb);
1306	dst_release(dst);
1307}
1308
1309void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1310{
1311	ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
 
1312}
1313EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1314
1315static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1316{
1317	struct net_device *dev = dst->dev;
1318	unsigned int mtu = dst_mtu(dst);
1319	struct net *net = dev_net(dev);
1320
1321	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1322
 
 
 
1323	if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1324		mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1325
 
 
1326	/*
1327	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1328	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1329	 * IPV6_MAXPLEN is also valid and means: "any MSS,
1330	 * rely only on pmtu discovery"
1331	 */
1332	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1333		mtu = IPV6_MAXPLEN;
1334	return mtu;
1335}
1336
1337static unsigned int ip6_mtu(const struct dst_entry *dst)
1338{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1339	struct inet6_dev *idev;
1340	unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 
1341
1342	if (mtu)
1343		goto out;
 
 
 
1344
1345	mtu = IPV6_MIN_MTU;
 
 
 
 
1346
1347	rcu_read_lock();
1348	idev = __in6_dev_get(dst->dev);
1349	if (idev)
1350		mtu = idev->cnf.mtu6;
1351	rcu_read_unlock();
1352
 
1353out:
1354	return min_t(unsigned int, mtu, IP6_MAX_MTU);
1355}
1356
1357static struct dst_entry *icmp6_dst_gc_list;
1358static DEFINE_SPINLOCK(icmp6_dst_lock);
1359
1360struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1361				  struct flowi6 *fl6)
1362{
1363	struct dst_entry *dst;
1364	struct rt6_info *rt;
1365	struct inet6_dev *idev = in6_dev_get(dev);
1366	struct net *net = dev_net(dev);
1367
1368	if (unlikely(!idev))
1369		return ERR_PTR(-ENODEV);
1370
1371	rt = ip6_dst_alloc(net, dev, 0, NULL);
1372	if (unlikely(!rt)) {
1373		in6_dev_put(idev);
1374		dst = ERR_PTR(-ENOMEM);
1375		goto out;
1376	}
1377
1378	rt->dst.flags |= DST_HOST;
1379	rt->dst.output  = ip6_output;
1380	atomic_set(&rt->dst.__refcnt, 1);
1381	rt->rt6i_gateway  = fl6->daddr;
1382	rt->rt6i_dst.addr = fl6->daddr;
1383	rt->rt6i_dst.plen = 128;
1384	rt->rt6i_idev     = idev;
1385	dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1386
1387	spin_lock_bh(&icmp6_dst_lock);
1388	rt->dst.next = icmp6_dst_gc_list;
1389	icmp6_dst_gc_list = &rt->dst;
1390	spin_unlock_bh(&icmp6_dst_lock);
1391
1392	fib6_force_start_gc(net);
1393
1394	dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1395
1396out:
1397	return dst;
1398}
1399
1400int icmp6_dst_gc(void)
1401{
1402	struct dst_entry *dst, **pprev;
1403	int more = 0;
 
 
 
 
 
1404
1405	spin_lock_bh(&icmp6_dst_lock);
1406	pprev = &icmp6_dst_gc_list;
1407
1408	while ((dst = *pprev) != NULL) {
1409		if (!atomic_read(&dst->__refcnt)) {
1410			*pprev = dst->next;
1411			dst_free(dst);
1412		} else {
1413			pprev = &dst->next;
1414			++more;
1415		}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1416	}
1417
1418	spin_unlock_bh(&icmp6_dst_lock);
 
 
 
 
 
 
 
 
1419
1420	return more;
1421}
1422
1423static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1424			    void *arg)
1425{
1426	struct dst_entry *dst, **pprev;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1427
1428	spin_lock_bh(&icmp6_dst_lock);
1429	pprev = &icmp6_dst_gc_list;
1430	while ((dst = *pprev) != NULL) {
1431		struct rt6_info *rt = (struct rt6_info *) dst;
1432		if (func(rt, arg)) {
1433			*pprev = dst->next;
1434			dst_free(dst);
1435		} else {
1436			pprev = &dst->next;
 
 
 
 
1437		}
 
 
 
 
 
 
 
 
 
 
 
 
 
1438	}
1439	spin_unlock_bh(&icmp6_dst_lock);
1440}
1441
1442static int ip6_dst_gc(struct dst_ops *ops)
1443{
1444	struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1445	int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1446	int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1447	int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1448	int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1449	unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1450	int entries;
1451
1452	entries = dst_entries_get_fast(ops);
1453	if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1454	    entries <= rt_max_size)
 
 
 
 
1455		goto out;
 
1456
1457	net->ipv6.ip6_rt_gc_expire++;
1458	fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, entries > rt_max_size);
1459	entries = dst_entries_get_slow(ops);
1460	if (entries < ops->gc_thresh)
1461		net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
 
 
 
 
 
1462out:
1463	net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1464	return entries > rt_max_size;
1465}
1466
1467/*
1468 *
1469 */
 
 
 
 
 
 
 
1470
1471int ip6_route_add(struct fib6_config *cfg)
 
 
1472{
1473	int err;
1474	struct net *net = cfg->fc_nlinfo.nl_net;
1475	struct rt6_info *rt = NULL;
1476	struct net_device *dev = NULL;
1477	struct inet6_dev *idev = NULL;
1478	struct fib6_table *table;
1479	int addr_type;
 
1480
1481	if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1482		return -EINVAL;
1483#ifndef CONFIG_IPV6_SUBTREES
1484	if (cfg->fc_src_len)
1485		return -EINVAL;
1486#endif
 
 
 
 
 
 
 
1487	if (cfg->fc_ifindex) {
1488		err = -ENODEV;
1489		dev = dev_get_by_index(net, cfg->fc_ifindex);
1490		if (!dev)
1491			goto out;
1492		idev = in6_dev_get(dev);
1493		if (!idev)
1494			goto out;
1495	}
1496
1497	if (cfg->fc_metric == 0)
1498		cfg->fc_metric = IP6_RT_PRIO_USER;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1499
1500	err = -ENOBUFS;
1501	if (cfg->fc_nlinfo.nlh &&
1502	    !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1503		table = fib6_get_table(net, cfg->fc_table);
1504		if (!table) {
1505			pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1506			table = fib6_new_table(net, cfg->fc_table);
1507		}
1508	} else {
1509		table = fib6_new_table(net, cfg->fc_table);
1510	}
1511
1512	if (!table)
1513		goto out;
1514
1515	rt = ip6_dst_alloc(net, NULL, (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT, table);
1516
1517	if (!rt) {
1518		err = -ENOMEM;
1519		goto out;
 
 
 
 
 
 
 
 
1520	}
1521
 
 
 
1522	if (cfg->fc_flags & RTF_EXPIRES)
1523		rt6_set_expires(rt, jiffies +
1524				clock_t_to_jiffies(cfg->fc_expires));
1525	else
1526		rt6_clean_expires(rt);
1527
1528	if (cfg->fc_protocol == RTPROT_UNSPEC)
1529		cfg->fc_protocol = RTPROT_BOOT;
1530	rt->rt6i_protocol = cfg->fc_protocol;
1531
1532	addr_type = ipv6_addr_type(&cfg->fc_dst);
 
 
 
1533
1534	if (addr_type & IPV6_ADDR_MULTICAST)
1535		rt->dst.input = ip6_mc_input;
1536	else if (cfg->fc_flags & RTF_LOCAL)
1537		rt->dst.input = ip6_input;
1538	else
1539		rt->dst.input = ip6_forward;
1540
1541	rt->dst.output = ip6_output;
1542
1543	ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1544	rt->rt6i_dst.plen = cfg->fc_dst_len;
1545	if (rt->rt6i_dst.plen == 128) {
1546		rt->dst.flags |= DST_HOST;
1547		dst_metrics_set_force_overwrite(&rt->dst);
1548	}
1549
1550#ifdef CONFIG_IPV6_SUBTREES
1551	ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1552	rt->rt6i_src.plen = cfg->fc_src_len;
1553#endif
1554
1555	rt->rt6i_metric = cfg->fc_metric;
1556
1557	/* We cannot add true routes via loopback here,
1558	   they would result in kernel looping; promote them to reject routes
1559	 */
1560	if ((cfg->fc_flags & RTF_REJECT) ||
1561	    (dev && (dev->flags & IFF_LOOPBACK) &&
1562	     !(addr_type & IPV6_ADDR_LOOPBACK) &&
1563	     !(cfg->fc_flags & RTF_LOCAL))) {
1564		/* hold loopback dev/idev if we haven't done so. */
1565		if (dev != net->loopback_dev) {
1566			if (dev) {
1567				dev_put(dev);
1568				in6_dev_put(idev);
1569			}
1570			dev = net->loopback_dev;
1571			dev_hold(dev);
1572			idev = in6_dev_get(dev);
1573			if (!idev) {
1574				err = -ENODEV;
1575				goto out;
1576			}
1577		}
1578		rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1579		switch (cfg->fc_type) {
1580		case RTN_BLACKHOLE:
1581			rt->dst.error = -EINVAL;
1582			rt->dst.output = dst_discard_sk;
1583			rt->dst.input = dst_discard;
1584			break;
1585		case RTN_PROHIBIT:
1586			rt->dst.error = -EACCES;
1587			rt->dst.output = ip6_pkt_prohibit_out;
1588			rt->dst.input = ip6_pkt_prohibit;
1589			break;
1590		case RTN_THROW:
1591		default:
1592			rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1593					: -ENETUNREACH;
1594			rt->dst.output = ip6_pkt_discard_out;
1595			rt->dst.input = ip6_pkt_discard;
1596			break;
1597		}
1598		goto install_route;
1599	}
1600
1601	if (cfg->fc_flags & RTF_GATEWAY) {
1602		const struct in6_addr *gw_addr;
1603		int gwa_type;
1604
1605		gw_addr = &cfg->fc_gateway;
1606		rt->rt6i_gateway = *gw_addr;
1607		gwa_type = ipv6_addr_type(gw_addr);
1608
1609		if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1610			struct rt6_info *grt;
1611
1612			/* IPv6 strictly inhibits using not link-local
1613			   addresses as nexthop address.
1614			   Otherwise, router will not able to send redirects.
1615			   It is very good, but in some (rare!) circumstances
1616			   (SIT, PtP, NBMA NOARP links) it is handy to allow
1617			   some exceptions. --ANK
1618			 */
1619			err = -EINVAL;
1620			if (!(gwa_type & IPV6_ADDR_UNICAST))
1621				goto out;
1622
1623			grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1624
1625			err = -EHOSTUNREACH;
1626			if (!grt)
1627				goto out;
1628			if (dev) {
1629				if (dev != grt->dst.dev) {
1630					ip6_rt_put(grt);
1631					goto out;
1632				}
1633			} else {
1634				dev = grt->dst.dev;
1635				idev = grt->rt6i_idev;
1636				dev_hold(dev);
1637				in6_dev_hold(grt->rt6i_idev);
1638			}
1639			if (!(grt->rt6i_flags & RTF_GATEWAY))
1640				err = 0;
1641			ip6_rt_put(grt);
1642
1643			if (err)
1644				goto out;
1645		}
1646		err = -EINVAL;
1647		if (!dev || (dev->flags & IFF_LOOPBACK))
1648			goto out;
 
1649	}
1650
1651	err = -ENODEV;
1652	if (!dev)
1653		goto out;
1654
1655	if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
 
 
1656		if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
 
1657			err = -EINVAL;
1658			goto out;
1659		}
1660		rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1661		rt->rt6i_prefsrc.plen = 128;
1662	} else
1663		rt->rt6i_prefsrc.plen = 0;
1664
1665	rt->rt6i_flags = cfg->fc_flags;
 
 
 
 
 
 
 
 
1666
1667install_route:
1668	rt->dst.dev = dev;
1669	rt->rt6i_idev = idev;
1670	rt->rt6i_table = table;
 
1671
1672	cfg->fc_nlinfo.nl_net = dev_net(dev);
 
 
1673
1674	return __ip6_ins_rt(rt, &cfg->fc_nlinfo, cfg->fc_mx, cfg->fc_mx_len);
 
1675
1676out:
1677	if (dev)
1678		dev_put(dev);
1679	if (idev)
1680		in6_dev_put(idev);
1681	if (rt)
1682		dst_free(&rt->dst);
1683	return err;
1684}
1685
1686static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1687{
1688	int err;
1689	struct fib6_table *table;
1690	struct net *net = dev_net(rt->dst.dev);
1691
1692	if (rt == net->ipv6.ip6_null_entry) {
1693		err = -ENOENT;
1694		goto out;
1695	}
1696
1697	table = rt->rt6i_table;
1698	write_lock_bh(&table->tb6_lock);
1699	err = fib6_del(rt, info);
1700	write_unlock_bh(&table->tb6_lock);
1701
1702out:
1703	ip6_rt_put(rt);
1704	return err;
1705}
1706
1707int ip6_del_rt(struct rt6_info *rt)
1708{
1709	struct nl_info info = {
1710		.nl_net = dev_net(rt->dst.dev),
 
1711	};
 
1712	return __ip6_del_rt(rt, &info);
1713}
1714
1715static int ip6_route_del(struct fib6_config *cfg)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1716{
1717	struct fib6_table *table;
 
1718	struct fib6_node *fn;
1719	struct rt6_info *rt;
1720	int err = -ESRCH;
1721
1722	table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1723	if (!table)
 
1724		return err;
 
1725
1726	read_lock_bh(&table->tb6_lock);
1727
1728	fn = fib6_locate(&table->tb6_root,
1729			 &cfg->fc_dst, cfg->fc_dst_len,
1730			 &cfg->fc_src, cfg->fc_src_len);
 
1731
1732	if (fn) {
1733		for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1734			if (cfg->fc_ifindex &&
1735			    (!rt->dst.dev ||
1736			     rt->dst.dev->ifindex != cfg->fc_ifindex))
1737				continue;
1738			if (cfg->fc_flags & RTF_GATEWAY &&
1739			    !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1740				continue;
1741			if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1742				continue;
1743			dst_hold(&rt->dst);
1744			read_unlock_bh(&table->tb6_lock);
1745
1746			return __ip6_del_rt(rt, &cfg->fc_nlinfo);
 
 
 
 
1747		}
1748	}
1749	read_unlock_bh(&table->tb6_lock);
1750
1751	return err;
1752}
1753
1754static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
1755{
1756	struct net *net = dev_net(skb->dev);
1757	struct netevent_redirect netevent;
1758	struct rt6_info *rt, *nrt = NULL;
 
1759	struct ndisc_options ndopts;
1760	struct inet6_dev *in6_dev;
1761	struct neighbour *neigh;
1762	struct rd_msg *msg;
1763	int optlen, on_link;
1764	u8 *lladdr;
1765
1766	optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1767	optlen -= sizeof(*msg);
1768
1769	if (optlen < 0) {
1770		net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1771		return;
1772	}
1773
1774	msg = (struct rd_msg *)icmp6_hdr(skb);
1775
1776	if (ipv6_addr_is_multicast(&msg->dest)) {
1777		net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1778		return;
1779	}
1780
1781	on_link = 0;
1782	if (ipv6_addr_equal(&msg->dest, &msg->target)) {
1783		on_link = 1;
1784	} else if (ipv6_addr_type(&msg->target) !=
1785		   (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
1786		net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1787		return;
1788	}
1789
1790	in6_dev = __in6_dev_get(skb->dev);
1791	if (!in6_dev)
1792		return;
1793	if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
 
1794		return;
1795
1796	/* RFC2461 8.1:
1797	 *	The IP source address of the Redirect MUST be the same as the current
1798	 *	first-hop router for the specified ICMP Destination Address.
1799	 */
1800
1801	if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
1802		net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1803		return;
1804	}
1805
1806	lladdr = NULL;
1807	if (ndopts.nd_opts_tgt_lladdr) {
1808		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
1809					     skb->dev);
1810		if (!lladdr) {
1811			net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1812			return;
1813		}
1814	}
1815
1816	rt = (struct rt6_info *) dst;
1817	if (rt == net->ipv6.ip6_null_entry) {
1818		net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1819		return;
1820	}
1821
1822	/* Redirect received -> path was valid.
1823	 * Look, redirects are sent only in response to data packets,
1824	 * so that this nexthop apparently is reachable. --ANK
1825	 */
1826	dst_confirm(&rt->dst);
1827
1828	neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
1829	if (!neigh)
1830		return;
1831
1832	/*
1833	 *	We have finally decided to accept it.
1834	 */
1835
1836	neigh_update(neigh, lladdr, NUD_STALE,
1837		     NEIGH_UPDATE_F_WEAK_OVERRIDE|
1838		     NEIGH_UPDATE_F_OVERRIDE|
1839		     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1840				     NEIGH_UPDATE_F_ISROUTER))
1841		     );
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1842
1843	nrt = ip6_rt_copy(rt, &msg->dest);
 
 
 
 
 
 
 
 
 
 
 
 
1844	if (!nrt)
1845		goto out;
1846
1847	nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1848	if (on_link)
1849		nrt->rt6i_flags &= ~RTF_GATEWAY;
1850
1851	nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1852
1853	if (ip6_ins_rt(nrt))
 
 
1854		goto out;
 
1855
1856	netevent.old = &rt->dst;
1857	netevent.new = &nrt->dst;
1858	netevent.daddr = &msg->dest;
1859	netevent.neigh = neigh;
1860	call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1861
1862	if (rt->rt6i_flags & RTF_CACHE) {
1863		rt = (struct rt6_info *) dst_clone(&rt->dst);
1864		ip6_del_rt(rt);
1865	}
1866
1867out:
 
1868	neigh_release(neigh);
1869}
1870
1871/*
1872 *	Misc support functions
1873 */
1874
1875static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1876				    const struct in6_addr *dest)
1877{
1878	struct net *net = dev_net(ort->dst.dev);
1879	struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1880					    ort->rt6i_table);
1881
1882	if (rt) {
1883		rt->dst.input = ort->dst.input;
1884		rt->dst.output = ort->dst.output;
1885		rt->dst.flags |= DST_HOST;
1886
1887		rt->rt6i_dst.addr = *dest;
1888		rt->rt6i_dst.plen = 128;
1889		dst_copy_metrics(&rt->dst, &ort->dst);
1890		rt->dst.error = ort->dst.error;
1891		rt->rt6i_idev = ort->rt6i_idev;
1892		if (rt->rt6i_idev)
1893			in6_dev_hold(rt->rt6i_idev);
1894		rt->dst.lastuse = jiffies;
1895
1896		if (ort->rt6i_flags & RTF_GATEWAY)
1897			rt->rt6i_gateway = ort->rt6i_gateway;
1898		else
1899			rt->rt6i_gateway = *dest;
1900		rt->rt6i_flags = ort->rt6i_flags;
1901		rt6_set_from(rt, ort);
1902		rt->rt6i_metric = 0;
1903
1904#ifdef CONFIG_IPV6_SUBTREES
1905		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1906#endif
1907		memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1908		rt->rt6i_table = ort->rt6i_table;
1909	}
1910	return rt;
1911}
1912
1913#ifdef CONFIG_IPV6_ROUTE_INFO
1914static struct rt6_info *rt6_get_route_info(struct net *net,
1915					   const struct in6_addr *prefix, int prefixlen,
1916					   const struct in6_addr *gwaddr, int ifindex)
 
1917{
 
 
1918	struct fib6_node *fn;
1919	struct rt6_info *rt = NULL;
1920	struct fib6_table *table;
1921
1922	table = fib6_get_table(net, RT6_TABLE_INFO);
1923	if (!table)
1924		return NULL;
1925
1926	read_lock_bh(&table->tb6_lock);
1927	fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1928	if (!fn)
1929		goto out;
1930
1931	for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1932		if (rt->dst.dev->ifindex != ifindex)
 
1933			continue;
1934		if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1935			continue;
1936		if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
 
 
 
 
 
1937			continue;
1938		dst_hold(&rt->dst);
1939		break;
1940	}
1941out:
1942	read_unlock_bh(&table->tb6_lock);
1943	return rt;
1944}
1945
1946static struct rt6_info *rt6_add_route_info(struct net *net,
1947					   const struct in6_addr *prefix, int prefixlen,
1948					   const struct in6_addr *gwaddr, int ifindex,
 
1949					   unsigned int pref)
1950{
1951	struct fib6_config cfg = {
1952		.fc_table	= RT6_TABLE_INFO,
1953		.fc_metric	= IP6_RT_PRIO_USER,
1954		.fc_ifindex	= ifindex,
1955		.fc_dst_len	= prefixlen,
1956		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1957				  RTF_UP | RTF_PREF(pref),
 
 
1958		.fc_nlinfo.portid = 0,
1959		.fc_nlinfo.nlh = NULL,
1960		.fc_nlinfo.nl_net = net,
1961	};
1962
 
1963	cfg.fc_dst = *prefix;
1964	cfg.fc_gateway = *gwaddr;
1965
1966	/* We should treat it as a default route if prefix length is 0. */
1967	if (!prefixlen)
1968		cfg.fc_flags |= RTF_DEFAULT;
1969
1970	ip6_route_add(&cfg);
1971
1972	return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1973}
1974#endif
1975
1976struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
 
 
1977{
1978	struct rt6_info *rt;
 
1979	struct fib6_table *table;
1980
1981	table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1982	if (!table)
1983		return NULL;
1984
1985	read_lock_bh(&table->tb6_lock);
1986	for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1987		if (dev == rt->dst.dev &&
1988		    ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1989		    ipv6_addr_equal(&rt->rt6i_gateway, addr))
 
 
 
 
 
 
 
1990			break;
1991	}
1992	if (rt)
1993		dst_hold(&rt->dst);
1994	read_unlock_bh(&table->tb6_lock);
1995	return rt;
1996}
1997
1998struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
 
1999				     struct net_device *dev,
2000				     unsigned int pref)
 
 
2001{
2002	struct fib6_config cfg = {
2003		.fc_table	= RT6_TABLE_DFLT,
2004		.fc_metric	= IP6_RT_PRIO_USER,
2005		.fc_ifindex	= dev->ifindex,
2006		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2007				  RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
 
 
2008		.fc_nlinfo.portid = 0,
2009		.fc_nlinfo.nlh = NULL,
2010		.fc_nlinfo.nl_net = dev_net(dev),
 
2011	};
2012
2013	cfg.fc_gateway = *gwaddr;
2014
2015	ip6_route_add(&cfg);
 
 
 
 
 
 
2016
2017	return rt6_get_dflt_router(gwaddr, dev);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2018}
2019
2020void rt6_purge_dflt_routers(struct net *net)
2021{
2022	struct rt6_info *rt;
2023	struct fib6_table *table;
 
 
2024
2025	/* NOTE: Keep consistent with rt6_get_dflt_router */
2026	table = fib6_get_table(net, RT6_TABLE_DFLT);
2027	if (!table)
2028		return;
2029
2030restart:
2031	read_lock_bh(&table->tb6_lock);
2032	for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2033		if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2034		    (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2035			dst_hold(&rt->dst);
2036			read_unlock_bh(&table->tb6_lock);
2037			ip6_del_rt(rt);
2038			goto restart;
2039		}
2040	}
2041	read_unlock_bh(&table->tb6_lock);
 
2042}
2043
2044static void rtmsg_to_fib6_config(struct net *net,
2045				 struct in6_rtmsg *rtmsg,
2046				 struct fib6_config *cfg)
2047{
2048	memset(cfg, 0, sizeof(*cfg));
 
 
 
 
 
 
 
 
 
2049
2050	cfg->fc_table = RT6_TABLE_MAIN;
2051	cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2052	cfg->fc_metric = rtmsg->rtmsg_metric;
2053	cfg->fc_expires = rtmsg->rtmsg_info;
2054	cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2055	cfg->fc_src_len = rtmsg->rtmsg_src_len;
2056	cfg->fc_flags = rtmsg->rtmsg_flags;
2057
2058	cfg->fc_nlinfo.nl_net = net;
2059
2060	cfg->fc_dst = rtmsg->rtmsg_dst;
2061	cfg->fc_src = rtmsg->rtmsg_src;
2062	cfg->fc_gateway = rtmsg->rtmsg_gateway;
 
2063}
2064
2065int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2066{
2067	struct fib6_config cfg;
2068	struct in6_rtmsg rtmsg;
2069	int err;
2070
2071	switch(cmd) {
2072	case SIOCADDRT:		/* Add a route */
2073	case SIOCDELRT:		/* Delete a route */
2074		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2075			return -EPERM;
2076		err = copy_from_user(&rtmsg, arg,
2077				     sizeof(struct in6_rtmsg));
2078		if (err)
2079			return -EFAULT;
2080
2081		rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2082
2083		rtnl_lock();
2084		switch (cmd) {
2085		case SIOCADDRT:
2086			err = ip6_route_add(&cfg);
2087			break;
2088		case SIOCDELRT:
2089			err = ip6_route_del(&cfg);
2090			break;
2091		default:
2092			err = -EINVAL;
2093		}
2094		rtnl_unlock();
2095
2096		return err;
 
 
 
 
 
 
 
 
 
 
2097	}
2098
2099	return -EINVAL;
2100}
2101
2102/*
2103 *	Drop the packet on the floor
2104 */
2105
2106static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2107{
2108	int type;
2109	struct dst_entry *dst = skb_dst(skb);
 
 
 
 
 
 
 
 
 
 
 
2110	switch (ipstats_mib_noroutes) {
2111	case IPSTATS_MIB_INNOROUTES:
2112		type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2113		if (type == IPV6_ADDR_ANY) {
2114			IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2115				      IPSTATS_MIB_INADDRERRORS);
2116			break;
2117		}
2118		/* FALLTHROUGH */
 
2119	case IPSTATS_MIB_OUTNOROUTES:
2120		IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2121			      ipstats_mib_noroutes);
2122		break;
2123	}
 
 
 
 
 
2124	icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2125	kfree_skb(skb);
2126	return 0;
2127}
2128
2129static int ip6_pkt_discard(struct sk_buff *skb)
2130{
2131	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2132}
2133
2134static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2135{
2136	skb->dev = skb_dst(skb)->dev;
2137	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2138}
2139
2140static int ip6_pkt_prohibit(struct sk_buff *skb)
2141{
2142	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2143}
2144
2145static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2146{
2147	skb->dev = skb_dst(skb)->dev;
2148	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2149}
2150
2151/*
2152 *	Allocate a dst for local (unicast / anycast) address.
2153 */
2154
2155struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2156				    const struct in6_addr *addr,
2157				    bool anycast)
2158{
2159	struct net *net = dev_net(idev->dev);
2160	struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2161					    DST_NOCOUNT, NULL);
2162	if (!rt)
2163		return ERR_PTR(-ENOMEM);
2164
2165	in6_dev_hold(idev);
2166
2167	rt->dst.flags |= DST_HOST;
2168	rt->dst.input = ip6_input;
2169	rt->dst.output = ip6_output;
2170	rt->rt6i_idev = idev;
2171
2172	rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2173	if (anycast)
2174		rt->rt6i_flags |= RTF_ANYCAST;
2175	else
2176		rt->rt6i_flags |= RTF_LOCAL;
2177
2178	rt->rt6i_gateway  = *addr;
2179	rt->rt6i_dst.addr = *addr;
2180	rt->rt6i_dst.plen = 128;
2181	rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
 
 
 
2182
2183	atomic_set(&rt->dst.__refcnt, 1);
 
 
2184
2185	return rt;
2186}
 
 
 
2187
2188int ip6_route_get_saddr(struct net *net,
2189			struct rt6_info *rt,
2190			const struct in6_addr *daddr,
2191			unsigned int prefs,
2192			struct in6_addr *saddr)
2193{
2194	struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2195	int err = 0;
2196	if (rt->rt6i_prefsrc.plen)
2197		*saddr = rt->rt6i_prefsrc.addr;
2198	else
2199		err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2200					 daddr, prefs, saddr);
2201	return err;
2202}
2203
2204/* remove deleted ip from prefsrc entries */
2205struct arg_dev_net_ip {
2206	struct net_device *dev;
2207	struct net *net;
2208	struct in6_addr *addr;
2209};
2210
2211static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2212{
2213	struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2214	struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2215	struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2216
2217	if (((void *)rt->dst.dev == dev || !dev) &&
2218	    rt != net->ipv6.ip6_null_entry &&
2219	    ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
 
 
2220		/* remove prefsrc entry */
2221		rt->rt6i_prefsrc.plen = 0;
 
2222	}
2223	return 0;
2224}
2225
2226void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2227{
2228	struct net *net = dev_net(ifp->idev->dev);
2229	struct arg_dev_net_ip adni = {
2230		.dev = ifp->idev->dev,
2231		.net = net,
2232		.addr = &ifp->addr,
2233	};
2234	fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2235}
2236
2237#define RTF_RA_ROUTER		(RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2238#define RTF_CACHE_GATEWAY	(RTF_GATEWAY | RTF_CACHE)
2239
2240/* Remove routers and update dst entries when gateway turn into host. */
2241static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2242{
2243	struct in6_addr *gateway = (struct in6_addr *)arg;
 
 
 
 
 
2244
2245	if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2246	     ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2247	     ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2248		return -1;
2249	}
 
 
 
 
 
 
2250	return 0;
2251}
2252
2253void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2254{
2255	fib6_clean_all(net, fib6_clean_tohost, gateway);
2256}
2257
2258struct arg_dev_net {
2259	struct net_device *dev;
2260	struct net *net;
 
 
 
2261};
2262
2263static int fib6_ifdown(struct rt6_info *rt, void *arg)
2264{
2265	const struct arg_dev_net *adn = arg;
2266	const struct net_device *dev = adn->dev;
2267
2268	if ((rt->dst.dev == dev || !dev) &&
2269	    rt != adn->net->ipv6.ip6_null_entry)
2270		return -1;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2271
2272	return 0;
2273}
2274
2275void rt6_ifdown(struct net *net, struct net_device *dev)
2276{
2277	struct arg_dev_net adn = {
2278		.dev = dev,
2279		.net = net,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2280	};
 
 
 
 
 
 
 
2281
2282	fib6_clean_all(net, fib6_ifdown, &adn);
2283	icmp6_clean_all(fib6_ifdown, &adn);
 
 
 
2284}
2285
2286struct rt6_mtu_change_arg {
2287	struct net_device *dev;
2288	unsigned int mtu;
 
2289};
2290
2291static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2292{
2293	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2294	struct inet6_dev *idev;
2295
2296	/* In IPv6 pmtu discovery is not optional,
2297	   so that RTAX_MTU lock cannot disable it.
2298	   We still use this lock to block changes
2299	   caused by addrconf/ndisc.
2300	*/
2301
2302	idev = __in6_dev_get(arg->dev);
2303	if (!idev)
2304		return 0;
2305
2306	/* For administrative MTU increase, there is no way to discover
2307	   IPv6 PMTU increase, so PMTU increase should be updated here.
2308	   Since RFC 1981 doesn't include administrative MTU increase
2309	   update PMTU increase is a MUST. (i.e. jumbo frame)
2310	 */
2311	/*
2312	   If new MTU is less than route PMTU, this new MTU will be the
2313	   lowest MTU in the path, update the route PMTU to reflect PMTU
2314	   decreases; if new MTU is greater than route PMTU, and the
2315	   old MTU is the lowest MTU in the path, update the route PMTU
2316	   to reflect the increase. In this case if the other nodes' MTU
2317	   also have the lowest MTU, TOO BIG MESSAGE will be lead to
2318	   PMTU discouvery.
2319	 */
2320	if (rt->dst.dev == arg->dev &&
2321	    !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2322	    (dst_mtu(&rt->dst) >= arg->mtu ||
2323	     (dst_mtu(&rt->dst) < arg->mtu &&
2324	      dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2325		dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2326	}
2327	return 0;
 
2328}
2329
2330void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2331{
2332	struct rt6_mtu_change_arg arg = {
2333		.dev = dev,
2334		.mtu = mtu,
2335	};
2336
2337	fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2338}
2339
2340static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
 
2341	[RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
 
2342	[RTA_OIF]               = { .type = NLA_U32 },
2343	[RTA_IIF]		= { .type = NLA_U32 },
2344	[RTA_PRIORITY]          = { .type = NLA_U32 },
2345	[RTA_METRICS]           = { .type = NLA_NESTED },
2346	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
 
 
 
 
 
 
 
 
 
 
 
2347};
2348
2349static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2350			      struct fib6_config *cfg)
 
2351{
2352	struct rtmsg *rtm;
2353	struct nlattr *tb[RTA_MAX+1];
 
2354	int err;
2355
2356	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
 
2357	if (err < 0)
2358		goto errout;
2359
2360	err = -EINVAL;
2361	rtm = nlmsg_data(nlh);
2362	memset(cfg, 0, sizeof(*cfg));
2363
2364	cfg->fc_table = rtm->rtm_table;
2365	cfg->fc_dst_len = rtm->rtm_dst_len;
2366	cfg->fc_src_len = rtm->rtm_src_len;
2367	cfg->fc_flags = RTF_UP;
2368	cfg->fc_protocol = rtm->rtm_protocol;
2369	cfg->fc_type = rtm->rtm_type;
 
 
 
 
 
 
 
 
 
 
 
 
2370
2371	if (rtm->rtm_type == RTN_UNREACHABLE ||
2372	    rtm->rtm_type == RTN_BLACKHOLE ||
2373	    rtm->rtm_type == RTN_PROHIBIT ||
2374	    rtm->rtm_type == RTN_THROW)
2375		cfg->fc_flags |= RTF_REJECT;
2376
2377	if (rtm->rtm_type == RTN_LOCAL)
2378		cfg->fc_flags |= RTF_LOCAL;
2379
2380	cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2381	cfg->fc_nlinfo.nlh = nlh;
2382	cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
 
 
 
 
 
 
 
 
 
 
 
2383
2384	if (tb[RTA_GATEWAY]) {
2385		nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2386		cfg->fc_flags |= RTF_GATEWAY;
2387	}
 
 
 
 
2388
2389	if (tb[RTA_DST]) {
2390		int plen = (rtm->rtm_dst_len + 7) >> 3;
2391
2392		if (nla_len(tb[RTA_DST]) < plen)
2393			goto errout;
2394
2395		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2396	}
2397
2398	if (tb[RTA_SRC]) {
2399		int plen = (rtm->rtm_src_len + 7) >> 3;
2400
2401		if (nla_len(tb[RTA_SRC]) < plen)
2402			goto errout;
2403
2404		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2405	}
2406
2407	if (tb[RTA_PREFSRC])
2408		nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2409
2410	if (tb[RTA_OIF])
2411		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2412
2413	if (tb[RTA_PRIORITY])
2414		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2415
2416	if (tb[RTA_METRICS]) {
2417		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2418		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2419	}
2420
2421	if (tb[RTA_TABLE])
2422		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2423
2424	if (tb[RTA_MULTIPATH]) {
2425		cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2426		cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2427	}
2428
2429	err = 0;
2430errout:
2431	return err;
2432}
2433
2434static int ip6_route_multipath(struct fib6_config *cfg, int add)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2435{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2436	struct fib6_config r_cfg;
2437	struct rtnexthop *rtnh;
 
 
 
 
2438	int remaining;
2439	int attrlen;
2440	int err = 0, last_err = 0;
 
 
 
 
 
 
 
 
2441
2442beginning:
2443	rtnh = (struct rtnexthop *)cfg->fc_mp;
2444	remaining = cfg->fc_mp_len;
 
2445
2446	/* Parse a Multipath Entry */
 
 
2447	while (rtnh_ok(rtnh, remaining)) {
2448		memcpy(&r_cfg, cfg, sizeof(*cfg));
2449		if (rtnh->rtnh_ifindex)
2450			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2451
2452		attrlen = rtnh_attrlen(rtnh);
2453		if (attrlen > 0) {
2454			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2455
2456			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2457			if (nla) {
2458				nla_memcpy(&r_cfg.fc_gateway, nla, 16);
 
 
 
 
2459				r_cfg.fc_flags |= RTF_GATEWAY;
2460			}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2461		}
2462		err = add ? ip6_route_add(&r_cfg) : ip6_route_del(&r_cfg);
 
 
 
 
2463		if (err) {
2464			last_err = err;
2465			/* If we are trying to remove a route, do not stop the
2466			 * loop when ip6_route_del() fails (because next hop is
2467			 * already gone), we should try to remove all next hops.
2468			 */
2469			if (add) {
2470				/* If add fails, we should try to delete all
2471				 * next hops that have been already added.
2472				 */
2473				add = 0;
2474				goto beginning;
2475			}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2476		}
 
 
 
 
 
 
 
2477		/* Because each route is added like a single route we remove
2478		 * this flag after the first nexthop (if there is a collision,
2479		 * we have already fail to add the first nexthop:
2480		 * fib6_add_rt2node() has reject it).
 
 
2481		 */
2482		cfg->fc_nlinfo.nlh->nlmsg_flags &= ~NLM_F_EXCL;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2483		rtnh = rtnh_next(rtnh, &remaining);
2484	}
2485
2486	return last_err;
2487}
2488
2489static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh)
 
2490{
2491	struct fib6_config cfg;
2492	int err;
2493
2494	err = rtm_to_fib6_config(skb, nlh, &cfg);
2495	if (err < 0)
2496		return err;
2497
 
 
 
 
 
 
2498	if (cfg.fc_mp)
2499		return ip6_route_multipath(&cfg, 0);
2500	else
2501		return ip6_route_del(&cfg);
 
 
2502}
2503
2504static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh)
 
2505{
2506	struct fib6_config cfg;
2507	int err;
2508
2509	err = rtm_to_fib6_config(skb, nlh, &cfg);
2510	if (err < 0)
2511		return err;
2512
 
 
 
2513	if (cfg.fc_mp)
2514		return ip6_route_multipath(&cfg, 1);
2515	else
2516		return ip6_route_add(&cfg);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2517}
2518
2519static inline size_t rt6_nlmsg_size(void)
2520{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2521	return NLMSG_ALIGN(sizeof(struct rtmsg))
2522	       + nla_total_size(16) /* RTA_SRC */
2523	       + nla_total_size(16) /* RTA_DST */
2524	       + nla_total_size(16) /* RTA_GATEWAY */
2525	       + nla_total_size(16) /* RTA_PREFSRC */
2526	       + nla_total_size(4) /* RTA_TABLE */
2527	       + nla_total_size(4) /* RTA_IIF */
2528	       + nla_total_size(4) /* RTA_OIF */
2529	       + nla_total_size(4) /* RTA_PRIORITY */
2530	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2531	       + nla_total_size(sizeof(struct rta_cacheinfo));
 
 
 
2532}
2533
2534static int rt6_fill_node(struct net *net,
2535			 struct sk_buff *skb, struct rt6_info *rt,
2536			 struct in6_addr *dst, struct in6_addr *src,
2537			 int iif, int type, u32 portid, u32 seq,
2538			 int prefix, int nowait, unsigned int flags)
2539{
2540	struct rtmsg *rtm;
2541	struct nlmsghdr *nlh;
2542	long expires;
2543	u32 table;
2544
2545	if (prefix) {	/* user wants prefix routes only */
2546		if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2547			/* success since this is not a prefix route */
2548			return 1;
2549		}
 
 
 
 
 
 
 
 
 
 
2550	}
2551
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2552	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
2553	if (!nlh)
2554		return -EMSGSIZE;
2555
 
 
 
 
 
 
 
 
 
 
2556	rtm = nlmsg_data(nlh);
2557	rtm->rtm_family = AF_INET6;
2558	rtm->rtm_dst_len = rt->rt6i_dst.plen;
2559	rtm->rtm_src_len = rt->rt6i_src.plen;
2560	rtm->rtm_tos = 0;
2561	if (rt->rt6i_table)
2562		table = rt->rt6i_table->tb6_id;
2563	else
2564		table = RT6_TABLE_UNSPEC;
2565	rtm->rtm_table = table;
2566	if (nla_put_u32(skb, RTA_TABLE, table))
2567		goto nla_put_failure;
2568	if (rt->rt6i_flags & RTF_REJECT) {
2569		switch (rt->dst.error) {
2570		case -EINVAL:
2571			rtm->rtm_type = RTN_BLACKHOLE;
2572			break;
2573		case -EACCES:
2574			rtm->rtm_type = RTN_PROHIBIT;
2575			break;
2576		case -EAGAIN:
2577			rtm->rtm_type = RTN_THROW;
2578			break;
2579		default:
2580			rtm->rtm_type = RTN_UNREACHABLE;
2581			break;
2582		}
2583	}
2584	else if (rt->rt6i_flags & RTF_LOCAL)
2585		rtm->rtm_type = RTN_LOCAL;
2586	else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2587		rtm->rtm_type = RTN_LOCAL;
2588	else
2589		rtm->rtm_type = RTN_UNICAST;
2590	rtm->rtm_flags = 0;
2591	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2592	rtm->rtm_protocol = rt->rt6i_protocol;
2593	if (rt->rt6i_flags & RTF_DYNAMIC)
2594		rtm->rtm_protocol = RTPROT_REDIRECT;
2595	else if (rt->rt6i_flags & RTF_ADDRCONF) {
2596		if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
2597			rtm->rtm_protocol = RTPROT_RA;
2598		else
2599			rtm->rtm_protocol = RTPROT_KERNEL;
2600	}
2601
2602	if (rt->rt6i_flags & RTF_CACHE)
2603		rtm->rtm_flags |= RTM_F_CLONED;
2604
2605	if (dst) {
2606		if (nla_put(skb, RTA_DST, 16, dst))
2607			goto nla_put_failure;
2608		rtm->rtm_dst_len = 128;
2609	} else if (rtm->rtm_dst_len)
2610		if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2611			goto nla_put_failure;
2612#ifdef CONFIG_IPV6_SUBTREES
2613	if (src) {
2614		if (nla_put(skb, RTA_SRC, 16, src))
2615			goto nla_put_failure;
2616		rtm->rtm_src_len = 128;
2617	} else if (rtm->rtm_src_len &&
2618		   nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2619		goto nla_put_failure;
2620#endif
2621	if (iif) {
2622#ifdef CONFIG_IPV6_MROUTE
2623		if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2624			int err = ip6mr_get_route(net, skb, rtm, nowait);
2625			if (err <= 0) {
2626				if (!nowait) {
2627					if (err == 0)
2628						return 0;
2629					goto nla_put_failure;
2630				} else {
2631					if (err == -EMSGSIZE)
2632						goto nla_put_failure;
2633				}
2634			}
2635		} else
2636#endif
2637			if (nla_put_u32(skb, RTA_IIF, iif))
2638				goto nla_put_failure;
2639	} else if (dst) {
2640		struct in6_addr saddr_buf;
2641		if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2642		    nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2643			goto nla_put_failure;
2644	}
2645
2646	if (rt->rt6i_prefsrc.plen) {
2647		struct in6_addr saddr_buf;
2648		saddr_buf = rt->rt6i_prefsrc.addr;
2649		if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2650			goto nla_put_failure;
2651	}
2652
2653	if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
 
2654		goto nla_put_failure;
2655
2656	if (rt->rt6i_flags & RTF_GATEWAY) {
2657		if (nla_put(skb, RTA_GATEWAY, 16, &rt->rt6i_gateway) < 0)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2658			goto nla_put_failure;
 
 
2659	}
2660
2661	if (rt->dst.dev &&
2662	    nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2663		goto nla_put_failure;
2664	if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2665		goto nla_put_failure;
 
 
 
 
 
 
 
 
2666
2667	expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
 
2668
2669	if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
2670		goto nla_put_failure;
2671
2672	return nlmsg_end(skb, nlh);
 
 
2673
2674nla_put_failure:
2675	nlmsg_cancel(skb, nlh);
2676	return -EMSGSIZE;
2677}
2678
2679int rt6_dump_route(struct rt6_info *rt, void *p_arg)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2680{
2681	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2682	int prefix;
 
 
 
2683
2684	if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2685		struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2686		prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2687	} else
2688		prefix = 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2689
2690	return rt6_fill_node(arg->net,
2691		     arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2692		     NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
2693		     prefix, 0, NLM_F_MULTI);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2694}
2695
2696static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh)
 
2697{
2698	struct net *net = sock_net(in_skb->sk);
2699	struct nlattr *tb[RTA_MAX+1];
 
 
 
2700	struct rt6_info *rt;
2701	struct sk_buff *skb;
2702	struct rtmsg *rtm;
2703	struct flowi6 fl6;
2704	int err, iif = 0, oif = 0;
2705
2706	err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2707	if (err < 0)
2708		goto errout;
2709
2710	err = -EINVAL;
2711	memset(&fl6, 0, sizeof(fl6));
 
 
2712
2713	if (tb[RTA_SRC]) {
2714		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2715			goto errout;
2716
2717		fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2718	}
2719
2720	if (tb[RTA_DST]) {
2721		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2722			goto errout;
2723
2724		fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2725	}
2726
2727	if (tb[RTA_IIF])
2728		iif = nla_get_u32(tb[RTA_IIF]);
2729
2730	if (tb[RTA_OIF])
2731		oif = nla_get_u32(tb[RTA_OIF]);
2732
2733	if (tb[RTA_MARK])
2734		fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
2735
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2736	if (iif) {
2737		struct net_device *dev;
2738		int flags = 0;
2739
2740		dev = __dev_get_by_index(net, iif);
 
 
2741		if (!dev) {
 
2742			err = -ENODEV;
2743			goto errout;
2744		}
2745
2746		fl6.flowi6_iif = iif;
2747
2748		if (!ipv6_addr_any(&fl6.saddr))
2749			flags |= RT6_LOOKUP_F_HAS_SADDR;
2750
2751		rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2752							       flags);
 
2753	} else {
2754		fl6.flowi6_oif = oif;
2755
2756		rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2757	}
2758
2759	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2760	if (!skb) {
2761		ip6_rt_put(rt);
2762		err = -ENOBUFS;
2763		goto errout;
2764	}
2765
2766	/* Reserve room for dummy headers, this skb can pass
2767	   through good chunk of routing engine.
2768	 */
2769	skb_reset_mac_header(skb);
2770	skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2771
2772	skb_dst_set(skb, &rt->dst);
2773
2774	err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2775			    RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
2776			    nlh->nlmsg_seq, 0, 0, 0);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2777	if (err < 0) {
2778		kfree_skb(skb);
2779		goto errout;
2780	}
2781
2782	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2783errout:
2784	return err;
2785}
2786
2787void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
 
2788{
2789	struct sk_buff *skb;
2790	struct net *net = info->nl_net;
2791	u32 seq;
2792	int err;
2793
2794	err = -ENOBUFS;
2795	seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2796
2797	skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2798	if (!skb)
2799		goto errout;
2800
2801	err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2802				event, info->portid, seq, 0, 0, 0);
2803	if (err < 0) {
2804		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2805		WARN_ON(err == -EMSGSIZE);
2806		kfree_skb(skb);
2807		goto errout;
2808	}
2809	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2810		    info->nlh, gfp_any());
2811	return;
2812errout:
2813	if (err < 0)
2814		rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2815}
 
2816
2817static int ip6_route_dev_notify(struct notifier_block *this,
2818				unsigned long event, void *ptr)
2819{
2820	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2821	struct net *net = dev_net(dev);
2822
2823	if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
 
 
 
 
2824		net->ipv6.ip6_null_entry->dst.dev = dev;
2825		net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2826#ifdef CONFIG_IPV6_MULTIPLE_TABLES
2827		net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2828		net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2829		net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2830		net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2831#endif
 
 
 
 
 
 
 
 
 
 
2832	}
2833
2834	return NOTIFY_OK;
2835}
2836
2837/*
2838 *	/proc
2839 */
2840
2841#ifdef CONFIG_PROC_FS
2842
2843static const struct file_operations ipv6_route_proc_fops = {
2844	.owner		= THIS_MODULE,
2845	.open		= ipv6_route_open,
2846	.read		= seq_read,
2847	.llseek		= seq_lseek,
2848	.release	= seq_release_net,
2849};
2850
2851static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2852{
2853	struct net *net = (struct net *)seq->private;
2854	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2855		   net->ipv6.rt6_stats->fib_nodes,
2856		   net->ipv6.rt6_stats->fib_route_nodes,
2857		   net->ipv6.rt6_stats->fib_rt_alloc,
2858		   net->ipv6.rt6_stats->fib_rt_entries,
2859		   net->ipv6.rt6_stats->fib_rt_cache,
2860		   dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2861		   net->ipv6.rt6_stats->fib_discarded_routes);
2862
2863	return 0;
2864}
2865
2866static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2867{
2868	return single_open_net(inode, file, rt6_stats_seq_show);
2869}
2870
2871static const struct file_operations rt6_stats_seq_fops = {
2872	.owner	 = THIS_MODULE,
2873	.open	 = rt6_stats_seq_open,
2874	.read	 = seq_read,
2875	.llseek	 = seq_lseek,
2876	.release = single_release_net,
2877};
2878#endif	/* CONFIG_PROC_FS */
2879
2880#ifdef CONFIG_SYSCTL
2881
2882static
2883int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
2884			      void __user *buffer, size_t *lenp, loff_t *ppos)
2885{
2886	struct net *net;
2887	int delay;
 
2888	if (!write)
2889		return -EINVAL;
2890
 
 
 
 
2891	net = (struct net *)ctl->extra1;
2892	delay = net->ipv6.sysctl.flush_delay;
2893	proc_dointvec(ctl, write, buffer, lenp, ppos);
2894	fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
2895	return 0;
2896}
2897
2898struct ctl_table ipv6_route_table_template[] = {
2899	{
2900		.procname	=	"flush",
2901		.data		=	&init_net.ipv6.sysctl.flush_delay,
2902		.maxlen		=	sizeof(int),
2903		.mode		=	0200,
2904		.proc_handler	=	ipv6_sysctl_rtcache_flush
2905	},
2906	{
2907		.procname	=	"gc_thresh",
2908		.data		=	&ip6_dst_ops_template.gc_thresh,
2909		.maxlen		=	sizeof(int),
2910		.mode		=	0644,
2911		.proc_handler	=	proc_dointvec,
2912	},
2913	{
2914		.procname	=	"max_size",
2915		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
2916		.maxlen		=	sizeof(int),
2917		.mode		=	0644,
2918		.proc_handler	=	proc_dointvec,
2919	},
2920	{
2921		.procname	=	"gc_min_interval",
2922		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2923		.maxlen		=	sizeof(int),
2924		.mode		=	0644,
2925		.proc_handler	=	proc_dointvec_jiffies,
2926	},
2927	{
2928		.procname	=	"gc_timeout",
2929		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2930		.maxlen		=	sizeof(int),
2931		.mode		=	0644,
2932		.proc_handler	=	proc_dointvec_jiffies,
2933	},
2934	{
2935		.procname	=	"gc_interval",
2936		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
2937		.maxlen		=	sizeof(int),
2938		.mode		=	0644,
2939		.proc_handler	=	proc_dointvec_jiffies,
2940	},
2941	{
2942		.procname	=	"gc_elasticity",
2943		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2944		.maxlen		=	sizeof(int),
2945		.mode		=	0644,
2946		.proc_handler	=	proc_dointvec,
2947	},
2948	{
2949		.procname	=	"mtu_expires",
2950		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2951		.maxlen		=	sizeof(int),
2952		.mode		=	0644,
2953		.proc_handler	=	proc_dointvec_jiffies,
2954	},
2955	{
2956		.procname	=	"min_adv_mss",
2957		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
2958		.maxlen		=	sizeof(int),
2959		.mode		=	0644,
2960		.proc_handler	=	proc_dointvec,
2961	},
2962	{
2963		.procname	=	"gc_min_interval_ms",
2964		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2965		.maxlen		=	sizeof(int),
2966		.mode		=	0644,
2967		.proc_handler	=	proc_dointvec_ms_jiffies,
2968	},
2969	{ }
 
 
 
 
 
 
 
 
2970};
2971
2972struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2973{
2974	struct ctl_table *table;
2975
2976	table = kmemdup(ipv6_route_table_template,
2977			sizeof(ipv6_route_table_template),
2978			GFP_KERNEL);
2979
2980	if (table) {
2981		table[0].data = &net->ipv6.sysctl.flush_delay;
2982		table[0].extra1 = net;
2983		table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2984		table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
 
2985		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2986		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2987		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2988		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2989		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2990		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2991		table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2992
2993		/* Don't export sysctls to unprivileged users */
2994		if (net->user_ns != &init_user_ns)
2995			table[0].procname = NULL;
2996	}
2997
2998	return table;
2999}
 
 
 
 
 
 
 
 
 
3000#endif
3001
3002static int __net_init ip6_route_net_init(struct net *net)
3003{
3004	int ret = -ENOMEM;
3005
3006	memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3007	       sizeof(net->ipv6.ip6_dst_ops));
3008
3009	if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3010		goto out_ip6_dst_ops;
3011
 
 
 
 
 
 
3012	net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3013					   sizeof(*net->ipv6.ip6_null_entry),
3014					   GFP_KERNEL);
3015	if (!net->ipv6.ip6_null_entry)
3016		goto out_ip6_dst_entries;
3017	net->ipv6.ip6_null_entry->dst.path =
3018		(struct dst_entry *)net->ipv6.ip6_null_entry;
3019	net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3020	dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3021			 ip6_template_metrics, true);
 
3022
3023#ifdef CONFIG_IPV6_MULTIPLE_TABLES
 
3024	net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3025					       sizeof(*net->ipv6.ip6_prohibit_entry),
3026					       GFP_KERNEL);
3027	if (!net->ipv6.ip6_prohibit_entry)
3028		goto out_ip6_null_entry;
3029	net->ipv6.ip6_prohibit_entry->dst.path =
3030		(struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3031	net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3032	dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3033			 ip6_template_metrics, true);
 
3034
3035	net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3036					       sizeof(*net->ipv6.ip6_blk_hole_entry),
3037					       GFP_KERNEL);
3038	if (!net->ipv6.ip6_blk_hole_entry)
3039		goto out_ip6_prohibit_entry;
3040	net->ipv6.ip6_blk_hole_entry->dst.path =
3041		(struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3042	net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3043	dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3044			 ip6_template_metrics, true);
 
 
 
 
3045#endif
3046
3047	net->ipv6.sysctl.flush_delay = 0;
3048	net->ipv6.sysctl.ip6_rt_max_size = 4096;
3049	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3050	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3051	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3052	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3053	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3054	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
 
3055
3056	net->ipv6.ip6_rt_gc_expire = 30*HZ;
3057
3058	ret = 0;
3059out:
3060	return ret;
3061
3062#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3063out_ip6_prohibit_entry:
3064	kfree(net->ipv6.ip6_prohibit_entry);
3065out_ip6_null_entry:
3066	kfree(net->ipv6.ip6_null_entry);
3067#endif
 
 
3068out_ip6_dst_entries:
3069	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3070out_ip6_dst_ops:
3071	goto out;
3072}
3073
3074static void __net_exit ip6_route_net_exit(struct net *net)
3075{
 
3076	kfree(net->ipv6.ip6_null_entry);
3077#ifdef CONFIG_IPV6_MULTIPLE_TABLES
3078	kfree(net->ipv6.ip6_prohibit_entry);
3079	kfree(net->ipv6.ip6_blk_hole_entry);
3080#endif
3081	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3082}
3083
3084static int __net_init ip6_route_net_init_late(struct net *net)
3085{
3086#ifdef CONFIG_PROC_FS
3087	proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3088	proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
 
 
 
 
 
 
 
 
3089#endif
3090	return 0;
3091}
3092
3093static void __net_exit ip6_route_net_exit_late(struct net *net)
3094{
3095#ifdef CONFIG_PROC_FS
3096	remove_proc_entry("ipv6_route", net->proc_net);
3097	remove_proc_entry("rt6_stats", net->proc_net);
3098#endif
3099}
3100
3101static struct pernet_operations ip6_route_net_ops = {
3102	.init = ip6_route_net_init,
3103	.exit = ip6_route_net_exit,
3104};
3105
3106static int __net_init ipv6_inetpeer_init(struct net *net)
3107{
3108	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3109
3110	if (!bp)
3111		return -ENOMEM;
3112	inet_peer_base_init(bp);
3113	net->ipv6.peers = bp;
3114	return 0;
3115}
3116
3117static void __net_exit ipv6_inetpeer_exit(struct net *net)
3118{
3119	struct inet_peer_base *bp = net->ipv6.peers;
3120
3121	net->ipv6.peers = NULL;
3122	inetpeer_invalidate_tree(bp);
3123	kfree(bp);
3124}
3125
3126static struct pernet_operations ipv6_inetpeer_ops = {
3127	.init	=	ipv6_inetpeer_init,
3128	.exit	=	ipv6_inetpeer_exit,
3129};
3130
3131static struct pernet_operations ip6_route_net_late_ops = {
3132	.init = ip6_route_net_init_late,
3133	.exit = ip6_route_net_exit_late,
3134};
3135
3136static struct notifier_block ip6_route_dev_notifier = {
3137	.notifier_call = ip6_route_dev_notify,
3138	.priority = 0,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3139};
3140
3141int __init ip6_route_init(void)
3142{
3143	int ret;
 
3144
3145	ret = -ENOMEM;
3146	ip6_dst_ops_template.kmem_cachep =
3147		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3148				  SLAB_HWCACHE_ALIGN, NULL);
3149	if (!ip6_dst_ops_template.kmem_cachep)
3150		goto out;
3151
3152	ret = dst_entries_init(&ip6_dst_blackhole_ops);
3153	if (ret)
3154		goto out_kmem_cache;
3155
3156	ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3157	if (ret)
3158		goto out_dst_entries;
3159
3160	ret = register_pernet_subsys(&ip6_route_net_ops);
3161	if (ret)
3162		goto out_register_inetpeer;
3163
3164	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3165
3166	/* Registering of the loopback is done before this portion of code,
3167	 * the loopback reference in rt6_info will not be taken, do it
3168	 * manually for init_net */
3169	init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3170	init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3171  #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3172	init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3173	init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3174	init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3175	init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3176  #endif
3177	ret = fib6_init();
3178	if (ret)
3179		goto out_register_subsys;
3180
3181	ret = xfrm6_init();
3182	if (ret)
3183		goto out_fib6_init;
3184
3185	ret = fib6_rules_init();
3186	if (ret)
3187		goto xfrm6_init;
3188
3189	ret = register_pernet_subsys(&ip6_route_net_late_ops);
3190	if (ret)
3191		goto fib6_rules_init;
3192
3193	ret = -ENOBUFS;
3194	if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3195	    __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3196	    __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3197		goto out_register_late_subsys;
3198
3199	ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3200	if (ret)
3201		goto out_register_late_subsys;
3202
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3203out:
3204	return ret;
3205
3206out_register_late_subsys:
 
3207	unregister_pernet_subsys(&ip6_route_net_late_ops);
3208fib6_rules_init:
3209	fib6_rules_cleanup();
3210xfrm6_init:
3211	xfrm6_fini();
3212out_fib6_init:
3213	fib6_gc_cleanup();
3214out_register_subsys:
3215	unregister_pernet_subsys(&ip6_route_net_ops);
3216out_register_inetpeer:
3217	unregister_pernet_subsys(&ipv6_inetpeer_ops);
3218out_dst_entries:
3219	dst_entries_destroy(&ip6_dst_blackhole_ops);
3220out_kmem_cache:
3221	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3222	goto out;
3223}
3224
3225void ip6_route_cleanup(void)
3226{
 
 
 
 
 
3227	unregister_netdevice_notifier(&ip6_route_dev_notifier);
3228	unregister_pernet_subsys(&ip6_route_net_late_ops);
3229	fib6_rules_cleanup();
3230	xfrm6_fini();
3231	fib6_gc_cleanup();
3232	unregister_pernet_subsys(&ipv6_inetpeer_ops);
3233	unregister_pernet_subsys(&ip6_route_net_ops);
3234	dst_entries_destroy(&ip6_dst_blackhole_ops);
3235	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3236}
v6.13.7
   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 void		ip6_negative_advice(struct sock *sk,
  91					    struct dst_entry *dst);
  92static void		ip6_dst_destroy(struct dst_entry *);
  93static void		ip6_dst_ifdown(struct dst_entry *,
  94				       struct net_device *dev);
  95static void		 ip6_dst_gc(struct dst_ops *ops);
  96
  97static int		ip6_pkt_discard(struct sk_buff *skb);
  98static int		ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  99static int		ip6_pkt_prohibit(struct sk_buff *skb);
 100static int		ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
 101static void		ip6_link_failure(struct sk_buff *skb);
 102static void		ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
 103					   struct sk_buff *skb, u32 mtu,
 104					   bool confirm_neigh);
 105static void		rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
 106					struct sk_buff *skb);
 107static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
 108			   int strict);
 109static size_t rt6_nlmsg_size(struct fib6_info *f6i);
 110static int rt6_fill_node(struct net *net, struct sk_buff *skb,
 111			 struct fib6_info *rt, struct dst_entry *dst,
 112			 struct in6_addr *dest, struct in6_addr *src,
 113			 int iif, int type, u32 portid, u32 seq,
 114			 unsigned int flags);
 115static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
 116					   const struct in6_addr *daddr,
 117					   const struct in6_addr *saddr);
 118
 119#ifdef CONFIG_IPV6_ROUTE_INFO
 120static struct fib6_info *rt6_add_route_info(struct net *net,
 121					   const struct in6_addr *prefix, int prefixlen,
 122					   const struct in6_addr *gwaddr,
 123					   struct net_device *dev,
 124					   unsigned int pref);
 125static struct fib6_info *rt6_get_route_info(struct net *net,
 126					   const struct in6_addr *prefix, int prefixlen,
 127					   const struct in6_addr *gwaddr,
 128					   struct net_device *dev);
 129#endif
 130
 131struct uncached_list {
 132	spinlock_t		lock;
 133	struct list_head	head;
 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 && 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_del_init(&rt->dst.rt_uncached);
 192		}
 193		spin_unlock_bh(&ul->lock);
 194	}
 
 195}
 196
 197static inline const void *choose_neigh_daddr(const struct in6_addr *p,
 198					     struct sk_buff *skb,
 199					     const void *daddr)
 200{
 
 
 201	if (!ipv6_addr_any(p))
 202		return (const void *) p;
 203	else if (skb)
 204		return &ipv6_hdr(skb)->daddr;
 205	return daddr;
 206}
 207
 208struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
 209				   struct net_device *dev,
 210				   struct sk_buff *skb,
 211				   const void *daddr)
 212{
 
 213	struct neighbour *n;
 214
 215	daddr = choose_neigh_daddr(gw, skb, daddr);
 216	n = __ipv6_neigh_lookup(dev, daddr);
 217	if (n)
 218		return n;
 219
 220	n = neigh_create(&nd_tbl, daddr, dev);
 221	return IS_ERR(n) ? NULL : n;
 222}
 223
 224static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
 225					      struct sk_buff *skb,
 226					      const void *daddr)
 227{
 228	const struct rt6_info *rt = dst_rt6_info(dst);
 229
 230	return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
 231				dst->dev, skb, daddr);
 232}
 233
 234static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
 235{
 236	const struct rt6_info *rt = dst_rt6_info(dst);
 237	struct net_device *dev = dst->dev;
 238
 239	daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr);
 240	if (!daddr)
 241		return;
 242	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 243		return;
 244	if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
 245		return;
 246	__ipv6_confirm_neigh(dev, daddr);
 247}
 248
 249static struct dst_ops ip6_dst_ops_template = {
 250	.family			=	AF_INET6,
 
 251	.gc			=	ip6_dst_gc,
 252	.gc_thresh		=	1024,
 253	.check			=	ip6_dst_check,
 254	.default_advmss		=	ip6_default_advmss,
 255	.mtu			=	ip6_mtu,
 256	.cow_metrics		=	dst_cow_metrics_generic,
 257	.destroy		=	ip6_dst_destroy,
 258	.ifdown			=	ip6_dst_ifdown,
 259	.negative_advice	=	ip6_negative_advice,
 260	.link_failure		=	ip6_link_failure,
 261	.update_pmtu		=	ip6_rt_update_pmtu,
 262	.redirect		=	rt6_do_redirect,
 263	.local_out		=	__ip6_local_out,
 264	.neigh_lookup		=	ip6_dst_neigh_lookup,
 265	.confirm_neigh		=	ip6_confirm_neigh,
 266};
 267
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 268static struct dst_ops ip6_dst_blackhole_ops = {
 269	.family			= AF_INET6,
 270	.default_advmss		= ip6_default_advmss,
 271	.neigh_lookup		= ip6_dst_neigh_lookup,
 272	.check			= ip6_dst_check,
 273	.destroy		= ip6_dst_destroy,
 274	.cow_metrics		= dst_cow_metrics_generic,
 275	.update_pmtu		= dst_blackhole_update_pmtu,
 276	.redirect		= dst_blackhole_redirect,
 277	.mtu			= dst_blackhole_mtu,
 
 278};
 279
 280static const u32 ip6_template_metrics[RTAX_MAX] = {
 281	[RTAX_HOPLIMIT - 1] = 0,
 282};
 283
 284static const struct fib6_info fib6_null_entry_template = {
 285	.fib6_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 286	.fib6_protocol  = RTPROT_KERNEL,
 287	.fib6_metric	= ~(u32)0,
 288	.fib6_ref	= REFCOUNT_INIT(1),
 289	.fib6_type	= RTN_UNREACHABLE,
 290	.fib6_metrics	= (struct dst_metrics *)&dst_default_metrics,
 291};
 292
 293static const struct rt6_info ip6_null_entry_template = {
 294	.dst = {
 295		.__rcuref	= RCUREF_INIT(1),
 296		.__use		= 1,
 297		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 298		.error		= -ENETUNREACH,
 299		.input		= ip6_pkt_discard,
 300		.output		= ip6_pkt_discard_out,
 301	},
 302	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 
 
 
 303};
 304
 305#ifdef CONFIG_IPV6_MULTIPLE_TABLES
 306
 307static const struct rt6_info ip6_prohibit_entry_template = {
 308	.dst = {
 309		.__rcuref	= RCUREF_INIT(1),
 310		.__use		= 1,
 311		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 312		.error		= -EACCES,
 313		.input		= ip6_pkt_prohibit,
 314		.output		= ip6_pkt_prohibit_out,
 315	},
 316	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 
 
 
 317};
 318
 319static const struct rt6_info ip6_blk_hole_entry_template = {
 320	.dst = {
 321		.__rcuref	= RCUREF_INIT(1),
 322		.__use		= 1,
 323		.obsolete	= DST_OBSOLETE_FORCE_CHK,
 324		.error		= -EINVAL,
 325		.input		= dst_discard,
 326		.output		= dst_discard_out,
 327	},
 328	.rt6i_flags	= (RTF_REJECT | RTF_NONEXTHOP),
 
 
 
 329};
 330
 331#endif
 332
 333static void rt6_info_init(struct rt6_info *rt)
 334{
 335	memset_after(rt, 0, dst);
 336}
 337
 338/* allocate dst with ip6_dst_ops */
 339struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
 340			       int flags)
 
 
 341{
 342	struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 343					DST_OBSOLETE_FORCE_CHK, flags);
 344
 345	if (rt) {
 346		rt6_info_init(rt);
 347		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
 
 
 
 
 348	}
 349
 350	return rt;
 351}
 352EXPORT_SYMBOL(ip6_dst_alloc);
 353
 354static void ip6_dst_destroy(struct dst_entry *dst)
 355{
 356	struct rt6_info *rt = dst_rt6_info(dst);
 357	struct fib6_info *from;
 358	struct inet6_dev *idev;
 359
 360	ip_dst_metrics_put(dst);
 361	rt6_uncached_list_del(rt);
 362
 363	idev = rt->rt6i_idev;
 364	if (idev) {
 365		rt->rt6i_idev = NULL;
 366		in6_dev_put(idev);
 367	}
 368
 369	from = unrcu_pointer(xchg(&rt->from, NULL));
 370	fib6_info_release(from);
 
 
 
 
 
 371}
 372
 373static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
 
 374{
 375	struct rt6_info *rt = dst_rt6_info(dst);
 376	struct inet6_dev *idev = rt->rt6i_idev;
 377	struct fib6_info *from;
 
 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	from = unrcu_pointer(xchg(&rt->from, NULL));
 388	fib6_info_release(from);
 389}
 390
 391static bool __rt6_check_expired(const struct rt6_info *rt)
 392{
 393	if (rt->rt6i_flags & RTF_EXPIRES)
 394		return time_after(jiffies, rt->dst.expires);
 395	else
 396		return false;
 397}
 398
 399static bool rt6_check_expired(const struct rt6_info *rt)
 400{
 401	struct fib6_info *from;
 402
 403	from = rcu_dereference(rt->from);
 404
 405	if (rt->rt6i_flags & RTF_EXPIRES) {
 406		if (time_after(jiffies, rt->dst.expires))
 407			return true;
 408	} else if (from) {
 409		return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
 410			fib6_check_expired(from);
 411	}
 412	return false;
 413}
 414
 415void fib6_select_path(const struct net *net, struct fib6_result *res,
 416		      struct flowi6 *fl6, int oif, bool have_oif_match,
 417		      const struct sk_buff *skb, int strict)
 
 
 
 418{
 419	struct fib6_info *match = res->f6i;
 420	struct fib6_info *sibling;
 421
 422	if (!match->nh && (!match->fib6_nsiblings || have_oif_match))
 423		goto out;
 424
 425	if (match->nh && have_oif_match && res->nh)
 426		return;
 427
 428	if (skb)
 429		IP6CB(skb)->flags |= IP6SKB_MULTIPATH;
 430
 431	/* We might have already computed the hash for ICMPv6 errors. In such
 432	 * case it will always be non-zero. Otherwise now is the time to do it.
 433	 */
 434	if (!fl6->mp_hash &&
 435	    (!match->nh || nexthop_is_multipath(match->nh)))
 436		fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
 437
 438	if (unlikely(match->nh)) {
 439		nexthop_path_fib6_result(res, fl6->mp_hash);
 440		return;
 441	}
 442
 443	if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
 444		goto out;
 445
 446	list_for_each_entry_rcu(sibling, &match->fib6_siblings,
 447				fib6_siblings) {
 448		const struct fib6_nh *nh = sibling->fib6_nh;
 449		int nh_upper_bound;
 450
 451		nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
 452		if (fl6->mp_hash > nh_upper_bound)
 453			continue;
 454		if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
 455			break;
 456		match = sibling;
 457		break;
 458	}
 
 
 459
 460out:
 461	res->f6i = match;
 462	res->nh = match->fib6_nh;
 463}
 464
 465/*
 466 *	Route lookup. rcu_read_lock() should be held.
 467 */
 468
 469static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
 470			       const struct in6_addr *saddr, int oif, int flags)
 471{
 472	const struct net_device *dev;
 
 473
 474	if (nh->fib_nh_flags & RTNH_F_DEAD)
 475		return false;
 476
 477	dev = nh->fib_nh_dev;
 478	if (oif) {
 479		if (dev->ifindex == oif)
 480			return true;
 481	} else {
 482		if (ipv6_chk_addr(net, saddr, dev,
 483				  flags & RT6_LOOKUP_F_IFACE))
 484			return true;
 485	}
 486
 487	return false;
 
 
 488}
 489
 490struct fib6_nh_dm_arg {
 491	struct net		*net;
 492	const struct in6_addr	*saddr;
 493	int			oif;
 494	int			flags;
 495	struct fib6_nh		*nh;
 496};
 497
 498static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
 499{
 500	struct fib6_nh_dm_arg *arg = _arg;
 501
 502	arg->nh = nh;
 503	return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
 504				  arg->flags);
 505}
 506
 507/* returns fib6_nh from nexthop or NULL */
 508static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
 509					struct fib6_result *res,
 510					const struct in6_addr *saddr,
 511					int oif, int flags)
 512{
 513	struct fib6_nh_dm_arg arg = {
 514		.net   = net,
 515		.saddr = saddr,
 516		.oif   = oif,
 517		.flags = flags,
 518	};
 519
 520	if (nexthop_is_blackhole(nh))
 521		return NULL;
 522
 523	if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
 524		return arg.nh;
 525
 526	return NULL;
 527}
 528
 529static void rt6_device_match(struct net *net, struct fib6_result *res,
 530			     const struct in6_addr *saddr, int oif, int flags)
 531{
 532	struct fib6_info *f6i = res->f6i;
 533	struct fib6_info *spf6i;
 534	struct fib6_nh *nh;
 535
 536	if (!oif && ipv6_addr_any(saddr)) {
 537		if (unlikely(f6i->nh)) {
 538			nh = nexthop_fib6_nh(f6i->nh);
 539			if (nexthop_is_blackhole(f6i->nh))
 540				goto out_blackhole;
 541		} else {
 542			nh = f6i->fib6_nh;
 
 
 543		}
 544		if (!(nh->fib_nh_flags & RTNH_F_DEAD))
 545			goto out;
 546	}
 547
 548	for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
 549		bool matched = false;
 550
 551		if (unlikely(spf6i->nh)) {
 552			nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
 553					      oif, flags);
 554			if (nh)
 555				matched = true;
 556		} else {
 557			nh = spf6i->fib6_nh;
 558			if (__rt6_device_match(net, nh, saddr, oif, flags))
 559				matched = true;
 560		}
 561		if (matched) {
 562			res->f6i = spf6i;
 563			goto out;
 564		}
 565	}
 566
 567	if (oif && flags & RT6_LOOKUP_F_IFACE) {
 568		res->f6i = net->ipv6.fib6_null_entry;
 569		nh = res->f6i->fib6_nh;
 570		goto out;
 571	}
 572
 573	if (unlikely(f6i->nh)) {
 574		nh = nexthop_fib6_nh(f6i->nh);
 575		if (nexthop_is_blackhole(f6i->nh))
 576			goto out_blackhole;
 577	} else {
 578		nh = f6i->fib6_nh;
 579	}
 580
 581	if (nh->fib_nh_flags & RTNH_F_DEAD) {
 582		res->f6i = net->ipv6.fib6_null_entry;
 583		nh = res->f6i->fib6_nh;
 584	}
 585out:
 586	res->nh = nh;
 587	res->fib6_type = res->f6i->fib6_type;
 588	res->fib6_flags = res->f6i->fib6_flags;
 589	return;
 590
 591out_blackhole:
 592	res->fib6_flags |= RTF_REJECT;
 593	res->fib6_type = RTN_BLACKHOLE;
 594	res->nh = nh;
 595}
 596
 597#ifdef CONFIG_IPV6_ROUTER_PREF
 598struct __rt6_probe_work {
 599	struct work_struct work;
 600	struct in6_addr target;
 601	struct net_device *dev;
 602	netdevice_tracker dev_tracker;
 603};
 604
 605static void rt6_probe_deferred(struct work_struct *w)
 606{
 607	struct in6_addr mcaddr;
 608	struct __rt6_probe_work *work =
 609		container_of(w, struct __rt6_probe_work, work);
 610
 611	addrconf_addr_solict_mult(&work->target, &mcaddr);
 612	ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
 613	netdev_put(work->dev, &work->dev_tracker);
 614	kfree(work);
 615}
 616
 617static void rt6_probe(struct fib6_nh *fib6_nh)
 618{
 619	struct __rt6_probe_work *work = NULL;
 620	const struct in6_addr *nh_gw;
 621	unsigned long last_probe;
 622	struct neighbour *neigh;
 623	struct net_device *dev;
 624	struct inet6_dev *idev;
 625
 626	/*
 627	 * Okay, this does not seem to be appropriate
 628	 * for now, however, we need to check if it
 629	 * is really so; aka Router Reachability Probing.
 630	 *
 631	 * Router Reachability Probe MUST be rate-limited
 632	 * to no more than one per minute.
 633	 */
 634	if (!fib6_nh->fib_nh_gw_family)
 635		return;
 636
 637	nh_gw = &fib6_nh->fib_nh_gw6;
 638	dev = fib6_nh->fib_nh_dev;
 639	rcu_read_lock();
 640	last_probe = READ_ONCE(fib6_nh->last_probe);
 641	idev = __in6_dev_get(dev);
 642	if (!idev)
 643		goto out;
 644	neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
 645	if (neigh) {
 646		if (READ_ONCE(neigh->nud_state) & NUD_VALID)
 
 647			goto out;
 
 
 
 
 
 648
 649		write_lock_bh(&neigh->lock);
 650		if (!(neigh->nud_state & NUD_VALID) &&
 651		    time_after(jiffies,
 652			       neigh->updated +
 653			       READ_ONCE(idev->cnf.rtr_probe_interval))) {
 654			work = kmalloc(sizeof(*work), GFP_ATOMIC);
 655			if (work)
 656				__neigh_set_probe_once(neigh);
 657		}
 658		write_unlock_bh(&neigh->lock);
 659	} else if (time_after(jiffies, last_probe +
 660				       READ_ONCE(idev->cnf.rtr_probe_interval))) {
 661		work = kmalloc(sizeof(*work), GFP_ATOMIC);
 662	}
 663
 664	if (!work || cmpxchg(&fib6_nh->last_probe,
 665			     last_probe, jiffies) != last_probe) {
 666		kfree(work);
 
 
 
 
 
 
 
 
 
 
 667	} else {
 668		INIT_WORK(&work->work, rt6_probe_deferred);
 669		work->target = *nh_gw;
 670		netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC);
 671		work->dev = dev;
 672		schedule_work(&work->work);
 673	}
 674
 675out:
 676	rcu_read_unlock();
 677}
 678#else
 679static inline void rt6_probe(struct fib6_nh *fib6_nh)
 680{
 681}
 682#endif
 683
 684/*
 685 * Default Router Selection (RFC 2461 6.3.6)
 686 */
 687static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
 
 
 
 
 
 
 
 
 
 
 
 688{
 
 689	enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 690	struct neighbour *neigh;
 691
 692	rcu_read_lock();
 693	neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
 694					  &fib6_nh->fib_nh_gw6);
 
 
 
 695	if (neigh) {
 696		u8 nud_state = READ_ONCE(neigh->nud_state);
 697
 698		if (nud_state & NUD_VALID)
 699			ret = RT6_NUD_SUCCEED;
 700#ifdef CONFIG_IPV6_ROUTER_PREF
 701		else if (!(nud_state & NUD_FAILED))
 702			ret = RT6_NUD_SUCCEED;
 703		else
 704			ret = RT6_NUD_FAIL_PROBE;
 705#endif
 
 706	} else {
 707		ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 708		      RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 709	}
 710	rcu_read_unlock();
 711
 712	return ret;
 713}
 714
 715static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
 716			   int strict)
 717{
 718	int m = 0;
 719
 720	if (!oif || nh->fib_nh_dev->ifindex == oif)
 721		m = 2;
 722
 
 723	if (!m && (strict & RT6_LOOKUP_F_IFACE))
 724		return RT6_NUD_FAIL_HARD;
 725#ifdef CONFIG_IPV6_ROUTER_PREF
 726	m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
 727#endif
 728	if ((strict & RT6_LOOKUP_F_REACHABLE) &&
 729	    !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
 730		int n = rt6_check_neigh(nh);
 731		if (n < 0)
 732			return n;
 733	}
 734	return m;
 735}
 736
 737static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
 738		       int oif, int strict, int *mpri, bool *do_rr)
 
 739{
 
 740	bool match_do_rr = false;
 741	bool rc = false;
 742	int m;
 743
 744	if (nh->fib_nh_flags & RTNH_F_DEAD)
 745		goto out;
 746
 747	if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
 748	    nh->fib_nh_flags & RTNH_F_LINKDOWN &&
 749	    !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
 750		goto out;
 751
 752	m = rt6_score_route(nh, fib6_flags, oif, strict);
 753	if (m == RT6_NUD_FAIL_DO_RR) {
 754		match_do_rr = true;
 755		m = 0; /* lowest valid score */
 756	} else if (m == RT6_NUD_FAIL_HARD) {
 757		goto out;
 758	}
 759
 760	if (strict & RT6_LOOKUP_F_REACHABLE)
 761		rt6_probe(nh);
 762
 763	/* note that m can be RT6_NUD_FAIL_PROBE at this point */
 764	if (m > *mpri) {
 765		*do_rr = match_do_rr;
 766		*mpri = m;
 767		rc = true;
 768	}
 769out:
 770	return rc;
 771}
 772
 773struct fib6_nh_frl_arg {
 774	u32		flags;
 775	int		oif;
 776	int		strict;
 777	int		*mpri;
 778	bool		*do_rr;
 779	struct fib6_nh	*nh;
 780};
 781
 782static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
 783{
 784	struct fib6_nh_frl_arg *arg = _arg;
 785
 786	arg->nh = nh;
 787	return find_match(nh, arg->flags, arg->oif, arg->strict,
 788			  arg->mpri, arg->do_rr);
 789}
 790
 791static void __find_rr_leaf(struct fib6_info *f6i_start,
 792			   struct fib6_info *nomatch, u32 metric,
 793			   struct fib6_result *res, struct fib6_info **cont,
 794			   int oif, int strict, bool *do_rr, int *mpri)
 795{
 796	struct fib6_info *f6i;
 797
 798	for (f6i = f6i_start;
 799	     f6i && f6i != nomatch;
 800	     f6i = rcu_dereference(f6i->fib6_next)) {
 801		bool matched = false;
 802		struct fib6_nh *nh;
 803
 804		if (cont && f6i->fib6_metric != metric) {
 805			*cont = f6i;
 806			return;
 807		}
 808
 809		if (fib6_check_expired(f6i))
 810			continue;
 811
 812		if (unlikely(f6i->nh)) {
 813			struct fib6_nh_frl_arg arg = {
 814				.flags  = f6i->fib6_flags,
 815				.oif    = oif,
 816				.strict = strict,
 817				.mpri   = mpri,
 818				.do_rr  = do_rr
 819			};
 820
 821			if (nexthop_is_blackhole(f6i->nh)) {
 822				res->fib6_flags = RTF_REJECT;
 823				res->fib6_type = RTN_BLACKHOLE;
 824				res->f6i = f6i;
 825				res->nh = nexthop_fib6_nh(f6i->nh);
 826				return;
 827			}
 828			if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
 829						     &arg)) {
 830				matched = true;
 831				nh = arg.nh;
 832			}
 833		} else {
 834			nh = f6i->fib6_nh;
 835			if (find_match(nh, f6i->fib6_flags, oif, strict,
 836				       mpri, do_rr))
 837				matched = true;
 838		}
 839		if (matched) {
 840			res->f6i = f6i;
 841			res->nh = nh;
 842			res->fib6_flags = f6i->fib6_flags;
 843			res->fib6_type = f6i->fib6_type;
 844		}
 845	}
 846}
 847
 848static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
 849			 struct fib6_info *rr_head, int oif, int strict,
 850			 bool *do_rr, struct fib6_result *res)
 
 851{
 852	u32 metric = rr_head->fib6_metric;
 853	struct fib6_info *cont = NULL;
 854	int mpri = -1;
 855
 856	__find_rr_leaf(rr_head, NULL, metric, res, &cont,
 857		       oif, strict, do_rr, &mpri);
 
 
 
 
 
 858
 859	__find_rr_leaf(leaf, rr_head, metric, res, &cont,
 860		       oif, strict, do_rr, &mpri);
 861
 862	if (res->f6i || !cont)
 863		return;
 864
 865	__find_rr_leaf(cont, NULL, metric, res, NULL,
 866		       oif, strict, do_rr, &mpri);
 867}
 868
 869static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
 870		       struct fib6_result *res, int strict)
 871{
 872	struct fib6_info *leaf = rcu_dereference(fn->leaf);
 873	struct fib6_info *rt0;
 874	bool do_rr = false;
 875	int key_plen;
 876
 877	/* make sure this function or its helpers sets f6i */
 878	res->f6i = NULL;
 879
 880	if (!leaf || leaf == net->ipv6.fib6_null_entry)
 881		goto out;
 882
 883	rt0 = rcu_dereference(fn->rr_ptr);
 884	if (!rt0)
 885		rt0 = leaf;
 886
 887	/* Double check to make sure fn is not an intermediate node
 888	 * and fn->leaf does not points to its child's leaf
 889	 * (This might happen if all routes under fn are deleted from
 890	 * the tree and fib6_repair_tree() is called on the node.)
 891	 */
 892	key_plen = rt0->fib6_dst.plen;
 893#ifdef CONFIG_IPV6_SUBTREES
 894	if (rt0->fib6_src.plen)
 895		key_plen = rt0->fib6_src.plen;
 896#endif
 897	if (fn->fn_bit != key_plen)
 898		goto out;
 899
 900	find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
 901	if (do_rr) {
 902		struct fib6_info *next = rcu_dereference(rt0->fib6_next);
 903
 904		/* no entries matched; do round-robin */
 905		if (!next || next->fib6_metric != rt0->fib6_metric)
 906			next = leaf;
 907
 908		if (next != rt0) {
 909			spin_lock_bh(&leaf->fib6_table->tb6_lock);
 910			/* make sure next is not being deleted from the tree */
 911			if (next->fib6_node)
 912				rcu_assign_pointer(fn->rr_ptr, next);
 913			spin_unlock_bh(&leaf->fib6_table->tb6_lock);
 914		}
 915	}
 916
 917out:
 918	if (!res->f6i) {
 919		res->f6i = net->ipv6.fib6_null_entry;
 920		res->nh = res->f6i->fib6_nh;
 921		res->fib6_flags = res->f6i->fib6_flags;
 922		res->fib6_type = res->f6i->fib6_type;
 923	}
 924}
 925
 926static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
 927{
 928	return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
 929	       res->nh->fib_nh_gw_family;
 930}
 931
 932#ifdef CONFIG_IPV6_ROUTE_INFO
 933int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 934		  const struct in6_addr *gwaddr)
 935{
 936	struct net *net = dev_net(dev);
 937	struct route_info *rinfo = (struct route_info *) opt;
 938	struct in6_addr prefix_buf, *prefix;
 939	struct fib6_table *table;
 940	unsigned int pref;
 941	unsigned long lifetime;
 942	struct fib6_info *rt;
 943
 944	if (len < sizeof(struct route_info)) {
 945		return -EINVAL;
 946	}
 947
 948	/* Sanity check for prefix_len and length */
 949	if (rinfo->length > 3) {
 950		return -EINVAL;
 951	} else if (rinfo->prefix_len > 128) {
 952		return -EINVAL;
 953	} else if (rinfo->prefix_len > 64) {
 954		if (rinfo->length < 2) {
 955			return -EINVAL;
 956		}
 957	} else if (rinfo->prefix_len > 0) {
 958		if (rinfo->length < 1) {
 959			return -EINVAL;
 960		}
 961	}
 962
 963	pref = rinfo->route_pref;
 964	if (pref == ICMPV6_ROUTER_PREF_INVALID)
 965		return -EINVAL;
 966
 967	lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 968
 969	if (rinfo->length == 3)
 970		prefix = (struct in6_addr *)rinfo->prefix;
 971	else {
 972		/* this function is safe */
 973		ipv6_addr_prefix(&prefix_buf,
 974				 (struct in6_addr *)rinfo->prefix,
 975				 rinfo->prefix_len);
 976		prefix = &prefix_buf;
 977	}
 978
 979	if (rinfo->prefix_len == 0)
 980		rt = rt6_get_dflt_router(net, gwaddr, dev);
 981	else
 982		rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 983					gwaddr, dev);
 984
 985	if (rt && !lifetime) {
 986		ip6_del_rt(net, rt, false);
 987		rt = NULL;
 988	}
 989
 990	if (!rt && lifetime)
 991		rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 992					dev, pref);
 993	else if (rt)
 994		rt->fib6_flags = RTF_ROUTEINFO |
 995				 (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 996
 997	if (rt) {
 998		table = rt->fib6_table;
 999		spin_lock_bh(&table->tb6_lock);
 
 
1000
1001		if (!addrconf_finite_timeout(lifetime)) {
1002			fib6_clean_expires(rt);
1003			fib6_remove_gc_list(rt);
1004		} else {
1005			fib6_set_expires(rt, jiffies + HZ * lifetime);
1006			fib6_add_gc_list(rt);
1007		}
1008
1009		spin_unlock_bh(&table->tb6_lock);
1010
1011		fib6_info_release(rt);
1012	}
1013	return 0;
1014}
1015#endif
1016
1017/*
1018 *	Misc support functions
1019 */
1020
1021/* called with rcu_lock held */
1022static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1023{
1024	struct net_device *dev = res->nh->fib_nh_dev;
1025
1026	if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1027		/* for copies of local routes, dst->dev needs to be the
1028		 * device if it is a master device, the master device if
1029		 * device is enslaved, and the loopback as the default
1030		 */
1031		if (netif_is_l3_slave(dev) &&
1032		    !rt6_need_strict(&res->f6i->fib6_dst.addr))
1033			dev = l3mdev_master_dev_rcu(dev);
1034		else if (!netif_is_l3_master(dev))
1035			dev = dev_net(dev)->loopback_dev;
1036		/* last case is netif_is_l3_master(dev) is true in which
1037		 * case we want dev returned to be dev
1038		 */
1039	}
1040
1041	return dev;
1042}
1043
1044static const int fib6_prop[RTN_MAX + 1] = {
1045	[RTN_UNSPEC]	= 0,
1046	[RTN_UNICAST]	= 0,
1047	[RTN_LOCAL]	= 0,
1048	[RTN_BROADCAST]	= 0,
1049	[RTN_ANYCAST]	= 0,
1050	[RTN_MULTICAST]	= 0,
1051	[RTN_BLACKHOLE]	= -EINVAL,
1052	[RTN_UNREACHABLE] = -EHOSTUNREACH,
1053	[RTN_PROHIBIT]	= -EACCES,
1054	[RTN_THROW]	= -EAGAIN,
1055	[RTN_NAT]	= -EINVAL,
1056	[RTN_XRESOLVE]	= -EINVAL,
1057};
1058
1059static int ip6_rt_type_to_error(u8 fib6_type)
1060{
1061	return fib6_prop[fib6_type];
1062}
1063
1064static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1065{
1066	unsigned short flags = 0;
1067
1068	if (rt->dst_nocount)
1069		flags |= DST_NOCOUNT;
1070	if (rt->dst_nopolicy)
1071		flags |= DST_NOPOLICY;
1072
1073	return flags;
1074}
1075
1076static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1077{
1078	rt->dst.error = ip6_rt_type_to_error(fib6_type);
1079
1080	switch (fib6_type) {
1081	case RTN_BLACKHOLE:
1082		rt->dst.output = dst_discard_out;
1083		rt->dst.input = dst_discard;
1084		break;
1085	case RTN_PROHIBIT:
1086		rt->dst.output = ip6_pkt_prohibit_out;
1087		rt->dst.input = ip6_pkt_prohibit;
1088		break;
1089	case RTN_THROW:
1090	case RTN_UNREACHABLE:
1091	default:
1092		rt->dst.output = ip6_pkt_discard_out;
1093		rt->dst.input = ip6_pkt_discard;
1094		break;
1095	}
1096}
1097
1098static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1099{
1100	struct fib6_info *f6i = res->f6i;
1101
1102	if (res->fib6_flags & RTF_REJECT) {
1103		ip6_rt_init_dst_reject(rt, res->fib6_type);
1104		return;
1105	}
1106
1107	rt->dst.error = 0;
1108	rt->dst.output = ip6_output;
1109
1110	if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1111		rt->dst.input = ip6_input;
1112	} else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1113		rt->dst.input = ip6_mc_input;
1114	} else {
1115		rt->dst.input = ip6_forward;
1116	}
1117
1118	if (res->nh->fib_nh_lws) {
1119		rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1120		lwtunnel_set_redirect(&rt->dst);
1121	}
1122
1123	rt->dst.lastuse = jiffies;
1124}
1125
1126/* Caller must already hold reference to @from */
1127static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1128{
1129	rt->rt6i_flags &= ~RTF_EXPIRES;
1130	rcu_assign_pointer(rt->from, from);
1131	ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1132}
1133
1134/* Caller must already hold reference to f6i in result */
1135static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1136{
1137	const struct fib6_nh *nh = res->nh;
1138	const struct net_device *dev = nh->fib_nh_dev;
1139	struct fib6_info *f6i = res->f6i;
1140
1141	ip6_rt_init_dst(rt, res);
1142
1143	rt->rt6i_dst = f6i->fib6_dst;
1144	rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1145	rt->rt6i_flags = res->fib6_flags;
1146	if (nh->fib_nh_gw_family) {
1147		rt->rt6i_gateway = nh->fib_nh_gw6;
1148		rt->rt6i_flags |= RTF_GATEWAY;
1149	}
1150	rt6_set_from(rt, f6i);
1151#ifdef CONFIG_IPV6_SUBTREES
1152	rt->rt6i_src = f6i->fib6_src;
1153#endif
1154}
1155
1156static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1157					struct in6_addr *saddr)
1158{
1159	struct fib6_node *pn, *sn;
1160	while (1) {
1161		if (fn->fn_flags & RTN_TL_ROOT)
1162			return NULL;
1163		pn = rcu_dereference(fn->parent);
1164		sn = FIB6_SUBTREE(pn);
1165		if (sn && sn != fn)
1166			fn = fib6_node_lookup(sn, NULL, saddr);
1167		else
1168			fn = pn;
1169		if (fn->fn_flags & RTN_RTINFO)
1170			return fn;
1171	}
1172}
1173
1174static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1175{
1176	struct rt6_info *rt = *prt;
1177
1178	if (dst_hold_safe(&rt->dst))
1179		return true;
1180	if (net) {
1181		rt = net->ipv6.ip6_null_entry;
1182		dst_hold(&rt->dst);
1183	} else {
1184		rt = NULL;
1185	}
1186	*prt = rt;
1187	return false;
1188}
1189
1190/* called with rcu_lock held */
1191static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1192{
1193	struct net_device *dev = res->nh->fib_nh_dev;
1194	struct fib6_info *f6i = res->f6i;
1195	unsigned short flags;
1196	struct rt6_info *nrt;
1197
1198	if (!fib6_info_hold_safe(f6i))
1199		goto fallback;
1200
1201	flags = fib6_info_dst_flags(f6i);
1202	nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1203	if (!nrt) {
1204		fib6_info_release(f6i);
1205		goto fallback;
1206	}
1207
1208	ip6_rt_copy_init(nrt, res);
1209	return nrt;
1210
1211fallback:
1212	nrt = dev_net(dev)->ipv6.ip6_null_entry;
1213	dst_hold(&nrt->dst);
1214	return nrt;
1215}
1216
1217INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net,
1218					     struct fib6_table *table,
1219					     struct flowi6 *fl6,
1220					     const struct sk_buff *skb,
1221					     int flags)
1222{
1223	struct fib6_result res = {};
1224	struct fib6_node *fn;
1225	struct rt6_info *rt;
1226
1227	rcu_read_lock();
1228	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1229restart:
1230	res.f6i = rcu_dereference(fn->leaf);
1231	if (!res.f6i)
1232		res.f6i = net->ipv6.fib6_null_entry;
1233	else
1234		rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1235				 flags);
1236
1237	if (res.f6i == net->ipv6.fib6_null_entry) {
1238		fn = fib6_backtrack(fn, &fl6->saddr);
1239		if (fn)
1240			goto restart;
1241
1242		rt = net->ipv6.ip6_null_entry;
1243		dst_hold(&rt->dst);
1244		goto out;
1245	} else if (res.fib6_flags & RTF_REJECT) {
1246		goto do_create;
1247	}
1248
1249	fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1250			 fl6->flowi6_oif != 0, skb, flags);
1251
1252	/* Search through exception table */
1253	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1254	if (rt) {
1255		if (ip6_hold_safe(net, &rt))
1256			dst_use_noref(&rt->dst, jiffies);
1257	} else {
1258do_create:
1259		rt = ip6_create_rt_rcu(&res);
1260	}
1261
1262out:
1263	trace_fib6_table_lookup(net, &res, table, fl6);
 
 
1264
1265	rcu_read_unlock();
1266
1267	return rt;
1268}
1269
1270struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1271				   const struct sk_buff *skb, int flags)
1272{
1273	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1274}
1275EXPORT_SYMBOL_GPL(ip6_route_lookup);
1276
1277struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1278			    const struct in6_addr *saddr, int oif,
1279			    const struct sk_buff *skb, int strict)
1280{
1281	struct flowi6 fl6 = {
1282		.flowi6_oif = oif,
1283		.daddr = *daddr,
1284	};
1285	struct dst_entry *dst;
1286	int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1287
1288	if (saddr) {
1289		memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1290		flags |= RT6_LOOKUP_F_HAS_SADDR;
1291	}
1292
1293	dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1294	if (dst->error == 0)
1295		return dst_rt6_info(dst);
1296
1297	dst_release(dst);
1298
1299	return NULL;
1300}
 
1301EXPORT_SYMBOL(rt6_lookup);
1302
1303/* ip6_ins_rt is called with FREE table->tb6_lock.
1304 * It takes new route entry, the addition fails by any reason the
1305 * route is released.
1306 * Caller must hold dst before calling it.
1307 */
1308
1309static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1310			struct netlink_ext_ack *extack)
1311{
1312	int err;
1313	struct fib6_table *table;
1314
1315	table = rt->fib6_table;
1316	spin_lock_bh(&table->tb6_lock);
1317	err = fib6_add(&table->tb6_root, rt, info, extack);
1318	spin_unlock_bh(&table->tb6_lock);
1319
1320	return err;
1321}
1322
1323int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1324{
1325	struct nl_info info = {	.nl_net = net, };
1326
1327	return __ip6_ins_rt(rt, &info, NULL);
 
1328}
1329
1330static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1331					   const struct in6_addr *daddr,
1332					   const struct in6_addr *saddr)
1333{
1334	struct fib6_info *f6i = res->f6i;
1335	struct net_device *dev;
1336	struct rt6_info *rt;
1337
1338	/*
1339	 *	Clone the route.
1340	 */
1341
1342	if (!fib6_info_hold_safe(f6i))
1343		return NULL;
1344
1345	dev = ip6_rt_get_dev_rcu(res);
1346	rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1347	if (!rt) {
1348		fib6_info_release(f6i);
1349		return NULL;
1350	}
1351
1352	ip6_rt_copy_init(rt, res);
1353	rt->rt6i_flags |= RTF_CACHE;
1354	rt->rt6i_dst.addr = *daddr;
1355	rt->rt6i_dst.plen = 128;
1356
1357	if (!rt6_is_gw_or_nonexthop(res)) {
1358		if (f6i->fib6_dst.plen != 128 &&
1359		    ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1360			rt->rt6i_flags |= RTF_ANYCAST;
1361#ifdef CONFIG_IPV6_SUBTREES
1362		if (rt->rt6i_src.plen && saddr) {
1363			rt->rt6i_src.addr = *saddr;
1364			rt->rt6i_src.plen = 128;
1365		}
1366#endif
1367	}
1368
1369	return rt;
1370}
1371
1372static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
 
1373{
1374	struct fib6_info *f6i = res->f6i;
1375	unsigned short flags = fib6_info_dst_flags(f6i);
1376	struct net_device *dev;
1377	struct rt6_info *pcpu_rt;
1378
1379	if (!fib6_info_hold_safe(f6i))
1380		return NULL;
1381
1382	rcu_read_lock();
1383	dev = ip6_rt_get_dev_rcu(res);
1384	pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT);
1385	rcu_read_unlock();
1386	if (!pcpu_rt) {
1387		fib6_info_release(f6i);
1388		return NULL;
1389	}
1390	ip6_rt_copy_init(pcpu_rt, res);
1391	pcpu_rt->rt6i_flags |= RTF_PCPU;
1392
1393	if (f6i->nh)
1394		pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev));
1395
1396	return pcpu_rt;
1397}
1398
1399static bool rt6_is_valid(const struct rt6_info *rt6)
 
1400{
1401	return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev));
1402}
 
 
 
 
1403
1404/* It should be called with rcu_read_lock() acquired */
1405static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1406{
1407	struct rt6_info *pcpu_rt;
1408
1409	pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
 
1410
1411	if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) {
1412		struct rt6_info *prev, **p;
1413
1414		p = this_cpu_ptr(res->nh->rt6i_pcpu);
1415		/* Paired with READ_ONCE() in __fib6_drop_pcpu_from() */
1416		prev = xchg(p, NULL);
1417		if (prev) {
1418			dst_dev_put(&prev->dst);
1419			dst_release(&prev->dst);
1420		}
1421
1422		pcpu_rt = NULL;
1423	}
1424
1425	return pcpu_rt;
1426}
1427
1428static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1429					    const struct fib6_result *res)
1430{
1431	struct rt6_info *pcpu_rt, *prev, **p;
1432
1433	pcpu_rt = ip6_rt_pcpu_alloc(res);
1434	if (!pcpu_rt)
1435		return NULL;
1436
1437	p = this_cpu_ptr(res->nh->rt6i_pcpu);
1438	prev = cmpxchg(p, NULL, pcpu_rt);
1439	BUG_ON(prev);
1440
1441	if (res->f6i->fib6_destroying) {
1442		struct fib6_info *from;
1443
1444		from = unrcu_pointer(xchg(&pcpu_rt->from, NULL));
1445		fib6_info_release(from);
1446	}
1447
1448	return pcpu_rt;
1449}
1450
1451/* exception hash table implementation
1452 */
1453static DEFINE_SPINLOCK(rt6_exception_lock);
1454
1455/* Remove rt6_ex from hash table and free the memory
1456 * Caller must hold rt6_exception_lock
1457 */
1458static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1459				 struct rt6_exception *rt6_ex)
1460{
1461	struct net *net;
1462
1463	if (!bucket || !rt6_ex)
1464		return;
1465
1466	net = dev_net(rt6_ex->rt6i->dst.dev);
1467	net->ipv6.rt6_stats->fib_rt_cache--;
1468
1469	/* purge completely the exception to allow releasing the held resources:
1470	 * some [sk] cache may keep the dst around for unlimited time
1471	 */
1472	dst_dev_put(&rt6_ex->rt6i->dst);
1473
1474	hlist_del_rcu(&rt6_ex->hlist);
1475	dst_release(&rt6_ex->rt6i->dst);
1476	kfree_rcu(rt6_ex, rcu);
1477	WARN_ON_ONCE(!bucket->depth);
1478	bucket->depth--;
1479}
1480
1481/* Remove oldest rt6_ex in bucket and free the memory
1482 * Caller must hold rt6_exception_lock
1483 */
1484static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1485{
1486	struct rt6_exception *rt6_ex, *oldest = NULL;
1487
1488	if (!bucket)
1489		return;
1490
1491	hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1492		if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1493			oldest = rt6_ex;
1494	}
1495	rt6_remove_exception(bucket, oldest);
1496}
1497
1498static u32 rt6_exception_hash(const struct in6_addr *dst,
1499			      const struct in6_addr *src)
1500{
1501	static siphash_aligned_key_t rt6_exception_key;
1502	struct {
1503		struct in6_addr dst;
1504		struct in6_addr src;
1505	} __aligned(SIPHASH_ALIGNMENT) combined = {
1506		.dst = *dst,
1507	};
1508	u64 val;
1509
1510	net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key));
1511
1512#ifdef CONFIG_IPV6_SUBTREES
1513	if (src)
1514		combined.src = *src;
1515#endif
1516	val = siphash(&combined, sizeof(combined), &rt6_exception_key);
1517
1518	return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1519}
1520
1521/* Helper function to find the cached rt in the hash table
1522 * and update bucket pointer to point to the bucket for this
1523 * (daddr, saddr) pair
1524 * Caller must hold rt6_exception_lock
1525 */
1526static struct rt6_exception *
1527__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1528			      const struct in6_addr *daddr,
1529			      const struct in6_addr *saddr)
1530{
1531	struct rt6_exception *rt6_ex;
1532	u32 hval;
1533
1534	if (!(*bucket) || !daddr)
1535		return NULL;
1536
1537	hval = rt6_exception_hash(daddr, saddr);
1538	*bucket += hval;
1539
1540	hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1541		struct rt6_info *rt6 = rt6_ex->rt6i;
1542		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1543
1544#ifdef CONFIG_IPV6_SUBTREES
1545		if (matched && saddr)
1546			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1547#endif
1548		if (matched)
1549			return rt6_ex;
1550	}
1551	return NULL;
1552}
1553
1554/* Helper function to find the cached rt in the hash table
1555 * and update bucket pointer to point to the bucket for this
1556 * (daddr, saddr) pair
1557 * Caller must hold rcu_read_lock()
1558 */
1559static struct rt6_exception *
1560__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1561			 const struct in6_addr *daddr,
1562			 const struct in6_addr *saddr)
1563{
1564	struct rt6_exception *rt6_ex;
1565	u32 hval;
1566
1567	WARN_ON_ONCE(!rcu_read_lock_held());
1568
1569	if (!(*bucket) || !daddr)
1570		return NULL;
1571
1572	hval = rt6_exception_hash(daddr, saddr);
1573	*bucket += hval;
1574
1575	hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1576		struct rt6_info *rt6 = rt6_ex->rt6i;
1577		bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1578
1579#ifdef CONFIG_IPV6_SUBTREES
1580		if (matched && saddr)
1581			matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1582#endif
1583		if (matched)
1584			return rt6_ex;
1585	}
1586	return NULL;
1587}
1588
1589static unsigned int fib6_mtu(const struct fib6_result *res)
1590{
1591	const struct fib6_nh *nh = res->nh;
1592	unsigned int mtu;
1593
1594	if (res->f6i->fib6_pmtu) {
1595		mtu = res->f6i->fib6_pmtu;
1596	} else {
1597		struct net_device *dev = nh->fib_nh_dev;
1598		struct inet6_dev *idev;
1599
1600		rcu_read_lock();
1601		idev = __in6_dev_get(dev);
1602		mtu = READ_ONCE(idev->cnf.mtu6);
1603		rcu_read_unlock();
1604	}
1605
1606	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1607
1608	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1609}
1610
1611#define FIB6_EXCEPTION_BUCKET_FLUSHED  0x1UL
1612
1613/* used when the flushed bit is not relevant, only access to the bucket
1614 * (ie., all bucket users except rt6_insert_exception);
1615 *
1616 * called under rcu lock; sometimes called with rt6_exception_lock held
1617 */
1618static
1619struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1620						       spinlock_t *lock)
1621{
1622	struct rt6_exception_bucket *bucket;
1623
1624	if (lock)
1625		bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1626						   lockdep_is_held(lock));
1627	else
1628		bucket = rcu_dereference(nh->rt6i_exception_bucket);
1629
1630	/* remove bucket flushed bit if set */
1631	if (bucket) {
1632		unsigned long p = (unsigned long)bucket;
1633
1634		p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1635		bucket = (struct rt6_exception_bucket *)p;
 
 
 
1636	}
1637
1638	return bucket;
1639}
1640
1641static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1642{
1643	unsigned long p = (unsigned long)bucket;
1644
1645	return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1646}
1647
1648/* called with rt6_exception_lock held */
1649static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1650					      spinlock_t *lock)
1651{
1652	struct rt6_exception_bucket *bucket;
1653	unsigned long p;
1654
1655	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1656					   lockdep_is_held(lock));
1657
1658	p = (unsigned long)bucket;
1659	p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1660	bucket = (struct rt6_exception_bucket *)p;
1661	rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1662}
1663
1664static int rt6_insert_exception(struct rt6_info *nrt,
1665				const struct fib6_result *res)
1666{
1667	struct net *net = dev_net(nrt->dst.dev);
1668	struct rt6_exception_bucket *bucket;
1669	struct fib6_info *f6i = res->f6i;
1670	struct in6_addr *src_key = NULL;
1671	struct rt6_exception *rt6_ex;
1672	struct fib6_nh *nh = res->nh;
1673	int max_depth;
1674	int err = 0;
1675
1676	spin_lock_bh(&rt6_exception_lock);
1677
1678	bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1679					  lockdep_is_held(&rt6_exception_lock));
1680	if (!bucket) {
1681		bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1682				 GFP_ATOMIC);
1683		if (!bucket) {
1684			err = -ENOMEM;
1685			goto out;
1686		}
1687		rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1688	} else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1689		err = -EINVAL;
1690		goto out;
1691	}
1692
1693#ifdef CONFIG_IPV6_SUBTREES
1694	/* fib6_src.plen != 0 indicates f6i is in subtree
1695	 * and exception table is indexed by a hash of
1696	 * both fib6_dst and fib6_src.
1697	 * Otherwise, the exception table is indexed by
1698	 * a hash of only fib6_dst.
1699	 */
1700	if (f6i->fib6_src.plen)
1701		src_key = &nrt->rt6i_src.addr;
1702#endif
1703	/* rt6_mtu_change() might lower mtu on f6i.
1704	 * Only insert this exception route if its mtu
1705	 * is less than f6i's mtu value.
1706	 */
1707	if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1708		err = -EINVAL;
1709		goto out;
1710	}
1711
1712	rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1713					       src_key);
1714	if (rt6_ex)
1715		rt6_remove_exception(bucket, rt6_ex);
1716
1717	rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1718	if (!rt6_ex) {
1719		err = -ENOMEM;
1720		goto out;
1721	}
1722	rt6_ex->rt6i = nrt;
1723	rt6_ex->stamp = jiffies;
1724	hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1725	bucket->depth++;
1726	net->ipv6.rt6_stats->fib_rt_cache++;
1727
1728	/* Randomize max depth to avoid some side channels attacks. */
1729	max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH);
1730	while (bucket->depth > max_depth)
1731		rt6_exception_remove_oldest(bucket);
1732
1733out:
1734	spin_unlock_bh(&rt6_exception_lock);
1735
1736	/* Update fn->fn_sernum to invalidate all cached dst */
1737	if (!err) {
1738		spin_lock_bh(&f6i->fib6_table->tb6_lock);
1739		fib6_update_sernum(net, f6i);
1740		spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1741		fib6_force_start_gc(net);
1742	}
 
 
 
 
 
1743
1744	return err;
1745}
1746
1747static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1748{
1749	struct rt6_exception_bucket *bucket;
1750	struct rt6_exception *rt6_ex;
1751	struct hlist_node *tmp;
1752	int i;
1753
1754	spin_lock_bh(&rt6_exception_lock);
1755
1756	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1757	if (!bucket)
1758		goto out;
1759
1760	/* Prevent rt6_insert_exception() to recreate the bucket list */
1761	if (!from)
1762		fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1763
1764	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1765		hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1766			if (!from ||
1767			    rcu_access_pointer(rt6_ex->rt6i->from) == from)
1768				rt6_remove_exception(bucket, rt6_ex);
1769		}
1770		WARN_ON_ONCE(!from && bucket->depth);
1771		bucket++;
1772	}
1773out:
1774	spin_unlock_bh(&rt6_exception_lock);
1775}
1776
1777static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
 
1778{
1779	struct fib6_info *f6i = arg;
1780
1781	fib6_nh_flush_exceptions(nh, f6i);
1782
1783	return 0;
1784}
1785
1786void rt6_flush_exceptions(struct fib6_info *f6i)
1787{
1788	if (f6i->nh)
1789		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1790					 f6i);
1791	else
1792		fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1793}
1794
1795/* Find cached rt in the hash table inside passed in rt
1796 * Caller has to hold rcu_read_lock()
1797 */
1798static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1799					   const struct in6_addr *daddr,
1800					   const struct in6_addr *saddr)
1801{
1802	const struct in6_addr *src_key = NULL;
1803	struct rt6_exception_bucket *bucket;
1804	struct rt6_exception *rt6_ex;
1805	struct rt6_info *ret = NULL;
1806
1807#ifdef CONFIG_IPV6_SUBTREES
1808	/* fib6i_src.plen != 0 indicates f6i is in subtree
1809	 * and exception table is indexed by a hash of
1810	 * both fib6_dst and fib6_src.
1811	 * However, the src addr used to create the hash
1812	 * might not be exactly the passed in saddr which
1813	 * is a /128 addr from the flow.
1814	 * So we need to use f6i->fib6_src to redo lookup
1815	 * if the passed in saddr does not find anything.
1816	 * (See the logic in ip6_rt_cache_alloc() on how
1817	 * rt->rt6i_src is updated.)
1818	 */
1819	if (res->f6i->fib6_src.plen)
1820		src_key = saddr;
1821find_ex:
1822#endif
1823	bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1824	rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1825
1826	if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1827		ret = rt6_ex->rt6i;
1828
1829#ifdef CONFIG_IPV6_SUBTREES
1830	/* Use fib6_src as src_key and redo lookup */
1831	if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1832		src_key = &res->f6i->fib6_src.addr;
1833		goto find_ex;
1834	}
1835#endif
1836
1837	return ret;
1838}
1839
1840/* Remove the passed in cached rt from the hash table that contains it */
1841static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1842				    const struct rt6_info *rt)
1843{
1844	const struct in6_addr *src_key = NULL;
1845	struct rt6_exception_bucket *bucket;
1846	struct rt6_exception *rt6_ex;
1847	int err;
1848
1849	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1850		return -ENOENT;
1851
1852	spin_lock_bh(&rt6_exception_lock);
1853	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1854
1855#ifdef CONFIG_IPV6_SUBTREES
1856	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1857	 * and exception table is indexed by a hash of
1858	 * both rt6i_dst and rt6i_src.
1859	 * Otherwise, the exception table is indexed by
1860	 * a hash of only rt6i_dst.
1861	 */
1862	if (plen)
1863		src_key = &rt->rt6i_src.addr;
1864#endif
1865	rt6_ex = __rt6_find_exception_spinlock(&bucket,
1866					       &rt->rt6i_dst.addr,
1867					       src_key);
1868	if (rt6_ex) {
1869		rt6_remove_exception(bucket, rt6_ex);
1870		err = 0;
1871	} else {
1872		err = -ENOENT;
1873	}
1874
1875	spin_unlock_bh(&rt6_exception_lock);
1876	return err;
1877}
1878
1879struct fib6_nh_excptn_arg {
1880	struct rt6_info	*rt;
1881	int		plen;
1882};
1883
1884static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1885{
1886	struct fib6_nh_excptn_arg *arg = _arg;
1887	int err;
1888
1889	err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1890	if (err == 0)
1891		return 1;
1892
1893	return 0;
1894}
1895
1896static int rt6_remove_exception_rt(struct rt6_info *rt)
1897{
1898	struct fib6_info *from;
1899
1900	from = rcu_dereference(rt->from);
1901	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1902		return -EINVAL;
1903
1904	if (from->nh) {
1905		struct fib6_nh_excptn_arg arg = {
1906			.rt = rt,
1907			.plen = from->fib6_src.plen
1908		};
1909		int rc;
1910
1911		/* rc = 1 means an entry was found */
1912		rc = nexthop_for_each_fib6_nh(from->nh,
1913					      rt6_nh_remove_exception_rt,
1914					      &arg);
1915		return rc ? 0 : -ENOENT;
1916	}
1917
1918	return fib6_nh_remove_exception(from->fib6_nh,
1919					from->fib6_src.plen, rt);
1920}
1921
1922/* Find rt6_ex which contains the passed in rt cache and
1923 * refresh its stamp
1924 */
1925static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1926				     const struct rt6_info *rt)
1927{
1928	const struct in6_addr *src_key = NULL;
1929	struct rt6_exception_bucket *bucket;
1930	struct rt6_exception *rt6_ex;
1931
1932	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1933#ifdef CONFIG_IPV6_SUBTREES
1934	/* rt6i_src.plen != 0 indicates 'from' is in subtree
1935	 * and exception table is indexed by a hash of
1936	 * both rt6i_dst and rt6i_src.
1937	 * Otherwise, the exception table is indexed by
1938	 * a hash of only rt6i_dst.
1939	 */
1940	if (plen)
1941		src_key = &rt->rt6i_src.addr;
1942#endif
1943	rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1944	if (rt6_ex)
1945		rt6_ex->stamp = jiffies;
1946}
1947
1948struct fib6_nh_match_arg {
1949	const struct net_device *dev;
1950	const struct in6_addr	*gw;
1951	struct fib6_nh		*match;
1952};
1953
1954/* determine if fib6_nh has given device and gateway */
1955static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1956{
1957	struct fib6_nh_match_arg *arg = _arg;
1958
1959	if (arg->dev != nh->fib_nh_dev ||
1960	    (arg->gw && !nh->fib_nh_gw_family) ||
1961	    (!arg->gw && nh->fib_nh_gw_family) ||
1962	    (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1963		return 0;
1964
1965	arg->match = nh;
1966
1967	/* found a match, break the loop */
1968	return 1;
1969}
1970
1971static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1972{
1973	struct fib6_info *from;
1974	struct fib6_nh *fib6_nh;
1975
1976	rcu_read_lock();
1977
1978	from = rcu_dereference(rt->from);
1979	if (!from || !(rt->rt6i_flags & RTF_CACHE))
1980		goto unlock;
1981
1982	if (from->nh) {
1983		struct fib6_nh_match_arg arg = {
1984			.dev = rt->dst.dev,
1985			.gw = &rt->rt6i_gateway,
1986		};
1987
1988		nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1989
1990		if (!arg.match)
1991			goto unlock;
1992		fib6_nh = arg.match;
1993	} else {
1994		fib6_nh = from->fib6_nh;
1995	}
1996	fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1997unlock:
1998	rcu_read_unlock();
1999}
2000
2001static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
2002					 struct rt6_info *rt, int mtu)
2003{
2004	/* If the new MTU is lower than the route PMTU, this new MTU will be the
2005	 * lowest MTU in the path: always allow updating the route PMTU to
2006	 * reflect PMTU decreases.
2007	 *
2008	 * If the new MTU is higher, and the route PMTU is equal to the local
2009	 * MTU, this means the old MTU is the lowest in the path, so allow
2010	 * updating it: if other nodes now have lower MTUs, PMTU discovery will
2011	 * handle this.
2012	 */
2013
2014	if (dst_mtu(&rt->dst) >= mtu)
2015		return true;
2016
2017	if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
2018		return true;
2019
2020	return false;
2021}
2022
2023static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
2024				       const struct fib6_nh *nh, int mtu)
2025{
2026	struct rt6_exception_bucket *bucket;
2027	struct rt6_exception *rt6_ex;
2028	int i;
2029
2030	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2031	if (!bucket)
2032		return;
2033
2034	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2035		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2036			struct rt6_info *entry = rt6_ex->rt6i;
2037
2038			/* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2039			 * route), the metrics of its rt->from have already
2040			 * been updated.
2041			 */
2042			if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2043			    rt6_mtu_change_route_allowed(idev, entry, mtu))
2044				dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2045		}
2046		bucket++;
2047	}
2048}
2049
2050#define RTF_CACHE_GATEWAY	(RTF_GATEWAY | RTF_CACHE)
2051
2052static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2053					    const struct in6_addr *gateway)
2054{
2055	struct rt6_exception_bucket *bucket;
2056	struct rt6_exception *rt6_ex;
2057	struct hlist_node *tmp;
2058	int i;
2059
2060	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2061		return;
2062
2063	spin_lock_bh(&rt6_exception_lock);
2064	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2065	if (bucket) {
2066		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2067			hlist_for_each_entry_safe(rt6_ex, tmp,
2068						  &bucket->chain, hlist) {
2069				struct rt6_info *entry = rt6_ex->rt6i;
2070
2071				if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2072				    RTF_CACHE_GATEWAY &&
2073				    ipv6_addr_equal(gateway,
2074						    &entry->rt6i_gateway)) {
2075					rt6_remove_exception(bucket, rt6_ex);
2076				}
2077			}
2078			bucket++;
2079		}
2080	}
2081
2082	spin_unlock_bh(&rt6_exception_lock);
2083}
2084
2085static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2086				      struct rt6_exception *rt6_ex,
2087				      struct fib6_gc_args *gc_args,
2088				      unsigned long now)
2089{
2090	struct rt6_info *rt = rt6_ex->rt6i;
2091
2092	/* we are pruning and obsoleting aged-out and non gateway exceptions
2093	 * even if others have still references to them, so that on next
2094	 * dst_check() such references can be dropped.
2095	 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2096	 * expired, independently from their aging, as per RFC 8201 section 4
2097	 */
2098	if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2099		if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2100			pr_debug("aging clone %p\n", rt);
2101			rt6_remove_exception(bucket, rt6_ex);
2102			return;
2103		}
2104	} else if (time_after(jiffies, rt->dst.expires)) {
2105		pr_debug("purging expired route %p\n", rt);
2106		rt6_remove_exception(bucket, rt6_ex);
2107		return;
2108	}
2109
2110	if (rt->rt6i_flags & RTF_GATEWAY) {
2111		struct neighbour *neigh;
2112
2113		neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2114
2115		if (!(neigh && (neigh->flags & NTF_ROUTER))) {
2116			pr_debug("purging route %p via non-router but gateway\n",
2117				 rt);
2118			rt6_remove_exception(bucket, rt6_ex);
2119			return;
2120		}
2121	}
2122
2123	gc_args->more++;
2124}
2125
2126static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2127				   struct fib6_gc_args *gc_args,
2128				   unsigned long now)
2129{
2130	struct rt6_exception_bucket *bucket;
2131	struct rt6_exception *rt6_ex;
2132	struct hlist_node *tmp;
2133	int i;
2134
2135	if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2136		return;
2137
2138	rcu_read_lock_bh();
2139	spin_lock(&rt6_exception_lock);
2140	bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2141	if (bucket) {
2142		for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2143			hlist_for_each_entry_safe(rt6_ex, tmp,
2144						  &bucket->chain, hlist) {
2145				rt6_age_examine_exception(bucket, rt6_ex,
2146							  gc_args, now);
2147			}
2148			bucket++;
2149		}
2150	}
2151	spin_unlock(&rt6_exception_lock);
2152	rcu_read_unlock_bh();
2153}
2154
2155struct fib6_nh_age_excptn_arg {
2156	struct fib6_gc_args	*gc_args;
2157	unsigned long		now;
2158};
2159
2160static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2161{
2162	struct fib6_nh_age_excptn_arg *arg = _arg;
2163
2164	fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2165	return 0;
2166}
2167
2168void rt6_age_exceptions(struct fib6_info *f6i,
2169			struct fib6_gc_args *gc_args,
2170			unsigned long now)
2171{
2172	if (f6i->nh) {
2173		struct fib6_nh_age_excptn_arg arg = {
2174			.gc_args = gc_args,
2175			.now = now
2176		};
2177
2178		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2179					 &arg);
2180	} else {
2181		fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2182	}
2183}
2184
2185/* must be called with rcu lock held */
2186int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2187		      struct flowi6 *fl6, struct fib6_result *res, int strict)
2188{
2189	struct fib6_node *fn, *saved_fn;
2190
2191	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2192	saved_fn = fn;
2193
2194redo_rt6_select:
2195	rt6_select(net, fn, oif, res, strict);
2196	if (res->f6i == net->ipv6.fib6_null_entry) {
2197		fn = fib6_backtrack(fn, &fl6->saddr);
2198		if (fn)
2199			goto redo_rt6_select;
2200		else if (strict & RT6_LOOKUP_F_REACHABLE) {
2201			/* also consider unreachable route */
2202			strict &= ~RT6_LOOKUP_F_REACHABLE;
2203			fn = saved_fn;
2204			goto redo_rt6_select;
2205		}
2206	}
2207
2208	trace_fib6_table_lookup(net, res, table, fl6);
2209
2210	return 0;
2211}
2212
2213struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2214			       int oif, struct flowi6 *fl6,
2215			       const struct sk_buff *skb, int flags)
2216{
2217	struct fib6_result res = {};
2218	struct rt6_info *rt = NULL;
2219	int strict = 0;
2220
2221	WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2222		     !rcu_read_lock_held());
2223
2224	strict |= flags & RT6_LOOKUP_F_IFACE;
2225	strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2226	if (READ_ONCE(net->ipv6.devconf_all->forwarding) == 0)
2227		strict |= RT6_LOOKUP_F_REACHABLE;
2228
2229	rcu_read_lock();
2230
2231	fib6_table_lookup(net, table, oif, fl6, &res, strict);
2232	if (res.f6i == net->ipv6.fib6_null_entry)
2233		goto out;
2234
2235	fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2236
2237	/*Search through exception table */
2238	rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2239	if (rt) {
2240		goto out;
2241	} else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2242			    !res.nh->fib_nh_gw_family)) {
2243		/* Create a RTF_CACHE clone which will not be
2244		 * owned by the fib6 tree.  It is for the special case where
2245		 * the daddr in the skb during the neighbor look-up is different
2246		 * from the fl6->daddr used to look-up route here.
2247		 */
2248		rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2249
2250		if (rt) {
2251			/* 1 refcnt is taken during ip6_rt_cache_alloc().
2252			 * As rt6_uncached_list_add() does not consume refcnt,
2253			 * this refcnt is always returned to the caller even
2254			 * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2255			 */
2256			rt6_uncached_list_add(rt);
2257			rcu_read_unlock();
2258
2259			return rt;
2260		}
2261	} else {
2262		/* Get a percpu copy */
2263		local_bh_disable();
2264		rt = rt6_get_pcpu_route(&res);
2265
2266		if (!rt)
2267			rt = rt6_make_pcpu_route(net, &res);
2268
2269		local_bh_enable();
2270	}
2271out:
2272	if (!rt)
2273		rt = net->ipv6.ip6_null_entry;
2274	if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2275		ip6_hold_safe(net, &rt);
2276	rcu_read_unlock();
2277
2278	return rt;
2279}
2280EXPORT_SYMBOL_GPL(ip6_pol_route);
2281
2282INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net,
2283					    struct fib6_table *table,
2284					    struct flowi6 *fl6,
2285					    const struct sk_buff *skb,
2286					    int flags)
2287{
2288	return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2289}
2290
2291struct dst_entry *ip6_route_input_lookup(struct net *net,
2292					 struct net_device *dev,
2293					 struct flowi6 *fl6,
2294					 const struct sk_buff *skb,
2295					 int flags)
2296{
2297	if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2298		flags |= RT6_LOOKUP_F_IFACE;
2299
2300	return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2301}
2302EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2303
2304static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2305				  struct flow_keys *keys,
2306				  struct flow_keys *flkeys)
2307{
2308	const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2309	const struct ipv6hdr *key_iph = outer_iph;
2310	struct flow_keys *_flkeys = flkeys;
2311	const struct ipv6hdr *inner_iph;
2312	const struct icmp6hdr *icmph;
2313	struct ipv6hdr _inner_iph;
2314	struct icmp6hdr _icmph;
2315
2316	if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2317		goto out;
2318
2319	icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2320				   sizeof(_icmph), &_icmph);
2321	if (!icmph)
2322		goto out;
2323
2324	if (!icmpv6_is_err(icmph->icmp6_type))
2325		goto out;
2326
2327	inner_iph = skb_header_pointer(skb,
2328				       skb_transport_offset(skb) + sizeof(*icmph),
2329				       sizeof(_inner_iph), &_inner_iph);
2330	if (!inner_iph)
2331		goto out;
2332
2333	key_iph = inner_iph;
2334	_flkeys = NULL;
2335out:
2336	if (_flkeys) {
2337		keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2338		keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2339		keys->tags.flow_label = _flkeys->tags.flow_label;
2340		keys->basic.ip_proto = _flkeys->basic.ip_proto;
2341	} else {
2342		keys->addrs.v6addrs.src = key_iph->saddr;
2343		keys->addrs.v6addrs.dst = key_iph->daddr;
2344		keys->tags.flow_label = ip6_flowlabel(key_iph);
2345		keys->basic.ip_proto = key_iph->nexthdr;
2346	}
2347}
2348
2349static u32 rt6_multipath_custom_hash_outer(const struct net *net,
2350					   const struct sk_buff *skb,
2351					   bool *p_has_inner)
2352{
2353	u32 hash_fields = ip6_multipath_hash_fields(net);
2354	struct flow_keys keys, hash_keys;
2355
2356	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2357		return 0;
2358
2359	memset(&hash_keys, 0, sizeof(hash_keys));
2360	skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP);
2361
2362	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2363	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2364		hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2365	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2366		hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2367	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2368		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2369	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2370		hash_keys.tags.flow_label = keys.tags.flow_label;
2371	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2372		hash_keys.ports.src = keys.ports.src;
2373	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2374		hash_keys.ports.dst = keys.ports.dst;
2375
2376	*p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION);
2377	return fib_multipath_hash_from_keys(net, &hash_keys);
2378}
2379
2380static u32 rt6_multipath_custom_hash_inner(const struct net *net,
2381					   const struct sk_buff *skb,
2382					   bool has_inner)
2383{
2384	u32 hash_fields = ip6_multipath_hash_fields(net);
2385	struct flow_keys keys, hash_keys;
2386
2387	/* We assume the packet carries an encapsulation, but if none was
2388	 * encountered during dissection of the outer flow, then there is no
2389	 * point in calling the flow dissector again.
2390	 */
2391	if (!has_inner)
2392		return 0;
2393
2394	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK))
2395		return 0;
2396
2397	memset(&hash_keys, 0, sizeof(hash_keys));
2398	skb_flow_dissect_flow_keys(skb, &keys, 0);
2399
2400	if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION))
2401		return 0;
2402
2403	if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2404		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2405		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2406			hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
2407		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2408			hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst;
2409	} else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2410		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2411		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP)
2412			hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
2413		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP)
2414			hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst;
2415		if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL)
2416			hash_keys.tags.flow_label = keys.tags.flow_label;
2417	}
2418
2419	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO)
2420		hash_keys.basic.ip_proto = keys.basic.ip_proto;
2421	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT)
2422		hash_keys.ports.src = keys.ports.src;
2423	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT)
2424		hash_keys.ports.dst = keys.ports.dst;
2425
2426	return fib_multipath_hash_from_keys(net, &hash_keys);
2427}
2428
2429static u32 rt6_multipath_custom_hash_skb(const struct net *net,
2430					 const struct sk_buff *skb)
2431{
2432	u32 mhash, mhash_inner;
2433	bool has_inner = true;
2434
2435	mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner);
2436	mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner);
2437
2438	return jhash_2words(mhash, mhash_inner, 0);
2439}
2440
2441static u32 rt6_multipath_custom_hash_fl6(const struct net *net,
2442					 const struct flowi6 *fl6)
2443{
2444	u32 hash_fields = ip6_multipath_hash_fields(net);
2445	struct flow_keys hash_keys;
2446
2447	if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK))
2448		return 0;
2449
2450	memset(&hash_keys, 0, sizeof(hash_keys));
2451	hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2452	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP)
2453		hash_keys.addrs.v6addrs.src = fl6->saddr;
2454	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP)
2455		hash_keys.addrs.v6addrs.dst = fl6->daddr;
2456	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO)
2457		hash_keys.basic.ip_proto = fl6->flowi6_proto;
2458	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL)
2459		hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2460	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT)
2461		hash_keys.ports.src = fl6->fl6_sport;
2462	if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT)
2463		hash_keys.ports.dst = fl6->fl6_dport;
2464
2465	return fib_multipath_hash_from_keys(net, &hash_keys);
2466}
2467
2468/* if skb is set it will be used and fl6 can be NULL */
2469u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2470		       const struct sk_buff *skb, struct flow_keys *flkeys)
2471{
2472	struct flow_keys hash_keys;
2473	u32 mhash = 0;
2474
2475	switch (ip6_multipath_hash_policy(net)) {
2476	case 0:
2477		memset(&hash_keys, 0, sizeof(hash_keys));
2478		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2479		if (skb) {
2480			ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2481		} else {
2482			hash_keys.addrs.v6addrs.src = fl6->saddr;
2483			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2484			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2485			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2486		}
2487		mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2488		break;
2489	case 1:
2490		if (skb) {
2491			unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2492			struct flow_keys keys;
2493
2494			/* short-circuit if we already have L4 hash present */
2495			if (skb->l4_hash)
2496				return skb_get_hash_raw(skb) >> 1;
2497
2498			memset(&hash_keys, 0, sizeof(hash_keys));
2499
2500			if (!flkeys) {
2501				skb_flow_dissect_flow_keys(skb, &keys, flag);
2502				flkeys = &keys;
2503			}
2504			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2505			hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2506			hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2507			hash_keys.ports.src = flkeys->ports.src;
2508			hash_keys.ports.dst = flkeys->ports.dst;
2509			hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2510		} else {
2511			memset(&hash_keys, 0, sizeof(hash_keys));
2512			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2513			hash_keys.addrs.v6addrs.src = fl6->saddr;
2514			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2515			hash_keys.ports.src = fl6->fl6_sport;
2516			hash_keys.ports.dst = fl6->fl6_dport;
2517			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2518		}
2519		mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2520		break;
2521	case 2:
2522		memset(&hash_keys, 0, sizeof(hash_keys));
2523		hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2524		if (skb) {
2525			struct flow_keys keys;
2526
2527			if (!flkeys) {
2528				skb_flow_dissect_flow_keys(skb, &keys, 0);
2529				flkeys = &keys;
2530			}
2531
2532			/* Inner can be v4 or v6 */
2533			if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2534				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2535				hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2536				hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2537			} else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2538				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2539				hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2540				hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2541				hash_keys.tags.flow_label = flkeys->tags.flow_label;
2542				hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2543			} else {
2544				/* Same as case 0 */
2545				hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2546				ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2547			}
2548		} else {
2549			/* Same as case 0 */
2550			hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2551			hash_keys.addrs.v6addrs.src = fl6->saddr;
2552			hash_keys.addrs.v6addrs.dst = fl6->daddr;
2553			hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2554			hash_keys.basic.ip_proto = fl6->flowi6_proto;
2555		}
2556		mhash = fib_multipath_hash_from_keys(net, &hash_keys);
2557		break;
2558	case 3:
2559		if (skb)
2560			mhash = rt6_multipath_custom_hash_skb(net, skb);
2561		else
2562			mhash = rt6_multipath_custom_hash_fl6(net, fl6);
2563		break;
2564	}
2565
2566	return mhash >> 1;
2567}
2568
2569/* Called with rcu held */
2570void ip6_route_input(struct sk_buff *skb)
2571{
2572	const struct ipv6hdr *iph = ipv6_hdr(skb);
2573	struct net *net = dev_net(skb->dev);
2574	int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2575	struct ip_tunnel_info *tun_info;
2576	struct flowi6 fl6 = {
2577		.flowi6_iif = skb->dev->ifindex,
2578		.daddr = iph->daddr,
2579		.saddr = iph->saddr,
2580		.flowlabel = ip6_flowinfo(iph),
2581		.flowi6_mark = skb->mark,
2582		.flowi6_proto = iph->nexthdr,
2583	};
2584	struct flow_keys *flkeys = NULL, _flkeys;
2585
2586	tun_info = skb_tunnel_info(skb);
2587	if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2588		fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2589
2590	if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2591		flkeys = &_flkeys;
2592
2593	if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2594		fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2595	skb_dst_drop(skb);
2596	skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2597						      &fl6, skb, flags));
2598}
2599
2600INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net,
2601					     struct fib6_table *table,
2602					     struct flowi6 *fl6,
2603					     const struct sk_buff *skb,
2604					     int flags)
2605{
2606	return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2607}
2608
2609static struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2610						      const struct sock *sk,
2611						      struct flowi6 *fl6,
2612						      int flags)
2613{
2614	bool any_src;
2615
2616	if (ipv6_addr_type(&fl6->daddr) &
2617	    (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2618		struct dst_entry *dst;
2619
2620		/* This function does not take refcnt on the dst */
2621		dst = l3mdev_link_scope_lookup(net, fl6);
2622		if (dst)
2623			return dst;
2624	}
2625
2626	fl6->flowi6_iif = LOOPBACK_IFINDEX;
2627
2628	flags |= RT6_LOOKUP_F_DST_NOREF;
2629	any_src = ipv6_addr_any(&fl6->saddr);
2630	if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2631	    (fl6->flowi6_oif && any_src))
2632		flags |= RT6_LOOKUP_F_IFACE;
2633
2634	if (!any_src)
2635		flags |= RT6_LOOKUP_F_HAS_SADDR;
2636	else if (sk)
2637		flags |= rt6_srcprefs2flags(READ_ONCE(inet6_sk(sk)->srcprefs));
2638
2639	return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2640}
2641
2642struct dst_entry *ip6_route_output_flags(struct net *net,
2643					 const struct sock *sk,
2644					 struct flowi6 *fl6,
2645					 int flags)
2646{
2647	struct dst_entry *dst;
2648	struct rt6_info *rt6;
2649
2650	rcu_read_lock();
2651	dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2652	rt6 = dst_rt6_info(dst);
2653	/* For dst cached in uncached_list, refcnt is already taken. */
2654	if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) {
2655		dst = &net->ipv6.ip6_null_entry->dst;
2656		dst_hold(dst);
2657	}
2658	rcu_read_unlock();
2659
2660	return dst;
2661}
2662EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2663
2664struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2665{
2666	struct rt6_info *rt, *ort = dst_rt6_info(dst_orig);
2667	struct net_device *loopback_dev = net->loopback_dev;
2668	struct dst_entry *new = NULL;
2669
2670	rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev,
2671		       DST_OBSOLETE_DEAD, 0);
2672	if (rt) {
2673		rt6_info_init(rt);
2674		atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
 
 
2675
2676		new = &rt->dst;
2677		new->__use = 1;
2678		new->input = dst_discard;
2679		new->output = dst_discard_out;
2680
2681		dst_copy_metrics(new, &ort->dst);
 
 
 
 
 
 
2682
2683		rt->rt6i_idev = in6_dev_get(loopback_dev);
2684		rt->rt6i_gateway = ort->rt6i_gateway;
2685		rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
 
2686
2687		memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2688#ifdef CONFIG_IPV6_SUBTREES
2689		memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2690#endif
 
 
2691	}
2692
2693	dst_release(dst_orig);
2694	return new ? new : ERR_PTR(-ENOMEM);
2695}
2696
2697/*
2698 *	Destination cache support functions
2699 */
2700
2701static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2702{
2703	u32 rt_cookie = 0;
2704
2705	if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2706		return false;
2707
2708	if (fib6_check_expired(f6i))
2709		return false;
2710
2711	return true;
2712}
2713
2714static struct dst_entry *rt6_check(struct rt6_info *rt,
2715				   struct fib6_info *from,
2716				   u32 cookie)
2717{
2718	u32 rt_cookie = 0;
2719
2720	if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2721	    rt_cookie != cookie)
2722		return NULL;
2723
2724	if (rt6_check_expired(rt))
2725		return NULL;
2726
2727	return &rt->dst;
2728}
2729
2730static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2731					    struct fib6_info *from,
2732					    u32 cookie)
2733{
2734	if (!__rt6_check_expired(rt) &&
2735	    rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2736	    fib6_check(from, cookie))
2737		return &rt->dst;
2738	else
2739		return NULL;
2740}
2741
2742INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst,
2743							u32 cookie)
2744{
2745	struct dst_entry *dst_ret;
2746	struct fib6_info *from;
2747	struct rt6_info *rt;
2748
2749	rt = dst_rt6_info(dst);
2750
2751	if (rt->sernum)
2752		return rt6_is_valid(rt) ? dst : NULL;
2753
2754	rcu_read_lock();
2755
2756	/* All IPV6 dsts are created with ->obsolete set to the value
2757	 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2758	 * into this function always.
2759	 */
 
 
2760
2761	from = rcu_dereference(rt->from);
 
2762
2763	if (from && (rt->rt6i_flags & RTF_PCPU ||
2764	    unlikely(!list_empty(&rt->dst.rt_uncached))))
2765		dst_ret = rt6_dst_from_check(rt, from, cookie);
2766	else
2767		dst_ret = rt6_check(rt, from, cookie);
2768
2769	rcu_read_unlock();
2770
2771	return dst_ret;
2772}
2773EXPORT_INDIRECT_CALLABLE(ip6_dst_check);
2774
2775static void ip6_negative_advice(struct sock *sk,
2776				struct dst_entry *dst)
2777{
2778	struct rt6_info *rt = dst_rt6_info(dst);
2779
2780	if (rt->rt6i_flags & RTF_CACHE) {
2781		rcu_read_lock();
2782		if (rt6_check_expired(rt)) {
2783			/* rt/dst can not be destroyed yet,
2784			 * because of rcu_read_lock()
2785			 */
2786			sk_dst_reset(sk);
2787			rt6_remove_exception_rt(rt);
 
2788		}
2789		rcu_read_unlock();
2790		return;
2791	}
2792	sk_dst_reset(sk);
2793}
2794
2795static void ip6_link_failure(struct sk_buff *skb)
2796{
2797	struct rt6_info *rt;
2798
2799	icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2800
2801	rt = dst_rt6_info(skb_dst(skb));
2802	if (rt) {
2803		rcu_read_lock();
2804		if (rt->rt6i_flags & RTF_CACHE) {
2805			rt6_remove_exception_rt(rt);
2806		} else {
2807			struct fib6_info *from;
2808			struct fib6_node *fn;
2809
2810			from = rcu_dereference(rt->from);
2811			if (from) {
2812				fn = rcu_dereference(from->fib6_node);
2813				if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2814					WRITE_ONCE(fn->fn_sernum, -1);
2815			}
2816		}
2817		rcu_read_unlock();
2818	}
2819}
2820
2821static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2822{
2823	if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2824		struct fib6_info *from;
2825
2826		rcu_read_lock();
2827		from = rcu_dereference(rt0->from);
2828		if (from)
2829			rt0->dst.expires = from->expires;
2830		rcu_read_unlock();
2831	}
2832
2833	dst_set_expires(&rt0->dst, timeout);
2834	rt0->rt6i_flags |= RTF_EXPIRES;
2835}
2836
2837static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2838{
2839	struct net *net = dev_net(rt->dst.dev);
2840
2841	dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2842	rt->rt6i_flags |= RTF_MODIFIED;
2843	rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2844}
2845
2846static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2847{
2848	return !(rt->rt6i_flags & RTF_CACHE) &&
2849		(rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2850}
2851
2852static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2853				 const struct ipv6hdr *iph, u32 mtu,
2854				 bool confirm_neigh)
2855{
2856	const struct in6_addr *daddr, *saddr;
2857	struct rt6_info *rt6 = dst_rt6_info(dst);
2858
2859	/* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU)
2860	 * IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it.
2861	 * [see also comment in rt6_mtu_change_route()]
2862	 */
2863
2864	if (iph) {
2865		daddr = &iph->daddr;
2866		saddr = &iph->saddr;
2867	} else if (sk) {
2868		daddr = &sk->sk_v6_daddr;
2869		saddr = &inet6_sk(sk)->saddr;
2870	} else {
2871		daddr = NULL;
2872		saddr = NULL;
2873	}
2874
2875	if (confirm_neigh)
2876		dst_confirm_neigh(dst, daddr);
2877
2878	if (mtu < IPV6_MIN_MTU)
2879		return;
2880	if (mtu >= dst_mtu(dst))
2881		return;
2882
2883	if (!rt6_cache_allowed_for_pmtu(rt6)) {
2884		rt6_do_update_pmtu(rt6, mtu);
2885		/* update rt6_ex->stamp for cache */
2886		if (rt6->rt6i_flags & RTF_CACHE)
2887			rt6_update_exception_stamp_rt(rt6);
2888	} else if (daddr) {
2889		struct fib6_result res = {};
2890		struct rt6_info *nrt6;
2891
2892		rcu_read_lock();
2893		res.f6i = rcu_dereference(rt6->from);
2894		if (!res.f6i)
2895			goto out_unlock;
2896
2897		res.fib6_flags = res.f6i->fib6_flags;
2898		res.fib6_type = res.f6i->fib6_type;
2899
2900		if (res.f6i->nh) {
2901			struct fib6_nh_match_arg arg = {
2902				.dev = dst->dev,
2903				.gw = &rt6->rt6i_gateway,
2904			};
2905
2906			nexthop_for_each_fib6_nh(res.f6i->nh,
2907						 fib6_nh_find_match, &arg);
2908
2909			/* fib6_info uses a nexthop that does not have fib6_nh
2910			 * using the dst->dev + gw. Should be impossible.
2911			 */
2912			if (!arg.match)
2913				goto out_unlock;
2914
2915			res.nh = arg.match;
2916		} else {
2917			res.nh = res.f6i->fib6_nh;
2918		}
2919
2920		nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2921		if (nrt6) {
2922			rt6_do_update_pmtu(nrt6, mtu);
2923			if (rt6_insert_exception(nrt6, &res))
2924				dst_release_immediate(&nrt6->dst);
 
 
 
 
 
2925		}
2926out_unlock:
2927		rcu_read_unlock();
2928	}
2929}
2930
2931static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2932			       struct sk_buff *skb, u32 mtu,
2933			       bool confirm_neigh)
2934{
2935	__ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu,
2936			     confirm_neigh);
2937}
2938
2939void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2940		     int oif, u32 mark, kuid_t uid)
2941{
2942	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2943	struct dst_entry *dst;
2944	struct flowi6 fl6 = {
2945		.flowi6_oif = oif,
2946		.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2947		.daddr = iph->daddr,
2948		.saddr = iph->saddr,
2949		.flowlabel = ip6_flowinfo(iph),
2950		.flowi6_uid = uid,
2951	};
2952
2953	dst = ip6_route_output(net, NULL, &fl6);
2954	if (!dst->error)
2955		__ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true);
2956	dst_release(dst);
2957}
2958EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2959
2960void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2961{
2962	int oif = sk->sk_bound_dev_if;
2963	struct dst_entry *dst;
2964
2965	if (!oif && skb->dev)
2966		oif = l3mdev_master_ifindex(skb->dev);
2967
2968	ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark),
2969			sk->sk_uid);
2970
2971	dst = __sk_dst_get(sk);
2972	if (!dst || !dst->obsolete ||
2973	    dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2974		return;
2975
2976	bh_lock_sock(sk);
2977	if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2978		ip6_datagram_dst_update(sk, false);
2979	bh_unlock_sock(sk);
2980}
2981EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2982
2983void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2984			   const struct flowi6 *fl6)
2985{
2986#ifdef CONFIG_IPV6_SUBTREES
2987	struct ipv6_pinfo *np = inet6_sk(sk);
2988#endif
2989
2990	ip6_dst_store(sk, dst,
2991		      ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2992		      &sk->sk_v6_daddr : NULL,
2993#ifdef CONFIG_IPV6_SUBTREES
2994		      ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2995		      &np->saddr :
2996#endif
2997		      NULL);
2998}
2999
3000static bool ip6_redirect_nh_match(const struct fib6_result *res,
3001				  struct flowi6 *fl6,
3002				  const struct in6_addr *gw,
3003				  struct rt6_info **ret)
3004{
3005	const struct fib6_nh *nh = res->nh;
3006
3007	if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
3008	    fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
3009		return false;
3010
3011	/* rt_cache's gateway might be different from its 'parent'
3012	 * in the case of an ip redirect.
3013	 * So we keep searching in the exception table if the gateway
3014	 * is different.
3015	 */
3016	if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
3017		struct rt6_info *rt_cache;
3018
3019		rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
3020		if (rt_cache &&
3021		    ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
3022			*ret = rt_cache;
3023			return true;
3024		}
3025		return false;
3026	}
3027	return true;
3028}
3029
3030struct fib6_nh_rd_arg {
3031	struct fib6_result	*res;
3032	struct flowi6		*fl6;
3033	const struct in6_addr	*gw;
3034	struct rt6_info		**ret;
3035};
3036
3037static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
3038{
3039	struct fib6_nh_rd_arg *arg = _arg;
3040
3041	arg->res->nh = nh;
3042	return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
3043}
3044
3045/* Handle redirects */
3046struct ip6rd_flowi {
3047	struct flowi6 fl6;
3048	struct in6_addr gateway;
3049};
3050
3051INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net,
3052					     struct fib6_table *table,
3053					     struct flowi6 *fl6,
3054					     const struct sk_buff *skb,
3055					     int flags)
3056{
3057	struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
3058	struct rt6_info *ret = NULL;
3059	struct fib6_result res = {};
3060	struct fib6_nh_rd_arg arg = {
3061		.res = &res,
3062		.fl6 = fl6,
3063		.gw  = &rdfl->gateway,
3064		.ret = &ret
3065	};
3066	struct fib6_info *rt;
3067	struct fib6_node *fn;
3068
3069	/* Get the "current" route for this destination and
3070	 * check if the redirect has come from appropriate router.
3071	 *
3072	 * RFC 4861 specifies that redirects should only be
3073	 * accepted if they come from the nexthop to the target.
3074	 * Due to the way the routes are chosen, this notion
3075	 * is a bit fuzzy and one might need to check all possible
3076	 * routes.
3077	 */
3078
3079	rcu_read_lock();
3080	fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
3081restart:
3082	for_each_fib6_node_rt_rcu(fn) {
3083		res.f6i = rt;
3084		if (fib6_check_expired(rt))
3085			continue;
3086		if (rt->fib6_flags & RTF_REJECT)
3087			break;
3088		if (unlikely(rt->nh)) {
3089			if (nexthop_is_blackhole(rt->nh))
3090				continue;
3091			/* on match, res->nh is filled in and potentially ret */
3092			if (nexthop_for_each_fib6_nh(rt->nh,
3093						     fib6_nh_redirect_match,
3094						     &arg))
3095				goto out;
3096		} else {
3097			res.nh = rt->fib6_nh;
3098			if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
3099						  &ret))
3100				goto out;
3101		}
3102	}
3103
3104	if (!rt)
3105		rt = net->ipv6.fib6_null_entry;
3106	else if (rt->fib6_flags & RTF_REJECT) {
3107		ret = net->ipv6.ip6_null_entry;
3108		goto out;
3109	}
3110
3111	if (rt == net->ipv6.fib6_null_entry) {
3112		fn = fib6_backtrack(fn, &fl6->saddr);
3113		if (fn)
3114			goto restart;
3115	}
3116
3117	res.f6i = rt;
3118	res.nh = rt->fib6_nh;
3119out:
3120	if (ret) {
3121		ip6_hold_safe(net, &ret);
3122	} else {
3123		res.fib6_flags = res.f6i->fib6_flags;
3124		res.fib6_type = res.f6i->fib6_type;
3125		ret = ip6_create_rt_rcu(&res);
3126	}
3127
3128	rcu_read_unlock();
3129
3130	trace_fib6_table_lookup(net, &res, table, fl6);
3131	return ret;
3132};
3133
3134static struct dst_entry *ip6_route_redirect(struct net *net,
3135					    const struct flowi6 *fl6,
3136					    const struct sk_buff *skb,
3137					    const struct in6_addr *gateway)
3138{
3139	int flags = RT6_LOOKUP_F_HAS_SADDR;
3140	struct ip6rd_flowi rdfl;
3141
3142	rdfl.fl6 = *fl6;
3143	rdfl.gateway = *gateway;
3144
3145	return fib6_rule_lookup(net, &rdfl.fl6, skb,
3146				flags, __ip6_route_redirect);
3147}
3148
3149void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
3150		  kuid_t uid)
3151{
3152	const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
3153	struct dst_entry *dst;
3154	struct flowi6 fl6 = {
3155		.flowi6_iif = LOOPBACK_IFINDEX,
3156		.flowi6_oif = oif,
3157		.flowi6_mark = mark,
3158		.daddr = iph->daddr,
3159		.saddr = iph->saddr,
3160		.flowlabel = ip6_flowinfo(iph),
3161		.flowi6_uid = uid,
3162	};
3163
3164	dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3165	rt6_do_redirect(dst, NULL, skb);
3166	dst_release(dst);
3167}
3168EXPORT_SYMBOL_GPL(ip6_redirect);
3169
3170void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
 
3171{
3172	const struct ipv6hdr *iph = ipv6_hdr(skb);
3173	const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3174	struct dst_entry *dst;
3175	struct flowi6 fl6 = {
3176		.flowi6_iif = LOOPBACK_IFINDEX,
3177		.flowi6_oif = oif,
3178		.daddr = msg->dest,
3179		.saddr = iph->daddr,
3180		.flowi6_uid = sock_net_uid(net, NULL),
3181	};
 
3182
3183	dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3184	rt6_do_redirect(dst, NULL, skb);
3185	dst_release(dst);
3186}
3187
3188void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3189{
3190	ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if,
3191		     READ_ONCE(sk->sk_mark), sk->sk_uid);
3192}
3193EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3194
3195static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3196{
3197	struct net_device *dev = dst->dev;
3198	unsigned int mtu = dst_mtu(dst);
3199	struct net *net;
3200
3201	mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3202
3203	rcu_read_lock();
3204
3205	net = dev_net_rcu(dev);
3206	if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3207		mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3208
3209	rcu_read_unlock();
3210
3211	/*
3212	 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3213	 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3214	 * IPV6_MAXPLEN is also valid and means: "any MSS,
3215	 * rely only on pmtu discovery"
3216	 */
3217	if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3218		mtu = IPV6_MAXPLEN;
3219	return mtu;
3220}
3221
3222INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst)
3223{
3224	return ip6_dst_mtu_maybe_forward(dst, false);
3225}
3226EXPORT_INDIRECT_CALLABLE(ip6_mtu);
3227
3228/* MTU selection:
3229 * 1. mtu on route is locked - use it
3230 * 2. mtu from nexthop exception
3231 * 3. mtu from egress device
3232 *
3233 * based on ip6_dst_mtu_forward and exception logic of
3234 * rt6_find_cached_rt; called with rcu_read_lock
3235 */
3236u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3237		      const struct in6_addr *daddr,
3238		      const struct in6_addr *saddr)
3239{
3240	const struct fib6_nh *nh = res->nh;
3241	struct fib6_info *f6i = res->f6i;
3242	struct inet6_dev *idev;
3243	struct rt6_info *rt;
3244	u32 mtu = 0;
3245
3246	if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3247		mtu = f6i->fib6_pmtu;
3248		if (mtu)
3249			goto out;
3250	}
3251
3252	rt = rt6_find_cached_rt(res, daddr, saddr);
3253	if (unlikely(rt)) {
3254		mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3255	} else {
3256		struct net_device *dev = nh->fib_nh_dev;
3257
3258		mtu = IPV6_MIN_MTU;
3259		idev = __in6_dev_get(dev);
3260		if (idev)
3261			mtu = max_t(u32, mtu, READ_ONCE(idev->cnf.mtu6));
3262	}
3263
3264	mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3265out:
3266	return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3267}
3268
 
 
 
3269struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3270				  struct flowi6 *fl6)
3271{
3272	struct dst_entry *dst;
3273	struct rt6_info *rt;
3274	struct inet6_dev *idev = in6_dev_get(dev);
3275	struct net *net = dev_net(dev);
3276
3277	if (unlikely(!idev))
3278		return ERR_PTR(-ENODEV);
3279
3280	rt = ip6_dst_alloc(net, dev, 0);
3281	if (unlikely(!rt)) {
3282		in6_dev_put(idev);
3283		dst = ERR_PTR(-ENOMEM);
3284		goto out;
3285	}
3286
3287	rt->dst.input = ip6_input;
3288	rt->dst.output  = ip6_output;
 
3289	rt->rt6i_gateway  = fl6->daddr;
3290	rt->rt6i_dst.addr = fl6->daddr;
3291	rt->rt6i_dst.plen = 128;
3292	rt->rt6i_idev     = idev;
3293	dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3294
3295	/* Add this dst into uncached_list so that rt6_disable_ip() can
3296	 * do proper release of the net_device
3297	 */
3298	rt6_uncached_list_add(rt);
 
 
3299
3300	dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3301
3302out:
3303	return dst;
3304}
3305
3306static void ip6_dst_gc(struct dst_ops *ops)
3307{
3308	struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3309	int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3310	int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3311	int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3312	unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3313	unsigned int val;
3314	int entries;
3315
3316	if (time_after(rt_last_gc + rt_min_interval, jiffies))
3317		goto out;
3318
3319	fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true);
3320	entries = dst_entries_get_slow(ops);
3321	if (entries < ops->gc_thresh)
3322		atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1);
3323out:
3324	val = atomic_read(&net->ipv6.ip6_rt_gc_expire);
3325	atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity));
3326}
3327
3328static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3329			       const struct in6_addr *gw_addr, u32 tbid,
3330			       int flags, struct fib6_result *res)
3331{
3332	struct flowi6 fl6 = {
3333		.flowi6_oif = cfg->fc_ifindex,
3334		.daddr = *gw_addr,
3335		.saddr = cfg->fc_prefsrc,
3336	};
3337	struct fib6_table *table;
3338	int err;
3339
3340	table = fib6_get_table(net, tbid);
3341	if (!table)
3342		return -EINVAL;
3343
3344	if (!ipv6_addr_any(&cfg->fc_prefsrc))
3345		flags |= RT6_LOOKUP_F_HAS_SADDR;
3346
3347	flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3348
3349	err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3350	if (!err && res->f6i != net->ipv6.fib6_null_entry)
3351		fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3352				 cfg->fc_ifindex != 0, NULL, flags);
3353
3354	return err;
3355}
3356
3357static int ip6_route_check_nh_onlink(struct net *net,
3358				     struct fib6_config *cfg,
3359				     const struct net_device *dev,
3360				     struct netlink_ext_ack *extack)
3361{
3362	u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3363	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3364	struct fib6_result res = {};
3365	int err;
3366
3367	err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3368	if (!err && !(res.fib6_flags & RTF_REJECT) &&
3369	    /* ignore match if it is the default route */
3370	    !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3371	    (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3372		NL_SET_ERR_MSG(extack,
3373			       "Nexthop has invalid gateway or device mismatch");
3374		err = -EINVAL;
3375	}
3376
3377	return err;
3378}
3379
3380static int ip6_route_check_nh(struct net *net,
3381			      struct fib6_config *cfg,
3382			      struct net_device **_dev,
3383			      netdevice_tracker *dev_tracker,
3384			      struct inet6_dev **idev)
3385{
3386	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3387	struct net_device *dev = _dev ? *_dev : NULL;
3388	int flags = RT6_LOOKUP_F_IFACE;
3389	struct fib6_result res = {};
3390	int err = -EHOSTUNREACH;
3391
3392	if (cfg->fc_table) {
3393		err = ip6_nh_lookup_table(net, cfg, gw_addr,
3394					  cfg->fc_table, flags, &res);
3395		/* gw_addr can not require a gateway or resolve to a reject
3396		 * route. If a device is given, it must match the result.
3397		 */
3398		if (err || res.fib6_flags & RTF_REJECT ||
3399		    res.nh->fib_nh_gw_family ||
3400		    (dev && dev != res.nh->fib_nh_dev))
3401			err = -EHOSTUNREACH;
3402	}
3403
3404	if (err < 0) {
3405		struct flowi6 fl6 = {
3406			.flowi6_oif = cfg->fc_ifindex,
3407			.daddr = *gw_addr,
3408		};
3409
3410		err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3411		if (err || res.fib6_flags & RTF_REJECT ||
3412		    res.nh->fib_nh_gw_family)
3413			err = -EHOSTUNREACH;
3414
3415		if (err)
3416			return err;
3417
3418		fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3419				 cfg->fc_ifindex != 0, NULL, flags);
3420	}
3421
3422	err = 0;
3423	if (dev) {
3424		if (dev != res.nh->fib_nh_dev)
3425			err = -EHOSTUNREACH;
3426	} else {
3427		*_dev = dev = res.nh->fib_nh_dev;
3428		netdev_hold(dev, dev_tracker, GFP_ATOMIC);
3429		*idev = in6_dev_get(dev);
3430	}
3431
3432	return err;
3433}
3434
3435static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3436			   struct net_device **_dev,
3437			   netdevice_tracker *dev_tracker,
3438			   struct inet6_dev **idev,
3439			   struct netlink_ext_ack *extack)
3440{
3441	const struct in6_addr *gw_addr = &cfg->fc_gateway;
3442	int gwa_type = ipv6_addr_type(gw_addr);
3443	bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3444	const struct net_device *dev = *_dev;
3445	bool need_addr_check = !dev;
3446	int err = -EINVAL;
3447
3448	/* if gw_addr is local we will fail to detect this in case
3449	 * address is still TENTATIVE (DAD in progress). rt6_lookup()
3450	 * will return already-added prefix route via interface that
3451	 * prefix route was assigned to, which might be non-loopback.
3452	 */
3453	if (dev &&
3454	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3455		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3456		goto out;
3457	}
3458
3459	if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3460		/* IPv6 strictly inhibits using not link-local
3461		 * addresses as nexthop address.
3462		 * Otherwise, router will not able to send redirects.
3463		 * It is very good, but in some (rare!) circumstances
3464		 * (SIT, PtP, NBMA NOARP links) it is handy to allow
3465		 * some exceptions. --ANK
3466		 * We allow IPv4-mapped nexthops to support RFC4798-type
3467		 * addressing
3468		 */
3469		if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3470			NL_SET_ERR_MSG(extack, "Invalid gateway address");
3471			goto out;
3472		}
3473
3474		rcu_read_lock();
3475
3476		if (cfg->fc_flags & RTNH_F_ONLINK)
3477			err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3478		else
3479			err = ip6_route_check_nh(net, cfg, _dev, dev_tracker,
3480						 idev);
3481
3482		rcu_read_unlock();
3483
3484		if (err)
3485			goto out;
3486	}
 
 
3487
3488	/* reload in case device was changed */
3489	dev = *_dev;
 
 
 
 
 
 
 
3490
3491	err = -EINVAL;
3492	if (!dev) {
3493		NL_SET_ERR_MSG(extack, "Egress device not specified");
3494		goto out;
3495	} else if (dev->flags & IFF_LOOPBACK) {
3496		NL_SET_ERR_MSG(extack,
3497			       "Egress device can not be loopback device for this route");
3498		goto out;
3499	}
3500
3501	/* if we did not check gw_addr above, do so now that the
3502	 * egress device has been resolved.
3503	 */
3504	if (need_addr_check &&
3505	    ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3506		NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3507		goto out;
3508	}
3509
3510	err = 0;
3511out:
3512	return err;
 
3513}
3514
3515static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3516{
3517	if ((flags & RTF_REJECT) ||
3518	    (dev && (dev->flags & IFF_LOOPBACK) &&
3519	     !(addr_type & IPV6_ADDR_LOOPBACK) &&
3520	     !(flags & (RTF_ANYCAST | RTF_LOCAL))))
3521		return true;
3522
3523	return false;
3524}
3525
3526int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3527		 struct fib6_config *cfg, gfp_t gfp_flags,
3528		 struct netlink_ext_ack *extack)
3529{
3530	netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker;
 
 
3531	struct net_device *dev = NULL;
3532	struct inet6_dev *idev = NULL;
 
3533	int addr_type;
3534	int err;
3535
3536	fib6_nh->fib_nh_family = AF_INET6;
3537#ifdef CONFIG_IPV6_ROUTER_PREF
3538	fib6_nh->last_probe = jiffies;
 
 
3539#endif
3540	if (cfg->fc_is_fdb) {
3541		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3542		fib6_nh->fib_nh_gw_family = AF_INET6;
3543		return 0;
3544	}
3545
3546	err = -ENODEV;
3547	if (cfg->fc_ifindex) {
3548		dev = netdev_get_by_index(net, cfg->fc_ifindex,
3549					  dev_tracker, gfp_flags);
3550		if (!dev)
3551			goto out;
3552		idev = in6_dev_get(dev);
3553		if (!idev)
3554			goto out;
3555	}
3556
3557	if (cfg->fc_flags & RTNH_F_ONLINK) {
3558		if (!dev) {
3559			NL_SET_ERR_MSG(extack,
3560				       "Nexthop device required for onlink");
3561			goto out;
3562		}
3563
3564		if (!(dev->flags & IFF_UP)) {
3565			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3566			err = -ENETDOWN;
3567			goto out;
3568		}
3569
3570		fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3571	}
3572
3573	fib6_nh->fib_nh_weight = 1;
3574
3575	/* We cannot add true routes via loopback here,
3576	 * they would result in kernel looping; promote them to reject routes
3577	 */
3578	addr_type = ipv6_addr_type(&cfg->fc_dst);
3579	if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3580		/* hold loopback dev/idev if we haven't done so. */
3581		if (dev != net->loopback_dev) {
3582			if (dev) {
3583				netdev_put(dev, dev_tracker);
3584				in6_dev_put(idev);
3585			}
3586			dev = net->loopback_dev;
3587			netdev_hold(dev, dev_tracker, gfp_flags);
3588			idev = in6_dev_get(dev);
3589			if (!idev) {
3590				err = -ENODEV;
3591				goto out;
3592			}
3593		}
3594		goto pcpu_alloc;
3595	}
3596
3597	if (cfg->fc_flags & RTF_GATEWAY) {
3598		err = ip6_validate_gw(net, cfg, &dev, dev_tracker,
3599				      &idev, extack);
3600		if (err)
3601			goto out;
3602
3603		fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3604		fib6_nh->fib_nh_gw_family = AF_INET6;
3605	}
3606
3607	err = -ENODEV;
3608	if (!dev)
3609		goto out;
3610
3611	if (!idev || idev->cnf.disable_ipv6) {
3612		NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3613		err = -EACCES;
3614		goto out;
3615	}
3616
3617	if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3618		NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3619		err = -ENETDOWN;
3620		goto out;
3621	}
3622
3623	if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3624	    !netif_carrier_ok(dev))
3625		fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3626
3627	err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap,
3628				 cfg->fc_encap_type, cfg, gfp_flags, extack);
3629	if (err)
3630		goto out;
3631
3632pcpu_alloc:
3633	fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3634	if (!fib6_nh->rt6i_pcpu) {
3635		err = -ENOMEM;
3636		goto out;
3637	}
3638
3639	fib6_nh->fib_nh_dev = dev;
3640	fib6_nh->fib_nh_oif = dev->ifindex;
3641	err = 0;
3642out:
3643	if (idev)
3644		in6_dev_put(idev);
3645
3646	if (err) {
3647		lwtstate_put(fib6_nh->fib_nh_lws);
3648		fib6_nh->fib_nh_lws = NULL;
3649		netdev_put(dev, dev_tracker);
3650	}
3651
3652	return err;
3653}
3654
3655void fib6_nh_release(struct fib6_nh *fib6_nh)
3656{
3657	struct rt6_exception_bucket *bucket;
3658
3659	rcu_read_lock();
3660
3661	fib6_nh_flush_exceptions(fib6_nh, NULL);
3662	bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3663	if (bucket) {
3664		rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3665		kfree(bucket);
3666	}
3667
3668	rcu_read_unlock();
3669
3670	fib6_nh_release_dsts(fib6_nh);
3671	free_percpu(fib6_nh->rt6i_pcpu);
3672
3673	fib_nh_common_release(&fib6_nh->nh_common);
3674}
3675
3676void fib6_nh_release_dsts(struct fib6_nh *fib6_nh)
3677{
3678	int cpu;
3679
3680	if (!fib6_nh->rt6i_pcpu)
3681		return;
3682
3683	for_each_possible_cpu(cpu) {
3684		struct rt6_info *pcpu_rt, **ppcpu_rt;
3685
3686		ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3687		pcpu_rt = xchg(ppcpu_rt, NULL);
3688		if (pcpu_rt) {
3689			dst_dev_put(&pcpu_rt->dst);
3690			dst_release(&pcpu_rt->dst);
3691		}
3692	}
3693}
3694
3695static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3696					      gfp_t gfp_flags,
3697					      struct netlink_ext_ack *extack)
3698{
3699	struct net *net = cfg->fc_nlinfo.nl_net;
3700	struct fib6_info *rt = NULL;
3701	struct nexthop *nh = NULL;
3702	struct fib6_table *table;
3703	struct fib6_nh *fib6_nh;
3704	int err = -EINVAL;
3705	int addr_type;
3706
3707	/* RTF_PCPU is an internal flag; can not be set by userspace */
3708	if (cfg->fc_flags & RTF_PCPU) {
3709		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3710		goto out;
3711	}
3712
3713	/* RTF_CACHE is an internal flag; can not be set by userspace */
3714	if (cfg->fc_flags & RTF_CACHE) {
3715		NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3716		goto out;
3717	}
3718
3719	if (cfg->fc_type > RTN_MAX) {
3720		NL_SET_ERR_MSG(extack, "Invalid route type");
3721		goto out;
3722	}
3723
3724	if (cfg->fc_dst_len > 128) {
3725		NL_SET_ERR_MSG(extack, "Invalid prefix length");
3726		goto out;
3727	}
3728	if (cfg->fc_src_len > 128) {
3729		NL_SET_ERR_MSG(extack, "Invalid source address length");
3730		goto out;
3731	}
3732#ifndef CONFIG_IPV6_SUBTREES
3733	if (cfg->fc_src_len) {
3734		NL_SET_ERR_MSG(extack,
3735			       "Specifying source address requires IPV6_SUBTREES to be enabled");
3736		goto out;
3737	}
3738#endif
3739	if (cfg->fc_nh_id) {
3740		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3741		if (!nh) {
3742			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3743			goto out;
3744		}
3745		err = fib6_check_nexthop(nh, cfg, extack);
3746		if (err)
3747			goto out;
3748	}
3749
3750	err = -ENOBUFS;
3751	if (cfg->fc_nlinfo.nlh &&
3752	    !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3753		table = fib6_get_table(net, cfg->fc_table);
3754		if (!table) {
3755			pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3756			table = fib6_new_table(net, cfg->fc_table);
3757		}
3758	} else {
3759		table = fib6_new_table(net, cfg->fc_table);
3760	}
3761
3762	if (!table)
3763		goto out;
3764
3765	err = -ENOMEM;
3766	rt = fib6_info_alloc(gfp_flags, !nh);
3767	if (!rt)
 
3768		goto out;
3769
3770	rt->fib6_metrics = ip_fib_metrics_init(cfg->fc_mx, cfg->fc_mx_len,
3771					       extack);
3772	if (IS_ERR(rt->fib6_metrics)) {
3773		err = PTR_ERR(rt->fib6_metrics);
3774		/* Do not leave garbage there. */
3775		rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3776		goto out_free;
3777	}
3778
3779	if (cfg->fc_flags & RTF_ADDRCONF)
3780		rt->dst_nocount = true;
3781
3782	if (cfg->fc_flags & RTF_EXPIRES)
3783		fib6_set_expires(rt, jiffies +
3784				clock_t_to_jiffies(cfg->fc_expires));
 
 
3785
3786	if (cfg->fc_protocol == RTPROT_UNSPEC)
3787		cfg->fc_protocol = RTPROT_BOOT;
3788	rt->fib6_protocol = cfg->fc_protocol;
3789
3790	rt->fib6_table = table;
3791	rt->fib6_metric = cfg->fc_metric;
3792	rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3793	rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3794
3795	ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3796	rt->fib6_dst.plen = cfg->fc_dst_len;
 
 
 
 
 
 
 
 
 
 
 
 
 
3797
3798#ifdef CONFIG_IPV6_SUBTREES
3799	ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3800	rt->fib6_src.plen = cfg->fc_src_len;
3801#endif
3802	if (nh) {
3803		if (rt->fib6_src.plen) {
3804			NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3805			goto out_free;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3806		}
3807		if (!nexthop_get(nh)) {
3808			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3809			goto out_free;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3810		}
3811		rt->nh = nh;
3812		fib6_nh = nexthop_fib6_nh(rt->nh);
3813	} else {
3814		err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3815		if (err)
3816			goto out;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3817
3818		fib6_nh = rt->fib6_nh;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3819
3820		/* We cannot add true routes via loopback here, they would
3821		 * result in kernel looping; promote them to reject routes
3822		 */
3823		addr_type = ipv6_addr_type(&cfg->fc_dst);
3824		if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3825				   addr_type))
3826			rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3827	}
3828
 
 
 
 
3829	if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3830		struct net_device *dev = fib6_nh->fib_nh_dev;
3831
3832		if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3833			NL_SET_ERR_MSG(extack, "Invalid source address");
3834			err = -EINVAL;
3835			goto out;
3836		}
3837		rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3838		rt->fib6_prefsrc.plen = 128;
3839	} else
3840		rt->fib6_prefsrc.plen = 0;
3841
3842	return rt;
3843out:
3844	fib6_info_release(rt);
3845	return ERR_PTR(err);
3846out_free:
3847	ip_fib_metrics_put(rt->fib6_metrics);
3848	kfree(rt);
3849	return ERR_PTR(err);
3850}
3851
3852int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3853		  struct netlink_ext_ack *extack)
3854{
3855	struct fib6_info *rt;
3856	int err;
3857
3858	rt = ip6_route_info_create(cfg, gfp_flags, extack);
3859	if (IS_ERR(rt))
3860		return PTR_ERR(rt);
3861
3862	err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3863	fib6_info_release(rt);
3864
 
 
 
 
 
 
 
3865	return err;
3866}
3867
3868static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3869{
3870	struct net *net = info->nl_net;
3871	struct fib6_table *table;
3872	int err;
3873
3874	if (rt == net->ipv6.fib6_null_entry) {
3875		err = -ENOENT;
3876		goto out;
3877	}
3878
3879	table = rt->fib6_table;
3880	spin_lock_bh(&table->tb6_lock);
3881	err = fib6_del(rt, info);
3882	spin_unlock_bh(&table->tb6_lock);
3883
3884out:
3885	fib6_info_release(rt);
3886	return err;
3887}
3888
3889int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
3890{
3891	struct nl_info info = {
3892		.nl_net = net,
3893		.skip_notify = skip_notify
3894	};
3895
3896	return __ip6_del_rt(rt, &info);
3897}
3898
3899static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3900{
3901	struct nl_info *info = &cfg->fc_nlinfo;
3902	struct net *net = info->nl_net;
3903	struct sk_buff *skb = NULL;
3904	struct fib6_table *table;
3905	int err = -ENOENT;
3906
3907	if (rt == net->ipv6.fib6_null_entry)
3908		goto out_put;
3909	table = rt->fib6_table;
3910	spin_lock_bh(&table->tb6_lock);
3911
3912	if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3913		struct fib6_info *sibling, *next_sibling;
3914		struct fib6_node *fn;
3915
3916		/* prefer to send a single notification with all hops */
3917		skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3918		if (skb) {
3919			u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3920
3921			if (rt6_fill_node(net, skb, rt, NULL,
3922					  NULL, NULL, 0, RTM_DELROUTE,
3923					  info->portid, seq, 0) < 0) {
3924				kfree_skb(skb);
3925				skb = NULL;
3926			} else
3927				info->skip_notify = 1;
3928		}
3929
3930		/* 'rt' points to the first sibling route. If it is not the
3931		 * leaf, then we do not need to send a notification. Otherwise,
3932		 * we need to check if the last sibling has a next route or not
3933		 * and emit a replace or delete notification, respectively.
3934		 */
3935		info->skip_notify_kernel = 1;
3936		fn = rcu_dereference_protected(rt->fib6_node,
3937					    lockdep_is_held(&table->tb6_lock));
3938		if (rcu_access_pointer(fn->leaf) == rt) {
3939			struct fib6_info *last_sibling, *replace_rt;
3940
3941			last_sibling = list_last_entry(&rt->fib6_siblings,
3942						       struct fib6_info,
3943						       fib6_siblings);
3944			replace_rt = rcu_dereference_protected(
3945					    last_sibling->fib6_next,
3946					    lockdep_is_held(&table->tb6_lock));
3947			if (replace_rt)
3948				call_fib6_entry_notifiers_replace(net,
3949								  replace_rt);
3950			else
3951				call_fib6_multipath_entry_notifiers(net,
3952						       FIB_EVENT_ENTRY_DEL,
3953						       rt, rt->fib6_nsiblings,
3954						       NULL);
3955		}
3956		list_for_each_entry_safe(sibling, next_sibling,
3957					 &rt->fib6_siblings,
3958					 fib6_siblings) {
3959			err = fib6_del(sibling, info);
3960			if (err)
3961				goto out_unlock;
3962		}
3963	}
3964
3965	err = fib6_del(rt, info);
3966out_unlock:
3967	spin_unlock_bh(&table->tb6_lock);
3968out_put:
3969	fib6_info_release(rt);
3970
3971	if (skb) {
3972		rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3973			    info->nlh, gfp_any());
3974	}
3975	return err;
3976}
3977
3978static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3979{
3980	int rc = -ESRCH;
3981
3982	if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3983		goto out;
3984
3985	if (cfg->fc_flags & RTF_GATEWAY &&
3986	    !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3987		goto out;
3988
3989	rc = rt6_remove_exception_rt(rt);
3990out:
3991	return rc;
3992}
3993
3994static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3995			     struct fib6_nh *nh)
3996{
3997	struct fib6_result res = {
3998		.f6i = rt,
3999		.nh = nh,
4000	};
4001	struct rt6_info *rt_cache;
4002
4003	rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
4004	if (rt_cache)
4005		return __ip6_del_cached_rt(rt_cache, cfg);
4006
4007	return 0;
4008}
4009
4010struct fib6_nh_del_cached_rt_arg {
4011	struct fib6_config *cfg;
4012	struct fib6_info *f6i;
4013};
4014
4015static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
4016{
4017	struct fib6_nh_del_cached_rt_arg *arg = _arg;
4018	int rc;
4019
4020	rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
4021	return rc != -ESRCH ? rc : 0;
4022}
4023
4024static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
4025{
4026	struct fib6_nh_del_cached_rt_arg arg = {
4027		.cfg = cfg,
4028		.f6i = f6i
4029	};
4030
4031	return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
4032}
4033
4034static int ip6_route_del(struct fib6_config *cfg,
4035			 struct netlink_ext_ack *extack)
4036{
4037	struct fib6_table *table;
4038	struct fib6_info *rt;
4039	struct fib6_node *fn;
 
4040	int err = -ESRCH;
4041
4042	table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
4043	if (!table) {
4044		NL_SET_ERR_MSG(extack, "FIB table does not exist");
4045		return err;
4046	}
4047
4048	rcu_read_lock();
4049
4050	fn = fib6_locate(&table->tb6_root,
4051			 &cfg->fc_dst, cfg->fc_dst_len,
4052			 &cfg->fc_src, cfg->fc_src_len,
4053			 !(cfg->fc_flags & RTF_CACHE));
4054
4055	if (fn) {
4056		for_each_fib6_node_rt_rcu(fn) {
4057			struct fib6_nh *nh;
4058
4059			if (rt->nh && cfg->fc_nh_id &&
4060			    rt->nh->id != cfg->fc_nh_id)
4061				continue;
4062
4063			if (cfg->fc_flags & RTF_CACHE) {
4064				int rc = 0;
4065
4066				if (rt->nh) {
4067					rc = ip6_del_cached_rt_nh(cfg, rt);
4068				} else if (cfg->fc_nh_id) {
4069					continue;
4070				} else {
4071					nh = rt->fib6_nh;
4072					rc = ip6_del_cached_rt(cfg, rt, nh);
4073				}
4074				if (rc != -ESRCH) {
4075					rcu_read_unlock();
4076					return rc;
4077				}
4078				continue;
4079			}
4080
4081			if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
4082				continue;
4083			if (cfg->fc_protocol &&
4084			    cfg->fc_protocol != rt->fib6_protocol)
4085				continue;
4086
4087			if (rt->nh) {
4088				if (!fib6_info_hold_safe(rt))
4089					continue;
4090				rcu_read_unlock();
4091
4092				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4093			}
4094			if (cfg->fc_nh_id)
4095				continue;
4096
4097			nh = rt->fib6_nh;
4098			if (cfg->fc_ifindex &&
4099			    (!nh->fib_nh_dev ||
4100			     nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
4101				continue;
4102			if (cfg->fc_flags & RTF_GATEWAY &&
4103			    !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
4104				continue;
4105			if (!fib6_info_hold_safe(rt))
4106				continue;
4107			rcu_read_unlock();
 
4108
4109			/* if gateway was specified only delete the one hop */
4110			if (cfg->fc_flags & RTF_GATEWAY)
4111				return __ip6_del_rt(rt, &cfg->fc_nlinfo);
4112
4113			return __ip6_del_rt_siblings(rt, cfg);
4114		}
4115	}
4116	rcu_read_unlock();
4117
4118	return err;
4119}
4120
4121static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
4122{
 
4123	struct netevent_redirect netevent;
4124	struct rt6_info *rt, *nrt = NULL;
4125	struct fib6_result res = {};
4126	struct ndisc_options ndopts;
4127	struct inet6_dev *in6_dev;
4128	struct neighbour *neigh;
4129	struct rd_msg *msg;
4130	int optlen, on_link;
4131	u8 *lladdr;
4132
4133	optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
4134	optlen -= sizeof(*msg);
4135
4136	if (optlen < 0) {
4137		net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
4138		return;
4139	}
4140
4141	msg = (struct rd_msg *)icmp6_hdr(skb);
4142
4143	if (ipv6_addr_is_multicast(&msg->dest)) {
4144		net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
4145		return;
4146	}
4147
4148	on_link = 0;
4149	if (ipv6_addr_equal(&msg->dest, &msg->target)) {
4150		on_link = 1;
4151	} else if (ipv6_addr_type(&msg->target) !=
4152		   (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
4153		net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
4154		return;
4155	}
4156
4157	in6_dev = __in6_dev_get(skb->dev);
4158	if (!in6_dev)
4159		return;
4160	if (READ_ONCE(in6_dev->cnf.forwarding) ||
4161	    !READ_ONCE(in6_dev->cnf.accept_redirects))
4162		return;
4163
4164	/* RFC2461 8.1:
4165	 *	The IP source address of the Redirect MUST be the same as the current
4166	 *	first-hop router for the specified ICMP Destination Address.
4167	 */
4168
4169	if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
4170		net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
4171		return;
4172	}
4173
4174	lladdr = NULL;
4175	if (ndopts.nd_opts_tgt_lladdr) {
4176		lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
4177					     skb->dev);
4178		if (!lladdr) {
4179			net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4180			return;
4181		}
4182	}
4183
4184	rt = dst_rt6_info(dst);
4185	if (rt->rt6i_flags & RTF_REJECT) {
4186		net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4187		return;
4188	}
4189
4190	/* Redirect received -> path was valid.
4191	 * Look, redirects are sent only in response to data packets,
4192	 * so that this nexthop apparently is reachable. --ANK
4193	 */
4194	dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4195
4196	neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4197	if (!neigh)
4198		return;
4199
4200	/*
4201	 *	We have finally decided to accept it.
4202	 */
4203
4204	ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4205		     NEIGH_UPDATE_F_WEAK_OVERRIDE|
4206		     NEIGH_UPDATE_F_OVERRIDE|
4207		     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4208				     NEIGH_UPDATE_F_ISROUTER)),
4209		     NDISC_REDIRECT, &ndopts);
4210
4211	rcu_read_lock();
4212	res.f6i = rcu_dereference(rt->from);
4213	if (!res.f6i)
4214		goto out;
4215
4216	if (res.f6i->nh) {
4217		struct fib6_nh_match_arg arg = {
4218			.dev = dst->dev,
4219			.gw = &rt->rt6i_gateway,
4220		};
4221
4222		nexthop_for_each_fib6_nh(res.f6i->nh,
4223					 fib6_nh_find_match, &arg);
4224
4225		/* fib6_info uses a nexthop that does not have fib6_nh
4226		 * using the dst->dev. Should be impossible
4227		 */
4228		if (!arg.match)
4229			goto out;
4230		res.nh = arg.match;
4231	} else {
4232		res.nh = res.f6i->fib6_nh;
4233	}
4234
4235	res.fib6_flags = res.f6i->fib6_flags;
4236	res.fib6_type = res.f6i->fib6_type;
4237	nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4238	if (!nrt)
4239		goto out;
4240
4241	nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4242	if (on_link)
4243		nrt->rt6i_flags &= ~RTF_GATEWAY;
4244
4245	nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4246
4247	/* rt6_insert_exception() will take care of duplicated exceptions */
4248	if (rt6_insert_exception(nrt, &res)) {
4249		dst_release_immediate(&nrt->dst);
4250		goto out;
4251	}
4252
4253	netevent.old = &rt->dst;
4254	netevent.new = &nrt->dst;
4255	netevent.daddr = &msg->dest;
4256	netevent.neigh = neigh;
4257	call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4258
 
 
 
 
 
4259out:
4260	rcu_read_unlock();
4261	neigh_release(neigh);
4262}
4263
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4264#ifdef CONFIG_IPV6_ROUTE_INFO
4265static struct fib6_info *rt6_get_route_info(struct net *net,
4266					   const struct in6_addr *prefix, int prefixlen,
4267					   const struct in6_addr *gwaddr,
4268					   struct net_device *dev)
4269{
4270	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4271	int ifindex = dev->ifindex;
4272	struct fib6_node *fn;
4273	struct fib6_info *rt = NULL;
4274	struct fib6_table *table;
4275
4276	table = fib6_get_table(net, tb_id);
4277	if (!table)
4278		return NULL;
4279
4280	rcu_read_lock();
4281	fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4282	if (!fn)
4283		goto out;
4284
4285	for_each_fib6_node_rt_rcu(fn) {
4286		/* these routes do not use nexthops */
4287		if (rt->nh)
4288			continue;
4289		if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4290			continue;
4291		if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4292		    !rt->fib6_nh->fib_nh_gw_family)
4293			continue;
4294		if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4295			continue;
4296		if (!fib6_info_hold_safe(rt))
4297			continue;
 
4298		break;
4299	}
4300out:
4301	rcu_read_unlock();
4302	return rt;
4303}
4304
4305static struct fib6_info *rt6_add_route_info(struct net *net,
4306					   const struct in6_addr *prefix, int prefixlen,
4307					   const struct in6_addr *gwaddr,
4308					   struct net_device *dev,
4309					   unsigned int pref)
4310{
4311	struct fib6_config cfg = {
 
4312		.fc_metric	= IP6_RT_PRIO_USER,
4313		.fc_ifindex	= dev->ifindex,
4314		.fc_dst_len	= prefixlen,
4315		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4316				  RTF_UP | RTF_PREF(pref),
4317		.fc_protocol = RTPROT_RA,
4318		.fc_type = RTN_UNICAST,
4319		.fc_nlinfo.portid = 0,
4320		.fc_nlinfo.nlh = NULL,
4321		.fc_nlinfo.nl_net = net,
4322	};
4323
4324	cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4325	cfg.fc_dst = *prefix;
4326	cfg.fc_gateway = *gwaddr;
4327
4328	/* We should treat it as a default route if prefix length is 0. */
4329	if (!prefixlen)
4330		cfg.fc_flags |= RTF_DEFAULT;
4331
4332	ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4333
4334	return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4335}
4336#endif
4337
4338struct fib6_info *rt6_get_dflt_router(struct net *net,
4339				     const struct in6_addr *addr,
4340				     struct net_device *dev)
4341{
4342	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4343	struct fib6_info *rt;
4344	struct fib6_table *table;
4345
4346	table = fib6_get_table(net, tb_id);
4347	if (!table)
4348		return NULL;
4349
4350	rcu_read_lock();
4351	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4352		struct fib6_nh *nh;
4353
4354		/* RA routes do not use nexthops */
4355		if (rt->nh)
4356			continue;
4357
4358		nh = rt->fib6_nh;
4359		if (dev == nh->fib_nh_dev &&
4360		    ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4361		    ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4362			break;
4363	}
4364	if (rt && !fib6_info_hold_safe(rt))
4365		rt = NULL;
4366	rcu_read_unlock();
4367	return rt;
4368}
4369
4370struct fib6_info *rt6_add_dflt_router(struct net *net,
4371				     const struct in6_addr *gwaddr,
4372				     struct net_device *dev,
4373				     unsigned int pref,
4374				     u32 defrtr_usr_metric,
4375				     int lifetime)
4376{
4377	struct fib6_config cfg = {
4378		.fc_table	= l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4379		.fc_metric	= defrtr_usr_metric,
4380		.fc_ifindex	= dev->ifindex,
4381		.fc_flags	= RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4382				  RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4383		.fc_protocol = RTPROT_RA,
4384		.fc_type = RTN_UNICAST,
4385		.fc_nlinfo.portid = 0,
4386		.fc_nlinfo.nlh = NULL,
4387		.fc_nlinfo.nl_net = net,
4388		.fc_expires = jiffies_to_clock_t(lifetime * HZ),
4389	};
4390
4391	cfg.fc_gateway = *gwaddr;
4392
4393	if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4394		struct fib6_table *table;
4395
4396		table = fib6_get_table(dev_net(dev), cfg.fc_table);
4397		if (table)
4398			table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4399	}
4400
4401	return rt6_get_dflt_router(net, gwaddr, dev);
4402}
4403
4404static void __rt6_purge_dflt_routers(struct net *net,
4405				     struct fib6_table *table)
4406{
4407	struct fib6_info *rt;
4408
4409restart:
4410	rcu_read_lock();
4411	for_each_fib6_node_rt_rcu(&table->tb6_root) {
4412		struct net_device *dev = fib6_info_nh_dev(rt);
4413		struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4414
4415		if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4416		    (!idev || idev->cnf.accept_ra != 2) &&
4417		    fib6_info_hold_safe(rt)) {
4418			rcu_read_unlock();
4419			ip6_del_rt(net, rt, false);
4420			goto restart;
4421		}
4422	}
4423	rcu_read_unlock();
4424
4425	table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4426}
4427
4428void rt6_purge_dflt_routers(struct net *net)
4429{
 
4430	struct fib6_table *table;
4431	struct hlist_head *head;
4432	unsigned int h;
4433
4434	rcu_read_lock();
 
 
 
4435
4436	for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4437		head = &net->ipv6.fib_table_hash[h];
4438		hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4439			if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4440				__rt6_purge_dflt_routers(net, table);
 
 
 
 
4441		}
4442	}
4443
4444	rcu_read_unlock();
4445}
4446
4447static void rtmsg_to_fib6_config(struct net *net,
4448				 struct in6_rtmsg *rtmsg,
4449				 struct fib6_config *cfg)
4450{
4451	*cfg = (struct fib6_config){
4452		.fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4453			 : RT6_TABLE_MAIN,
4454		.fc_ifindex = rtmsg->rtmsg_ifindex,
4455		.fc_metric = rtmsg->rtmsg_metric,
4456		.fc_expires = rtmsg->rtmsg_info,
4457		.fc_dst_len = rtmsg->rtmsg_dst_len,
4458		.fc_src_len = rtmsg->rtmsg_src_len,
4459		.fc_flags = rtmsg->rtmsg_flags,
4460		.fc_type = rtmsg->rtmsg_type,
4461
4462		.fc_nlinfo.nl_net = net,
 
 
 
 
 
 
 
 
4463
4464		.fc_dst = rtmsg->rtmsg_dst,
4465		.fc_src = rtmsg->rtmsg_src,
4466		.fc_gateway = rtmsg->rtmsg_gateway,
4467	};
4468}
4469
4470int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg)
4471{
4472	struct fib6_config cfg;
 
4473	int err;
4474
4475	if (cmd != SIOCADDRT && cmd != SIOCDELRT)
4476		return -EINVAL;
4477	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4478		return -EPERM;
 
 
 
 
 
 
 
4479
4480	rtmsg_to_fib6_config(net, rtmsg, &cfg);
 
 
 
 
 
 
 
 
 
 
 
4481
4482	rtnl_lock();
4483	switch (cmd) {
4484	case SIOCADDRT:
4485		/* Only do the default setting of fc_metric in route adding */
4486		if (cfg.fc_metric == 0)
4487			cfg.fc_metric = IP6_RT_PRIO_USER;
4488		err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4489		break;
4490	case SIOCDELRT:
4491		err = ip6_route_del(&cfg, NULL);
4492		break;
4493	}
4494	rtnl_unlock();
4495	return err;
4496}
4497
4498/*
4499 *	Drop the packet on the floor
4500 */
4501
4502static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4503{
 
4504	struct dst_entry *dst = skb_dst(skb);
4505	struct net *net = dev_net(dst->dev);
4506	struct inet6_dev *idev;
4507	SKB_DR(reason);
4508	int type;
4509
4510	if (netif_is_l3_master(skb->dev) ||
4511	    dst->dev == net->loopback_dev)
4512		idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4513	else
4514		idev = ip6_dst_idev(dst);
4515
4516	switch (ipstats_mib_noroutes) {
4517	case IPSTATS_MIB_INNOROUTES:
4518		type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4519		if (type == IPV6_ADDR_ANY) {
4520			SKB_DR_SET(reason, IP_INADDRERRORS);
4521			IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4522			break;
4523		}
4524		SKB_DR_SET(reason, IP_INNOROUTES);
4525		fallthrough;
4526	case IPSTATS_MIB_OUTNOROUTES:
4527		SKB_DR_OR(reason, IP_OUTNOROUTES);
4528		IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4529		break;
4530	}
4531
4532	/* Start over by dropping the dst for l3mdev case */
4533	if (netif_is_l3_master(skb->dev))
4534		skb_dst_drop(skb);
4535
4536	icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4537	kfree_skb_reason(skb, reason);
4538	return 0;
4539}
4540
4541static int ip6_pkt_discard(struct sk_buff *skb)
4542{
4543	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4544}
4545
4546static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4547{
4548	skb->dev = skb_dst(skb)->dev;
4549	return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4550}
4551
4552static int ip6_pkt_prohibit(struct sk_buff *skb)
4553{
4554	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4555}
4556
4557static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4558{
4559	skb->dev = skb_dst(skb)->dev;
4560	return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4561}
4562
4563/*
4564 *	Allocate a dst for local (unicast / anycast) address.
4565 */
4566
4567struct fib6_info *addrconf_f6i_alloc(struct net *net,
4568				     struct inet6_dev *idev,
4569				     const struct in6_addr *addr,
4570				     bool anycast, gfp_t gfp_flags,
4571				     struct netlink_ext_ack *extack)
4572{
4573	struct fib6_config cfg = {
4574		.fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4575		.fc_ifindex = idev->dev->ifindex,
4576		.fc_flags = RTF_UP | RTF_NONEXTHOP,
4577		.fc_dst = *addr,
4578		.fc_dst_len = 128,
4579		.fc_protocol = RTPROT_KERNEL,
4580		.fc_nlinfo.nl_net = net,
4581		.fc_ignore_dev_down = true,
4582	};
4583	struct fib6_info *f6i;
 
 
 
 
 
4584
4585	if (anycast) {
4586		cfg.fc_type = RTN_ANYCAST;
4587		cfg.fc_flags |= RTF_ANYCAST;
4588	} else {
4589		cfg.fc_type = RTN_LOCAL;
4590		cfg.fc_flags |= RTF_LOCAL;
4591	}
4592
4593	f6i = ip6_route_info_create(&cfg, gfp_flags, extack);
4594	if (!IS_ERR(f6i)) {
4595		f6i->dst_nocount = true;
4596
4597		if (!anycast &&
4598		    (READ_ONCE(net->ipv6.devconf_all->disable_policy) ||
4599		     READ_ONCE(idev->cnf.disable_policy)))
4600			f6i->dst_nopolicy = true;
4601	}
4602
4603	return f6i;
 
 
 
 
 
 
 
 
 
 
 
 
 
4604}
4605
4606/* remove deleted ip from prefsrc entries */
4607struct arg_dev_net_ip {
 
4608	struct net *net;
4609	struct in6_addr *addr;
4610};
4611
4612static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4613{
 
4614	struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4615	struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4616
4617	if (!rt->nh &&
4618	    rt != net->ipv6.fib6_null_entry &&
4619	    ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) &&
4620	    !ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) {
4621		spin_lock_bh(&rt6_exception_lock);
4622		/* remove prefsrc entry */
4623		rt->fib6_prefsrc.plen = 0;
4624		spin_unlock_bh(&rt6_exception_lock);
4625	}
4626	return 0;
4627}
4628
4629void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4630{
4631	struct net *net = dev_net(ifp->idev->dev);
4632	struct arg_dev_net_ip adni = {
 
4633		.net = net,
4634		.addr = &ifp->addr,
4635	};
4636	fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4637}
4638
4639#define RTF_RA_ROUTER		(RTF_ADDRCONF | RTF_DEFAULT)
 
4640
4641/* Remove routers and update dst entries when gateway turn into host. */
4642static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4643{
4644	struct in6_addr *gateway = (struct in6_addr *)arg;
4645	struct fib6_nh *nh;
4646
4647	/* RA routes do not use nexthops */
4648	if (rt->nh)
4649		return 0;
4650
4651	nh = rt->fib6_nh;
4652	if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4653	    nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4654		return -1;
4655
4656	/* Further clean up cached routes in exception table.
4657	 * This is needed because cached route may have a different
4658	 * gateway than its 'parent' in the case of an ip redirect.
4659	 */
4660	fib6_nh_exceptions_clean_tohost(nh, gateway);
4661
4662	return 0;
4663}
4664
4665void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4666{
4667	fib6_clean_all(net, fib6_clean_tohost, gateway);
4668}
4669
4670struct arg_netdev_event {
4671	const struct net_device *dev;
4672	union {
4673		unsigned char nh_flags;
4674		unsigned long event;
4675	};
4676};
4677
4678static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4679{
4680	struct fib6_info *iter;
4681	struct fib6_node *fn;
4682
4683	fn = rcu_dereference_protected(rt->fib6_node,
4684			lockdep_is_held(&rt->fib6_table->tb6_lock));
4685	iter = rcu_dereference_protected(fn->leaf,
4686			lockdep_is_held(&rt->fib6_table->tb6_lock));
4687	while (iter) {
4688		if (iter->fib6_metric == rt->fib6_metric &&
4689		    rt6_qualify_for_ecmp(iter))
4690			return iter;
4691		iter = rcu_dereference_protected(iter->fib6_next,
4692				lockdep_is_held(&rt->fib6_table->tb6_lock));
4693	}
4694
4695	return NULL;
4696}
4697
4698/* only called for fib entries with builtin fib6_nh */
4699static bool rt6_is_dead(const struct fib6_info *rt)
4700{
4701	if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4702	    (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4703	     ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4704		return true;
4705
4706	return false;
4707}
4708
4709static int rt6_multipath_total_weight(const struct fib6_info *rt)
4710{
4711	struct fib6_info *iter;
4712	int total = 0;
4713
4714	if (!rt6_is_dead(rt))
4715		total += rt->fib6_nh->fib_nh_weight;
4716
4717	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4718		if (!rt6_is_dead(iter))
4719			total += iter->fib6_nh->fib_nh_weight;
4720	}
4721
4722	return total;
4723}
4724
4725static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4726{
4727	int upper_bound = -1;
4728
4729	if (!rt6_is_dead(rt)) {
4730		*weight += rt->fib6_nh->fib_nh_weight;
4731		upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4732						    total) - 1;
4733	}
4734	atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4735}
4736
4737static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4738{
4739	struct fib6_info *iter;
4740	int weight = 0;
4741
4742	rt6_upper_bound_set(rt, &weight, total);
4743
4744	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4745		rt6_upper_bound_set(iter, &weight, total);
4746}
4747
4748void rt6_multipath_rebalance(struct fib6_info *rt)
4749{
4750	struct fib6_info *first;
4751	int total;
4752
4753	/* In case the entire multipath route was marked for flushing,
4754	 * then there is no need to rebalance upon the removal of every
4755	 * sibling route.
4756	 */
4757	if (!rt->fib6_nsiblings || rt->should_flush)
4758		return;
4759
4760	/* During lookup routes are evaluated in order, so we need to
4761	 * make sure upper bounds are assigned from the first sibling
4762	 * onwards.
4763	 */
4764	first = rt6_multipath_first_sibling(rt);
4765	if (WARN_ON_ONCE(!first))
4766		return;
4767
4768	total = rt6_multipath_total_weight(first);
4769	rt6_multipath_upper_bound_set(first, total);
4770}
4771
4772static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4773{
4774	const struct arg_netdev_event *arg = p_arg;
4775	struct net *net = dev_net(arg->dev);
4776
4777	if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4778	    rt->fib6_nh->fib_nh_dev == arg->dev) {
4779		rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4780		fib6_update_sernum_upto_root(net, rt);
4781		rt6_multipath_rebalance(rt);
4782	}
4783
4784	return 0;
4785}
4786
4787void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4788{
4789	struct arg_netdev_event arg = {
4790		.dev = dev,
4791		{
4792			.nh_flags = nh_flags,
4793		},
4794	};
4795
4796	if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4797		arg.nh_flags |= RTNH_F_LINKDOWN;
4798
4799	fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4800}
4801
4802/* only called for fib entries with inline fib6_nh */
4803static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4804				   const struct net_device *dev)
4805{
4806	struct fib6_info *iter;
4807
4808	if (rt->fib6_nh->fib_nh_dev == dev)
4809		return true;
4810	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4811		if (iter->fib6_nh->fib_nh_dev == dev)
4812			return true;
4813
4814	return false;
4815}
4816
4817static void rt6_multipath_flush(struct fib6_info *rt)
4818{
4819	struct fib6_info *iter;
4820
4821	rt->should_flush = 1;
4822	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4823		iter->should_flush = 1;
4824}
4825
4826static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4827					     const struct net_device *down_dev)
4828{
4829	struct fib6_info *iter;
4830	unsigned int dead = 0;
4831
4832	if (rt->fib6_nh->fib_nh_dev == down_dev ||
4833	    rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4834		dead++;
4835	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4836		if (iter->fib6_nh->fib_nh_dev == down_dev ||
4837		    iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4838			dead++;
4839
4840	return dead;
4841}
4842
4843static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4844				       const struct net_device *dev,
4845				       unsigned char nh_flags)
4846{
4847	struct fib6_info *iter;
4848
4849	if (rt->fib6_nh->fib_nh_dev == dev)
4850		rt->fib6_nh->fib_nh_flags |= nh_flags;
4851	list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4852		if (iter->fib6_nh->fib_nh_dev == dev)
4853			iter->fib6_nh->fib_nh_flags |= nh_flags;
4854}
4855
4856/* called with write lock held for table with rt */
4857static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4858{
4859	const struct arg_netdev_event *arg = p_arg;
4860	const struct net_device *dev = arg->dev;
4861	struct net *net = dev_net(dev);
4862
4863	if (rt == net->ipv6.fib6_null_entry || rt->nh)
4864		return 0;
4865
4866	switch (arg->event) {
4867	case NETDEV_UNREGISTER:
4868		return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4869	case NETDEV_DOWN:
4870		if (rt->should_flush)
4871			return -1;
4872		if (!rt->fib6_nsiblings)
4873			return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4874		if (rt6_multipath_uses_dev(rt, dev)) {
4875			unsigned int count;
4876
4877			count = rt6_multipath_dead_count(rt, dev);
4878			if (rt->fib6_nsiblings + 1 == count) {
4879				rt6_multipath_flush(rt);
4880				return -1;
4881			}
4882			rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4883						   RTNH_F_LINKDOWN);
4884			fib6_update_sernum(net, rt);
4885			rt6_multipath_rebalance(rt);
4886		}
4887		return -2;
4888	case NETDEV_CHANGE:
4889		if (rt->fib6_nh->fib_nh_dev != dev ||
4890		    rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4891			break;
4892		rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4893		rt6_multipath_rebalance(rt);
4894		break;
4895	}
4896
4897	return 0;
4898}
4899
4900void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4901{
4902	struct arg_netdev_event arg = {
4903		.dev = dev,
4904		{
4905			.event = event,
4906		},
4907	};
4908	struct net *net = dev_net(dev);
4909
4910	if (net->ipv6.sysctl.skip_notify_on_dev_down)
4911		fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4912	else
4913		fib6_clean_all(net, fib6_ifdown, &arg);
4914}
4915
4916void rt6_disable_ip(struct net_device *dev, unsigned long event)
4917{
4918	rt6_sync_down_dev(dev, event);
4919	rt6_uncached_list_flush_dev(dev);
4920	neigh_ifdown(&nd_tbl, dev);
4921}
4922
4923struct rt6_mtu_change_arg {
4924	struct net_device *dev;
4925	unsigned int mtu;
4926	struct fib6_info *f6i;
4927};
4928
4929static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4930{
4931	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4932	struct fib6_info *f6i = arg->f6i;
4933
4934	/* For administrative MTU increase, there is no way to discover
4935	 * IPv6 PMTU increase, so PMTU increase should be updated here.
4936	 * Since RFC 1981 doesn't include administrative MTU increase
4937	 * update PMTU increase is a MUST. (i.e. jumbo frame)
4938	 */
4939	if (nh->fib_nh_dev == arg->dev) {
4940		struct inet6_dev *idev = __in6_dev_get(arg->dev);
4941		u32 mtu = f6i->fib6_pmtu;
4942
4943		if (mtu >= arg->mtu ||
4944		    (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4945			fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4946
4947		spin_lock_bh(&rt6_exception_lock);
4948		rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4949		spin_unlock_bh(&rt6_exception_lock);
4950	}
4951
4952	return 0;
4953}
4954
4955static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4956{
4957	struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4958	struct inet6_dev *idev;
4959
4960	/* In IPv6 pmtu discovery is not optional,
4961	   so that RTAX_MTU lock cannot disable it.
4962	   We still use this lock to block changes
4963	   caused by addrconf/ndisc.
4964	*/
4965
4966	idev = __in6_dev_get(arg->dev);
4967	if (!idev)
4968		return 0;
4969
4970	if (fib6_metric_locked(f6i, RTAX_MTU))
4971		return 0;
4972
4973	arg->f6i = f6i;
4974	if (f6i->nh) {
4975		/* fib6_nh_mtu_change only returns 0, so this is safe */
4976		return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4977						arg);
 
 
 
 
 
 
 
 
 
 
 
 
4978	}
4979
4980	return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4981}
4982
4983void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4984{
4985	struct rt6_mtu_change_arg arg = {
4986		.dev = dev,
4987		.mtu = mtu,
4988	};
4989
4990	fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4991}
4992
4993static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4994	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
4995	[RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4996	[RTA_PREFSRC]		= { .len = sizeof(struct in6_addr) },
4997	[RTA_OIF]               = { .type = NLA_U32 },
4998	[RTA_IIF]		= { .type = NLA_U32 },
4999	[RTA_PRIORITY]          = { .type = NLA_U32 },
5000	[RTA_METRICS]           = { .type = NLA_NESTED },
5001	[RTA_MULTIPATH]		= { .len = sizeof(struct rtnexthop) },
5002	[RTA_PREF]              = { .type = NLA_U8 },
5003	[RTA_ENCAP_TYPE]	= { .type = NLA_U16 },
5004	[RTA_ENCAP]		= { .type = NLA_NESTED },
5005	[RTA_EXPIRES]		= { .type = NLA_U32 },
5006	[RTA_UID]		= { .type = NLA_U32 },
5007	[RTA_MARK]		= { .type = NLA_U32 },
5008	[RTA_TABLE]		= { .type = NLA_U32 },
5009	[RTA_IP_PROTO]		= { .type = NLA_U8 },
5010	[RTA_SPORT]		= { .type = NLA_U16 },
5011	[RTA_DPORT]		= { .type = NLA_U16 },
5012	[RTA_NH_ID]		= { .type = NLA_U32 },
5013};
5014
5015static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
5016			      struct fib6_config *cfg,
5017			      struct netlink_ext_ack *extack)
5018{
5019	struct rtmsg *rtm;
5020	struct nlattr *tb[RTA_MAX+1];
5021	unsigned int pref;
5022	int err;
5023
5024	err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5025				     rtm_ipv6_policy, extack);
5026	if (err < 0)
5027		goto errout;
5028
5029	err = -EINVAL;
5030	rtm = nlmsg_data(nlh);
 
5031
5032	if (rtm->rtm_tos) {
5033		NL_SET_ERR_MSG(extack,
5034			       "Invalid dsfield (tos): option not available for IPv6");
5035		goto errout;
5036	}
5037
5038	*cfg = (struct fib6_config){
5039		.fc_table = rtm->rtm_table,
5040		.fc_dst_len = rtm->rtm_dst_len,
5041		.fc_src_len = rtm->rtm_src_len,
5042		.fc_flags = RTF_UP,
5043		.fc_protocol = rtm->rtm_protocol,
5044		.fc_type = rtm->rtm_type,
5045
5046		.fc_nlinfo.portid = NETLINK_CB(skb).portid,
5047		.fc_nlinfo.nlh = nlh,
5048		.fc_nlinfo.nl_net = sock_net(skb->sk),
5049	};
5050
5051	if (rtm->rtm_type == RTN_UNREACHABLE ||
5052	    rtm->rtm_type == RTN_BLACKHOLE ||
5053	    rtm->rtm_type == RTN_PROHIBIT ||
5054	    rtm->rtm_type == RTN_THROW)
5055		cfg->fc_flags |= RTF_REJECT;
5056
5057	if (rtm->rtm_type == RTN_LOCAL)
5058		cfg->fc_flags |= RTF_LOCAL;
5059
5060	if (rtm->rtm_flags & RTM_F_CLONED)
5061		cfg->fc_flags |= RTF_CACHE;
5062
5063	cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
5064
5065	if (tb[RTA_NH_ID]) {
5066		if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
5067		    tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
5068			NL_SET_ERR_MSG(extack,
5069				       "Nexthop specification and nexthop id are mutually exclusive");
5070			goto errout;
5071		}
5072		cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
5073	}
5074
5075	if (tb[RTA_GATEWAY]) {
5076		cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
5077		cfg->fc_flags |= RTF_GATEWAY;
5078	}
5079	if (tb[RTA_VIA]) {
5080		NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
5081		goto errout;
5082	}
5083
5084	if (tb[RTA_DST]) {
5085		int plen = (rtm->rtm_dst_len + 7) >> 3;
5086
5087		if (nla_len(tb[RTA_DST]) < plen)
5088			goto errout;
5089
5090		nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
5091	}
5092
5093	if (tb[RTA_SRC]) {
5094		int plen = (rtm->rtm_src_len + 7) >> 3;
5095
5096		if (nla_len(tb[RTA_SRC]) < plen)
5097			goto errout;
5098
5099		nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
5100	}
5101
5102	if (tb[RTA_PREFSRC])
5103		cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
5104
5105	if (tb[RTA_OIF])
5106		cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
5107
5108	if (tb[RTA_PRIORITY])
5109		cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
5110
5111	if (tb[RTA_METRICS]) {
5112		cfg->fc_mx = nla_data(tb[RTA_METRICS]);
5113		cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
5114	}
5115
5116	if (tb[RTA_TABLE])
5117		cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
5118
5119	if (tb[RTA_MULTIPATH]) {
5120		cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
5121		cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
5122
5123		err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
5124						     cfg->fc_mp_len, extack);
5125		if (err < 0)
5126			goto errout;
5127	}
5128
5129	if (tb[RTA_PREF]) {
5130		pref = nla_get_u8(tb[RTA_PREF]);
5131		if (pref != ICMPV6_ROUTER_PREF_LOW &&
5132		    pref != ICMPV6_ROUTER_PREF_HIGH)
5133			pref = ICMPV6_ROUTER_PREF_MEDIUM;
5134		cfg->fc_flags |= RTF_PREF(pref);
5135	}
5136
5137	if (tb[RTA_ENCAP])
5138		cfg->fc_encap = tb[RTA_ENCAP];
5139
5140	if (tb[RTA_ENCAP_TYPE]) {
5141		cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
5142
5143		err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
5144		if (err < 0)
5145			goto errout;
5146	}
5147
5148	if (tb[RTA_EXPIRES]) {
5149		unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
5150
5151		if (addrconf_finite_timeout(timeout)) {
5152			cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
5153			cfg->fc_flags |= RTF_EXPIRES;
5154		}
5155	}
5156
5157	err = 0;
5158errout:
5159	return err;
5160}
5161
5162struct rt6_nh {
5163	struct fib6_info *fib6_info;
5164	struct fib6_config r_cfg;
5165	struct list_head next;
5166};
5167
5168static int ip6_route_info_append(struct net *net,
5169				 struct list_head *rt6_nh_list,
5170				 struct fib6_info *rt,
5171				 struct fib6_config *r_cfg)
5172{
5173	struct rt6_nh *nh;
5174	int err = -EEXIST;
5175
5176	list_for_each_entry(nh, rt6_nh_list, next) {
5177		/* check if fib6_info already exists */
5178		if (rt6_duplicate_nexthop(nh->fib6_info, rt))
5179			return err;
5180	}
5181
5182	nh = kzalloc(sizeof(*nh), GFP_KERNEL);
5183	if (!nh)
5184		return -ENOMEM;
5185	nh->fib6_info = rt;
5186	memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
5187	list_add_tail(&nh->next, rt6_nh_list);
5188
5189	return 0;
5190}
5191
5192static void ip6_route_mpath_notify(struct fib6_info *rt,
5193				   struct fib6_info *rt_last,
5194				   struct nl_info *info,
5195				   __u16 nlflags)
5196{
5197	/* if this is an APPEND route, then rt points to the first route
5198	 * inserted and rt_last points to last route inserted. Userspace
5199	 * wants a consistent dump of the route which starts at the first
5200	 * nexthop. Since sibling routes are always added at the end of
5201	 * the list, find the first sibling of the last route appended
5202	 */
5203	rcu_read_lock();
5204
5205	if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5206		rt = list_first_or_null_rcu(&rt_last->fib6_siblings,
5207					    struct fib6_info,
5208					    fib6_siblings);
5209	}
5210
5211	if (rt)
5212		inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5213
5214	rcu_read_unlock();
5215}
5216
5217static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
5218{
5219	bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
5220	bool should_notify = false;
5221	struct fib6_info *leaf;
5222	struct fib6_node *fn;
5223
5224	rcu_read_lock();
5225	fn = rcu_dereference(rt->fib6_node);
5226	if (!fn)
5227		goto out;
5228
5229	leaf = rcu_dereference(fn->leaf);
5230	if (!leaf)
5231		goto out;
5232
5233	if (rt == leaf ||
5234	    (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
5235	     rt6_qualify_for_ecmp(leaf)))
5236		should_notify = true;
5237out:
5238	rcu_read_unlock();
5239
5240	return should_notify;
5241}
5242
5243static int fib6_gw_from_attr(struct in6_addr *gw, struct nlattr *nla,
5244			     struct netlink_ext_ack *extack)
5245{
5246	if (nla_len(nla) < sizeof(*gw)) {
5247		NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_GATEWAY");
5248		return -EINVAL;
5249	}
5250
5251	*gw = nla_get_in6_addr(nla);
5252
5253	return 0;
5254}
5255
5256static int ip6_route_multipath_add(struct fib6_config *cfg,
5257				   struct netlink_ext_ack *extack)
5258{
5259	struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5260	struct nl_info *info = &cfg->fc_nlinfo;
5261	struct fib6_config r_cfg;
5262	struct rtnexthop *rtnh;
5263	struct fib6_info *rt;
5264	struct rt6_nh *err_nh;
5265	struct rt6_nh *nh, *nh_safe;
5266	__u16 nlflags;
5267	int remaining;
5268	int attrlen;
5269	int err = 1;
5270	int nhn = 0;
5271	int replace = (cfg->fc_nlinfo.nlh &&
5272		       (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5273	LIST_HEAD(rt6_nh_list);
5274
5275	nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5276	if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5277		nlflags |= NLM_F_APPEND;
5278
 
 
5279	remaining = cfg->fc_mp_len;
5280	rtnh = (struct rtnexthop *)cfg->fc_mp;
5281
5282	/* Parse a Multipath Entry and build a list (rt6_nh_list) of
5283	 * fib6_info structs per nexthop
5284	 */
5285	while (rtnh_ok(rtnh, remaining)) {
5286		memcpy(&r_cfg, cfg, sizeof(*cfg));
5287		if (rtnh->rtnh_ifindex)
5288			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5289
5290		attrlen = rtnh_attrlen(rtnh);
5291		if (attrlen > 0) {
5292			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5293
5294			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5295			if (nla) {
5296				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5297							extack);
5298				if (err)
5299					goto cleanup;
5300
5301				r_cfg.fc_flags |= RTF_GATEWAY;
5302			}
5303			r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5304
5305			/* RTA_ENCAP_TYPE length checked in
5306			 * lwtunnel_valid_encap_type_attr
5307			 */
5308			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5309			if (nla)
5310				r_cfg.fc_encap_type = nla_get_u16(nla);
5311		}
5312
5313		r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5314		rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5315		if (IS_ERR(rt)) {
5316			err = PTR_ERR(rt);
5317			rt = NULL;
5318			goto cleanup;
5319		}
5320		if (!rt6_qualify_for_ecmp(rt)) {
5321			err = -EINVAL;
5322			NL_SET_ERR_MSG(extack,
5323				       "Device only routes can not be added for IPv6 using the multipath API.");
5324			fib6_info_release(rt);
5325			goto cleanup;
5326		}
5327
5328		rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5329
5330		err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5331					    rt, &r_cfg);
5332		if (err) {
5333			fib6_info_release(rt);
5334			goto cleanup;
5335		}
5336
5337		rtnh = rtnh_next(rtnh, &remaining);
5338	}
5339
5340	if (list_empty(&rt6_nh_list)) {
5341		NL_SET_ERR_MSG(extack,
5342			       "Invalid nexthop configuration - no valid nexthops");
5343		return -EINVAL;
5344	}
5345
5346	/* for add and replace send one notification with all nexthops.
5347	 * Skip the notification in fib6_add_rt2node and send one with
5348	 * the full route when done
5349	 */
5350	info->skip_notify = 1;
5351
5352	/* For add and replace, send one notification with all nexthops. For
5353	 * append, send one notification with all appended nexthops.
5354	 */
5355	info->skip_notify_kernel = 1;
5356
5357	err_nh = NULL;
5358	list_for_each_entry(nh, &rt6_nh_list, next) {
5359		err = __ip6_ins_rt(nh->fib6_info, info, extack);
5360
5361		if (err) {
5362			if (replace && nhn)
5363				NL_SET_ERR_MSG_MOD(extack,
5364						   "multipath route replace failed (check consistency of installed routes)");
5365			err_nh = nh;
5366			goto add_errout;
5367		}
5368		/* save reference to last route successfully inserted */
5369		rt_last = nh->fib6_info;
5370
5371		/* save reference to first route for notification */
5372		if (!rt_notif)
5373			rt_notif = nh->fib6_info;
5374
5375		/* Because each route is added like a single route we remove
5376		 * these flags after the first nexthop: if there is a collision,
5377		 * we have already failed to add the first nexthop:
5378		 * fib6_add_rt2node() has rejected it; when replacing, old
5379		 * nexthops have been replaced by first new, the rest should
5380		 * be added to it.
5381		 */
5382		if (cfg->fc_nlinfo.nlh) {
5383			cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5384							     NLM_F_REPLACE);
5385			cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
5386		}
5387		nhn++;
5388	}
5389
5390	/* An in-kernel notification should only be sent in case the new
5391	 * multipath route is added as the first route in the node, or if
5392	 * it was appended to it. We pass 'rt_notif' since it is the first
5393	 * sibling and might allow us to skip some checks in the replace case.
5394	 */
5395	if (ip6_route_mpath_should_notify(rt_notif)) {
5396		enum fib_event_type fib_event;
5397
5398		if (rt_notif->fib6_nsiblings != nhn - 1)
5399			fib_event = FIB_EVENT_ENTRY_APPEND;
5400		else
5401			fib_event = FIB_EVENT_ENTRY_REPLACE;
5402
5403		err = call_fib6_multipath_entry_notifiers(info->nl_net,
5404							  fib_event, rt_notif,
5405							  nhn - 1, extack);
5406		if (err) {
5407			/* Delete all the siblings that were just added */
5408			err_nh = NULL;
5409			goto add_errout;
5410		}
5411	}
5412
5413	/* success ... tell user about new route */
5414	ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5415	goto cleanup;
5416
5417add_errout:
5418	/* send notification for routes that were added so that
5419	 * the delete notifications sent by ip6_route_del are
5420	 * coherent
5421	 */
5422	if (rt_notif)
5423		ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5424
5425	/* Delete routes that were already added */
5426	list_for_each_entry(nh, &rt6_nh_list, next) {
5427		if (err_nh == nh)
5428			break;
5429		ip6_route_del(&nh->r_cfg, extack);
5430	}
5431
5432cleanup:
5433	list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5434		fib6_info_release(nh->fib6_info);
5435		list_del(&nh->next);
5436		kfree(nh);
5437	}
5438
5439	return err;
5440}
5441
5442static int ip6_route_multipath_del(struct fib6_config *cfg,
5443				   struct netlink_ext_ack *extack)
5444{
5445	struct fib6_config r_cfg;
5446	struct rtnexthop *rtnh;
5447	int last_err = 0;
5448	int remaining;
5449	int attrlen;
5450	int err;
5451
5452	remaining = cfg->fc_mp_len;
5453	rtnh = (struct rtnexthop *)cfg->fc_mp;
5454
5455	/* Parse a Multipath Entry */
5456	while (rtnh_ok(rtnh, remaining)) {
5457		memcpy(&r_cfg, cfg, sizeof(*cfg));
5458		if (rtnh->rtnh_ifindex)
5459			r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5460
5461		attrlen = rtnh_attrlen(rtnh);
5462		if (attrlen > 0) {
5463			struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5464
5465			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5466			if (nla) {
5467				err = fib6_gw_from_attr(&r_cfg.fc_gateway, nla,
5468							extack);
5469				if (err) {
5470					last_err = err;
5471					goto next_rtnh;
5472				}
5473
5474				r_cfg.fc_flags |= RTF_GATEWAY;
5475			}
5476		}
5477		err = ip6_route_del(&r_cfg, extack);
5478		if (err)
5479			last_err = err;
5480
5481next_rtnh:
5482		rtnh = rtnh_next(rtnh, &remaining);
5483	}
5484
5485	return last_err;
5486}
5487
5488static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5489			      struct netlink_ext_ack *extack)
5490{
5491	struct fib6_config cfg;
5492	int err;
5493
5494	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5495	if (err < 0)
5496		return err;
5497
5498	if (cfg.fc_nh_id &&
5499	    !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5500		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5501		return -EINVAL;
5502	}
5503
5504	if (cfg.fc_mp)
5505		return ip6_route_multipath_del(&cfg, extack);
5506	else {
5507		cfg.fc_delete_all_nh = 1;
5508		return ip6_route_del(&cfg, extack);
5509	}
5510}
5511
5512static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5513			      struct netlink_ext_ack *extack)
5514{
5515	struct fib6_config cfg;
5516	int err;
5517
5518	err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5519	if (err < 0)
5520		return err;
5521
5522	if (cfg.fc_metric == 0)
5523		cfg.fc_metric = IP6_RT_PRIO_USER;
5524
5525	if (cfg.fc_mp)
5526		return ip6_route_multipath_add(&cfg, extack);
5527	else
5528		return ip6_route_add(&cfg, GFP_KERNEL, extack);
5529}
5530
5531/* add the overhead of this fib6_nh to nexthop_len */
5532static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5533{
5534	int *nexthop_len = arg;
5535
5536	*nexthop_len += nla_total_size(0)	 /* RTA_MULTIPATH */
5537		     + NLA_ALIGN(sizeof(struct rtnexthop))
5538		     + nla_total_size(16); /* RTA_GATEWAY */
5539
5540	if (nh->fib_nh_lws) {
5541		/* RTA_ENCAP_TYPE */
5542		*nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5543		/* RTA_ENCAP */
5544		*nexthop_len += nla_total_size(2);
5545	}
5546
5547	return 0;
5548}
5549
5550static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5551{
5552	int nexthop_len;
5553
5554	if (f6i->nh) {
5555		nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5556		nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5557					 &nexthop_len);
5558	} else {
5559		struct fib6_nh *nh = f6i->fib6_nh;
5560		struct fib6_info *sibling;
5561
5562		nexthop_len = 0;
5563		if (f6i->fib6_nsiblings) {
5564			rt6_nh_nlmsg_size(nh, &nexthop_len);
5565
5566			rcu_read_lock();
5567
5568			list_for_each_entry_rcu(sibling, &f6i->fib6_siblings,
5569						fib6_siblings) {
5570				rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len);
5571			}
5572
5573			rcu_read_unlock();
5574		}
5575		nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5576	}
5577
5578	return NLMSG_ALIGN(sizeof(struct rtmsg))
5579	       + nla_total_size(16) /* RTA_SRC */
5580	       + nla_total_size(16) /* RTA_DST */
5581	       + nla_total_size(16) /* RTA_GATEWAY */
5582	       + nla_total_size(16) /* RTA_PREFSRC */
5583	       + nla_total_size(4) /* RTA_TABLE */
5584	       + nla_total_size(4) /* RTA_IIF */
5585	       + nla_total_size(4) /* RTA_OIF */
5586	       + nla_total_size(4) /* RTA_PRIORITY */
5587	       + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5588	       + nla_total_size(sizeof(struct rta_cacheinfo))
5589	       + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5590	       + nla_total_size(1) /* RTA_PREF */
5591	       + nexthop_len;
5592}
5593
5594static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5595				 unsigned char *flags)
 
 
 
5596{
5597	if (nexthop_is_multipath(nh)) {
5598		struct nlattr *mp;
 
 
5599
5600		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5601		if (!mp)
5602			goto nla_put_failure;
5603
5604		if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5605			goto nla_put_failure;
5606
5607		nla_nest_end(skb, mp);
5608	} else {
5609		struct fib6_nh *fib6_nh;
5610
5611		fib6_nh = nexthop_fib6_nh(nh);
5612		if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5613				     flags, false) < 0)
5614			goto nla_put_failure;
5615	}
5616
5617	return 0;
5618
5619nla_put_failure:
5620	return -EMSGSIZE;
5621}
5622
5623static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5624			 struct fib6_info *rt, struct dst_entry *dst,
5625			 struct in6_addr *dest, struct in6_addr *src,
5626			 int iif, int type, u32 portid, u32 seq,
5627			 unsigned int flags)
5628{
5629	struct rt6_info *rt6 = dst_rt6_info(dst);
5630	struct rt6key *rt6_dst, *rt6_src;
5631	u32 *pmetrics, table, rt6_flags;
5632	unsigned char nh_flags = 0;
5633	struct nlmsghdr *nlh;
5634	struct rtmsg *rtm;
5635	long expires = 0;
5636
5637	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5638	if (!nlh)
5639		return -EMSGSIZE;
5640
5641	if (rt6) {
5642		rt6_dst = &rt6->rt6i_dst;
5643		rt6_src = &rt6->rt6i_src;
5644		rt6_flags = rt6->rt6i_flags;
5645	} else {
5646		rt6_dst = &rt->fib6_dst;
5647		rt6_src = &rt->fib6_src;
5648		rt6_flags = rt->fib6_flags;
5649	}
5650
5651	rtm = nlmsg_data(nlh);
5652	rtm->rtm_family = AF_INET6;
5653	rtm->rtm_dst_len = rt6_dst->plen;
5654	rtm->rtm_src_len = rt6_src->plen;
5655	rtm->rtm_tos = 0;
5656	if (rt->fib6_table)
5657		table = rt->fib6_table->tb6_id;
5658	else
5659		table = RT6_TABLE_UNSPEC;
5660	rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5661	if (nla_put_u32(skb, RTA_TABLE, table))
5662		goto nla_put_failure;
5663
5664	rtm->rtm_type = rt->fib6_type;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5665	rtm->rtm_flags = 0;
5666	rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5667	rtm->rtm_protocol = rt->fib6_protocol;
 
 
 
 
 
 
 
 
5668
5669	if (rt6_flags & RTF_CACHE)
5670		rtm->rtm_flags |= RTM_F_CLONED;
5671
5672	if (dest) {
5673		if (nla_put_in6_addr(skb, RTA_DST, dest))
5674			goto nla_put_failure;
5675		rtm->rtm_dst_len = 128;
5676	} else if (rtm->rtm_dst_len)
5677		if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5678			goto nla_put_failure;
5679#ifdef CONFIG_IPV6_SUBTREES
5680	if (src) {
5681		if (nla_put_in6_addr(skb, RTA_SRC, src))
5682			goto nla_put_failure;
5683		rtm->rtm_src_len = 128;
5684	} else if (rtm->rtm_src_len &&
5685		   nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5686		goto nla_put_failure;
5687#endif
5688	if (iif) {
5689#ifdef CONFIG_IPV6_MROUTE
5690		if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5691			int err = ip6mr_get_route(net, skb, rtm, portid);
5692
5693			if (err == 0)
5694				return 0;
5695			if (err < 0)
5696				goto nla_put_failure;
 
 
 
 
 
5697		} else
5698#endif
5699			if (nla_put_u32(skb, RTA_IIF, iif))
5700				goto nla_put_failure;
5701	} else if (dest) {
5702		struct in6_addr saddr_buf;
5703		if (ip6_route_get_saddr(net, rt, dest, 0, 0, &saddr_buf) == 0 &&
5704		    nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5705			goto nla_put_failure;
5706	}
5707
5708	if (rt->fib6_prefsrc.plen) {
5709		struct in6_addr saddr_buf;
5710		saddr_buf = rt->fib6_prefsrc.addr;
5711		if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5712			goto nla_put_failure;
5713	}
5714
5715	pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5716	if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5717		goto nla_put_failure;
5718
5719	if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5720		goto nla_put_failure;
5721
5722	/* For multipath routes, walk the siblings list and add
5723	 * each as a nexthop within RTA_MULTIPATH.
5724	 */
5725	if (rt6) {
5726		if (rt6_flags & RTF_GATEWAY &&
5727		    nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5728			goto nla_put_failure;
5729
5730		if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5731			goto nla_put_failure;
5732
5733		if (dst->lwtstate &&
5734		    lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
5735			goto nla_put_failure;
5736	} else if (rt->fib6_nsiblings) {
5737		struct fib6_info *sibling;
5738		struct nlattr *mp;
5739
5740		mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5741		if (!mp)
5742			goto nla_put_failure;
5743
5744		if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5745				    rt->fib6_nh->fib_nh_weight, AF_INET6,
5746				    0) < 0)
5747			goto nla_put_failure;
5748
5749		rcu_read_lock();
5750
5751		list_for_each_entry_rcu(sibling, &rt->fib6_siblings,
5752					fib6_siblings) {
5753			if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5754					    sibling->fib6_nh->fib_nh_weight,
5755					    AF_INET6, 0) < 0) {
5756				rcu_read_unlock();
5757
5758				goto nla_put_failure;
5759			}
5760		}
5761
5762		rcu_read_unlock();
5763
5764		nla_nest_end(skb, mp);
5765	} else if (rt->nh) {
5766		if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5767			goto nla_put_failure;
5768
5769		if (nexthop_is_blackhole(rt->nh))
5770			rtm->rtm_type = RTN_BLACKHOLE;
5771
5772		if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) &&
5773		    rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5774			goto nla_put_failure;
5775
5776		rtm->rtm_flags |= nh_flags;
5777	} else {
5778		if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5779				     &nh_flags, false) < 0)
5780			goto nla_put_failure;
5781
5782		rtm->rtm_flags |= nh_flags;
5783	}
5784
5785	if (rt6_flags & RTF_EXPIRES) {
5786		expires = dst ? dst->expires : rt->expires;
5787		expires -= jiffies;
5788	}
5789
5790	if (!dst) {
5791		if (READ_ONCE(rt->offload))
5792			rtm->rtm_flags |= RTM_F_OFFLOAD;
5793		if (READ_ONCE(rt->trap))
5794			rtm->rtm_flags |= RTM_F_TRAP;
5795		if (READ_ONCE(rt->offload_failed))
5796			rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED;
5797	}
5798
5799	if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5800		goto nla_put_failure;
5801
5802	if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5803		goto nla_put_failure;
5804
5805
5806	nlmsg_end(skb, nlh);
5807	return 0;
5808
5809nla_put_failure:
5810	nlmsg_cancel(skb, nlh);
5811	return -EMSGSIZE;
5812}
5813
5814static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5815{
5816	const struct net_device *dev = arg;
5817
5818	if (nh->fib_nh_dev == dev)
5819		return 1;
5820
5821	return 0;
5822}
5823
5824static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5825			       const struct net_device *dev)
5826{
5827	if (f6i->nh) {
5828		struct net_device *_dev = (struct net_device *)dev;
5829
5830		return !!nexthop_for_each_fib6_nh(f6i->nh,
5831						  fib6_info_nh_uses_dev,
5832						  _dev);
5833	}
5834
5835	if (f6i->fib6_nh->fib_nh_dev == dev)
5836		return true;
5837
5838	if (f6i->fib6_nsiblings) {
5839		struct fib6_info *sibling, *next_sibling;
5840
5841		list_for_each_entry_safe(sibling, next_sibling,
5842					 &f6i->fib6_siblings, fib6_siblings) {
5843			if (sibling->fib6_nh->fib_nh_dev == dev)
5844				return true;
5845		}
5846	}
5847
5848	return false;
5849}
5850
5851struct fib6_nh_exception_dump_walker {
5852	struct rt6_rtnl_dump_arg *dump;
5853	struct fib6_info *rt;
5854	unsigned int flags;
5855	unsigned int skip;
5856	unsigned int count;
5857};
5858
5859static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5860{
5861	struct fib6_nh_exception_dump_walker *w = arg;
5862	struct rt6_rtnl_dump_arg *dump = w->dump;
5863	struct rt6_exception_bucket *bucket;
5864	struct rt6_exception *rt6_ex;
5865	int i, err;
5866
5867	bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5868	if (!bucket)
5869		return 0;
5870
5871	for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5872		hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5873			if (w->skip) {
5874				w->skip--;
5875				continue;
5876			}
5877
5878			/* Expiration of entries doesn't bump sernum, insertion
5879			 * does. Removal is triggered by insertion, so we can
5880			 * rely on the fact that if entries change between two
5881			 * partial dumps, this node is scanned again completely,
5882			 * see rt6_insert_exception() and fib6_dump_table().
5883			 *
5884			 * Count expired entries we go through as handled
5885			 * entries that we'll skip next time, in case of partial
5886			 * node dump. Otherwise, if entries expire meanwhile,
5887			 * we'll skip the wrong amount.
5888			 */
5889			if (rt6_check_expired(rt6_ex->rt6i)) {
5890				w->count++;
5891				continue;
5892			}
5893
5894			err = rt6_fill_node(dump->net, dump->skb, w->rt,
5895					    &rt6_ex->rt6i->dst, NULL, NULL, 0,
5896					    RTM_NEWROUTE,
5897					    NETLINK_CB(dump->cb->skb).portid,
5898					    dump->cb->nlh->nlmsg_seq, w->flags);
5899			if (err)
5900				return err;
5901
5902			w->count++;
5903		}
5904		bucket++;
5905	}
5906
5907	return 0;
5908}
5909
5910/* Return -1 if done with node, number of handled routes on partial dump */
5911int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5912{
5913	struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5914	struct fib_dump_filter *filter = &arg->filter;
5915	unsigned int flags = NLM_F_MULTI;
5916	struct net *net = arg->net;
5917	int count = 0;
5918
5919	if (rt == net->ipv6.fib6_null_entry)
5920		return -1;
5921
5922	if ((filter->flags & RTM_F_PREFIX) &&
5923	    !(rt->fib6_flags & RTF_PREFIX_RT)) {
5924		/* success since this is not a prefix route */
5925		return -1;
5926	}
5927	if (filter->filter_set &&
5928	    ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5929	     (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5930	     (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5931		return -1;
5932	}
5933
5934	if (filter->filter_set ||
5935	    !filter->dump_routes || !filter->dump_exceptions) {
5936		flags |= NLM_F_DUMP_FILTERED;
5937	}
5938
5939	if (filter->dump_routes) {
5940		if (skip) {
5941			skip--;
5942		} else {
5943			if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5944					  0, RTM_NEWROUTE,
5945					  NETLINK_CB(arg->cb->skb).portid,
5946					  arg->cb->nlh->nlmsg_seq, flags)) {
5947				return 0;
5948			}
5949			count++;
5950		}
5951	}
5952
5953	if (filter->dump_exceptions) {
5954		struct fib6_nh_exception_dump_walker w = { .dump = arg,
5955							   .rt = rt,
5956							   .flags = flags,
5957							   .skip = skip,
5958							   .count = 0 };
5959		int err;
5960
5961		rcu_read_lock();
5962		if (rt->nh) {
5963			err = nexthop_for_each_fib6_nh(rt->nh,
5964						       rt6_nh_dump_exceptions,
5965						       &w);
5966		} else {
5967			err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5968		}
5969		rcu_read_unlock();
5970
5971		if (err)
5972			return count + w.count;
5973	}
5974
5975	return -1;
5976}
5977
5978static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5979					const struct nlmsghdr *nlh,
5980					struct nlattr **tb,
5981					struct netlink_ext_ack *extack)
5982{
5983	struct rtmsg *rtm;
5984	int i, err;
5985
5986	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5987		NL_SET_ERR_MSG_MOD(extack,
5988				   "Invalid header for get route request");
5989		return -EINVAL;
5990	}
5991
5992	if (!netlink_strict_get_check(skb))
5993		return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5994					      rtm_ipv6_policy, extack);
5995
5996	rtm = nlmsg_data(nlh);
5997	if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5998	    (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5999	    rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
6000	    rtm->rtm_type) {
6001		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
6002		return -EINVAL;
6003	}
6004	if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
6005		NL_SET_ERR_MSG_MOD(extack,
6006				   "Invalid flags for get route request");
6007		return -EINVAL;
6008	}
6009
6010	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
6011					    rtm_ipv6_policy, extack);
6012	if (err)
6013		return err;
6014
6015	if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
6016	    (tb[RTA_DST] && !rtm->rtm_dst_len)) {
6017		NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
6018		return -EINVAL;
6019	}
6020
6021	for (i = 0; i <= RTA_MAX; i++) {
6022		if (!tb[i])
6023			continue;
6024
6025		switch (i) {
6026		case RTA_SRC:
6027		case RTA_DST:
6028		case RTA_IIF:
6029		case RTA_OIF:
6030		case RTA_MARK:
6031		case RTA_UID:
6032		case RTA_SPORT:
6033		case RTA_DPORT:
6034		case RTA_IP_PROTO:
6035			break;
6036		default:
6037			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
6038			return -EINVAL;
6039		}
6040	}
6041
6042	return 0;
6043}
6044
6045static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
6046			      struct netlink_ext_ack *extack)
6047{
6048	struct net *net = sock_net(in_skb->sk);
6049	struct nlattr *tb[RTA_MAX+1];
6050	int err, iif = 0, oif = 0;
6051	struct fib6_info *from;
6052	struct dst_entry *dst;
6053	struct rt6_info *rt;
6054	struct sk_buff *skb;
6055	struct rtmsg *rtm;
6056	struct flowi6 fl6 = {};
6057	bool fibmatch;
6058
6059	err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
6060	if (err < 0)
6061		goto errout;
6062
6063	err = -EINVAL;
6064	rtm = nlmsg_data(nlh);
6065	fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
6066	fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
6067
6068	if (tb[RTA_SRC]) {
6069		if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
6070			goto errout;
6071
6072		fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
6073	}
6074
6075	if (tb[RTA_DST]) {
6076		if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
6077			goto errout;
6078
6079		fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
6080	}
6081
6082	if (tb[RTA_IIF])
6083		iif = nla_get_u32(tb[RTA_IIF]);
6084
6085	if (tb[RTA_OIF])
6086		oif = nla_get_u32(tb[RTA_OIF]);
6087
6088	if (tb[RTA_MARK])
6089		fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
6090
6091	if (tb[RTA_UID])
6092		fl6.flowi6_uid = make_kuid(current_user_ns(),
6093					   nla_get_u32(tb[RTA_UID]));
6094	else
6095		fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
6096
6097	if (tb[RTA_SPORT])
6098		fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
6099
6100	if (tb[RTA_DPORT])
6101		fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
6102
6103	if (tb[RTA_IP_PROTO]) {
6104		err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
6105						  &fl6.flowi6_proto, AF_INET6,
6106						  extack);
6107		if (err)
6108			goto errout;
6109	}
6110
6111	if (iif) {
6112		struct net_device *dev;
6113		int flags = 0;
6114
6115		rcu_read_lock();
6116
6117		dev = dev_get_by_index_rcu(net, iif);
6118		if (!dev) {
6119			rcu_read_unlock();
6120			err = -ENODEV;
6121			goto errout;
6122		}
6123
6124		fl6.flowi6_iif = iif;
6125
6126		if (!ipv6_addr_any(&fl6.saddr))
6127			flags |= RT6_LOOKUP_F_HAS_SADDR;
6128
6129		dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
6130
6131		rcu_read_unlock();
6132	} else {
6133		fl6.flowi6_oif = oif;
6134
6135		dst = ip6_route_output(net, NULL, &fl6);
6136	}
6137
6138
6139	rt = dst_rt6_info(dst);
6140	if (rt->dst.error) {
6141		err = rt->dst.error;
6142		ip6_rt_put(rt);
6143		goto errout;
6144	}
6145
6146	if (rt == net->ipv6.ip6_null_entry) {
6147		err = rt->dst.error;
6148		ip6_rt_put(rt);
6149		goto errout;
6150	}
6151
6152	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
6153	if (!skb) {
6154		ip6_rt_put(rt);
6155		err = -ENOBUFS;
6156		goto errout;
6157	}
6158
 
 
 
 
 
 
6159	skb_dst_set(skb, &rt->dst);
6160
6161	rcu_read_lock();
6162	from = rcu_dereference(rt->from);
6163	if (from) {
6164		if (fibmatch)
6165			err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
6166					    iif, RTM_NEWROUTE,
6167					    NETLINK_CB(in_skb).portid,
6168					    nlh->nlmsg_seq, 0);
6169		else
6170			err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
6171					    &fl6.saddr, iif, RTM_NEWROUTE,
6172					    NETLINK_CB(in_skb).portid,
6173					    nlh->nlmsg_seq, 0);
6174	} else {
6175		err = -ENETUNREACH;
6176	}
6177	rcu_read_unlock();
6178
6179	if (err < 0) {
6180		kfree_skb(skb);
6181		goto errout;
6182	}
6183
6184	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
6185errout:
6186	return err;
6187}
6188
6189void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
6190		     unsigned int nlm_flags)
6191{
6192	struct sk_buff *skb;
6193	struct net *net = info->nl_net;
6194	u32 seq;
6195	int err;
6196
6197	err = -ENOBUFS;
6198	seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6199
6200	skb = nlmsg_new(rt6_nlmsg_size(rt), GFP_ATOMIC);
6201	if (!skb)
6202		goto errout;
6203
6204	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6205			    event, info->portid, seq, nlm_flags);
6206	if (err < 0) {
6207		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6208		WARN_ON(err == -EMSGSIZE);
6209		kfree_skb(skb);
6210		goto errout;
6211	}
6212	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6213		    info->nlh, GFP_ATOMIC);
6214	return;
6215errout:
6216	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6217}
6218
6219void fib6_rt_update(struct net *net, struct fib6_info *rt,
6220		    struct nl_info *info)
6221{
6222	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
6223	struct sk_buff *skb;
6224	int err = -ENOBUFS;
6225
6226	skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
6227	if (!skb)
6228		goto errout;
6229
6230	err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
6231			    RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
6232	if (err < 0) {
6233		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6234		WARN_ON(err == -EMSGSIZE);
6235		kfree_skb(skb);
6236		goto errout;
6237	}
6238	rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
6239		    info->nlh, gfp_any());
6240	return;
6241errout:
6242	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6243}
6244
6245void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i,
6246			    bool offload, bool trap, bool offload_failed)
6247{
6248	struct sk_buff *skb;
6249	int err;
6250
6251	if (READ_ONCE(f6i->offload) == offload &&
6252	    READ_ONCE(f6i->trap) == trap &&
6253	    READ_ONCE(f6i->offload_failed) == offload_failed)
6254		return;
6255
6256	WRITE_ONCE(f6i->offload, offload);
6257	WRITE_ONCE(f6i->trap, trap);
6258
6259	/* 2 means send notifications only if offload_failed was changed. */
6260	if (net->ipv6.sysctl.fib_notify_on_flag_change == 2 &&
6261	    READ_ONCE(f6i->offload_failed) == offload_failed)
6262		return;
6263
6264	WRITE_ONCE(f6i->offload_failed, offload_failed);
6265
6266	if (!rcu_access_pointer(f6i->fib6_node))
6267		/* The route was removed from the tree, do not send
6268		 * notification.
6269		 */
6270		return;
6271
6272	if (!net->ipv6.sysctl.fib_notify_on_flag_change)
6273		return;
6274
6275	skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL);
6276	if (!skb) {
6277		err = -ENOBUFS;
6278		goto errout;
6279	}
6280
6281	err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0,
6282			    0, 0);
6283	if (err < 0) {
6284		/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
6285		WARN_ON(err == -EMSGSIZE);
6286		kfree_skb(skb);
6287		goto errout;
6288	}
6289
6290	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL);
6291	return;
6292
6293errout:
6294	rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
6295}
6296EXPORT_SYMBOL(fib6_info_hw_flags_set);
6297
6298static int ip6_route_dev_notify(struct notifier_block *this,
6299				unsigned long event, void *ptr)
6300{
6301	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6302	struct net *net = dev_net(dev);
6303
6304	if (!(dev->flags & IFF_LOOPBACK))
6305		return NOTIFY_OK;
6306
6307	if (event == NETDEV_REGISTER) {
6308		net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
6309		net->ipv6.ip6_null_entry->dst.dev = dev;
6310		net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
6311#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6312		net->ipv6.ip6_prohibit_entry->dst.dev = dev;
6313		net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
6314		net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
6315		net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
6316#endif
6317	 } else if (event == NETDEV_UNREGISTER &&
6318		    dev->reg_state != NETREG_UNREGISTERED) {
6319		/* NETDEV_UNREGISTER could be fired for multiple times by
6320		 * netdev_wait_allrefs(). Make sure we only call this once.
6321		 */
6322		in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
6323#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6324		in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
6325		in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
6326#endif
6327	}
6328
6329	return NOTIFY_OK;
6330}
6331
6332/*
6333 *	/proc
6334 */
6335
6336#ifdef CONFIG_PROC_FS
 
 
 
 
 
 
 
 
 
6337static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6338{
6339	struct net *net = (struct net *)seq->private;
6340	seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6341		   net->ipv6.rt6_stats->fib_nodes,
6342		   net->ipv6.rt6_stats->fib_route_nodes,
6343		   atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6344		   net->ipv6.rt6_stats->fib_rt_entries,
6345		   net->ipv6.rt6_stats->fib_rt_cache,
6346		   dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6347		   net->ipv6.rt6_stats->fib_discarded_routes);
6348
6349	return 0;
6350}
 
 
 
 
 
 
 
 
 
 
 
 
 
6351#endif	/* CONFIG_PROC_FS */
6352
6353#ifdef CONFIG_SYSCTL
6354
6355static int ipv6_sysctl_rtcache_flush(const struct ctl_table *ctl, int write,
6356			      void *buffer, size_t *lenp, loff_t *ppos)
 
6357{
6358	struct net *net;
6359	int delay;
6360	int ret;
6361	if (!write)
6362		return -EINVAL;
6363
6364	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6365	if (ret)
6366		return ret;
6367
6368	net = (struct net *)ctl->extra1;
6369	delay = net->ipv6.sysctl.flush_delay;
 
6370	fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6371	return 0;
6372}
6373
6374static struct ctl_table ipv6_route_table_template[] = {
6375	{
6376		.procname	=	"max_size",
6377		.data		=	&init_net.ipv6.sysctl.ip6_rt_max_size,
6378		.maxlen		=	sizeof(int),
6379		.mode		=	0644,
6380		.proc_handler	=	proc_dointvec,
6381	},
6382	{
6383		.procname	=	"gc_thresh",
6384		.data		=	&ip6_dst_ops_template.gc_thresh,
6385		.maxlen		=	sizeof(int),
6386		.mode		=	0644,
6387		.proc_handler	=	proc_dointvec,
6388	},
6389	{
6390		.procname	=	"flush",
6391		.data		=	&init_net.ipv6.sysctl.flush_delay,
6392		.maxlen		=	sizeof(int),
6393		.mode		=	0200,
6394		.proc_handler	=	ipv6_sysctl_rtcache_flush
6395	},
6396	{
6397		.procname	=	"gc_min_interval",
6398		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6399		.maxlen		=	sizeof(int),
6400		.mode		=	0644,
6401		.proc_handler	=	proc_dointvec_jiffies,
6402	},
6403	{
6404		.procname	=	"gc_timeout",
6405		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6406		.maxlen		=	sizeof(int),
6407		.mode		=	0644,
6408		.proc_handler	=	proc_dointvec_jiffies,
6409	},
6410	{
6411		.procname	=	"gc_interval",
6412		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_interval,
6413		.maxlen		=	sizeof(int),
6414		.mode		=	0644,
6415		.proc_handler	=	proc_dointvec_jiffies,
6416	},
6417	{
6418		.procname	=	"gc_elasticity",
6419		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6420		.maxlen		=	sizeof(int),
6421		.mode		=	0644,
6422		.proc_handler	=	proc_dointvec,
6423	},
6424	{
6425		.procname	=	"mtu_expires",
6426		.data		=	&init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6427		.maxlen		=	sizeof(int),
6428		.mode		=	0644,
6429		.proc_handler	=	proc_dointvec_jiffies,
6430	},
6431	{
6432		.procname	=	"min_adv_mss",
6433		.data		=	&init_net.ipv6.sysctl.ip6_rt_min_advmss,
6434		.maxlen		=	sizeof(int),
6435		.mode		=	0644,
6436		.proc_handler	=	proc_dointvec,
6437	},
6438	{
6439		.procname	=	"gc_min_interval_ms",
6440		.data		=	&init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6441		.maxlen		=	sizeof(int),
6442		.mode		=	0644,
6443		.proc_handler	=	proc_dointvec_ms_jiffies,
6444	},
6445	{
6446		.procname	=	"skip_notify_on_dev_down",
6447		.data		=	&init_net.ipv6.sysctl.skip_notify_on_dev_down,
6448		.maxlen		=	sizeof(u8),
6449		.mode		=	0644,
6450		.proc_handler	=	proc_dou8vec_minmax,
6451		.extra1		=	SYSCTL_ZERO,
6452		.extra2		=	SYSCTL_ONE,
6453	},
6454};
6455
6456struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6457{
6458	struct ctl_table *table;
6459
6460	table = kmemdup(ipv6_route_table_template,
6461			sizeof(ipv6_route_table_template),
6462			GFP_KERNEL);
6463
6464	if (table) {
6465		table[0].data = &net->ipv6.sysctl.ip6_rt_max_size;
 
6466		table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6467		table[2].data = &net->ipv6.sysctl.flush_delay;
6468		table[2].extra1 = net;
6469		table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6470		table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6471		table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6472		table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6473		table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6474		table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6475		table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6476		table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
 
 
 
6477	}
6478
6479	return table;
6480}
6481
6482size_t ipv6_route_sysctl_table_size(struct net *net)
6483{
6484	/* Don't export sysctls to unprivileged users */
6485	if (net->user_ns != &init_user_ns)
6486		return 1;
6487
6488	return ARRAY_SIZE(ipv6_route_table_template);
6489}
6490#endif
6491
6492static int __net_init ip6_route_net_init(struct net *net)
6493{
6494	int ret = -ENOMEM;
6495
6496	memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6497	       sizeof(net->ipv6.ip6_dst_ops));
6498
6499	if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6500		goto out_ip6_dst_ops;
6501
6502	net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6503	if (!net->ipv6.fib6_null_entry)
6504		goto out_ip6_dst_entries;
6505	memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6506	       sizeof(*net->ipv6.fib6_null_entry));
6507
6508	net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6509					   sizeof(*net->ipv6.ip6_null_entry),
6510					   GFP_KERNEL);
6511	if (!net->ipv6.ip6_null_entry)
6512		goto out_fib6_null_entry;
 
 
6513	net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6514	dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6515			 ip6_template_metrics, true);
6516	INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached);
6517
6518#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6519	net->ipv6.fib6_has_custom_rules = false;
6520	net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6521					       sizeof(*net->ipv6.ip6_prohibit_entry),
6522					       GFP_KERNEL);
6523	if (!net->ipv6.ip6_prohibit_entry)
6524		goto out_ip6_null_entry;
 
 
6525	net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6526	dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6527			 ip6_template_metrics, true);
6528	INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached);
6529
6530	net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6531					       sizeof(*net->ipv6.ip6_blk_hole_entry),
6532					       GFP_KERNEL);
6533	if (!net->ipv6.ip6_blk_hole_entry)
6534		goto out_ip6_prohibit_entry;
 
 
6535	net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6536	dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6537			 ip6_template_metrics, true);
6538	INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached);
6539#ifdef CONFIG_IPV6_SUBTREES
6540	net->ipv6.fib6_routes_require_src = 0;
6541#endif
6542#endif
6543
6544	net->ipv6.sysctl.flush_delay = 0;
6545	net->ipv6.sysctl.ip6_rt_max_size = INT_MAX;
6546	net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6547	net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6548	net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6549	net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6550	net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6551	net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6552	net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6553
6554	atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ);
6555
6556	ret = 0;
6557out:
6558	return ret;
6559
6560#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6561out_ip6_prohibit_entry:
6562	kfree(net->ipv6.ip6_prohibit_entry);
6563out_ip6_null_entry:
6564	kfree(net->ipv6.ip6_null_entry);
6565#endif
6566out_fib6_null_entry:
6567	kfree(net->ipv6.fib6_null_entry);
6568out_ip6_dst_entries:
6569	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6570out_ip6_dst_ops:
6571	goto out;
6572}
6573
6574static void __net_exit ip6_route_net_exit(struct net *net)
6575{
6576	kfree(net->ipv6.fib6_null_entry);
6577	kfree(net->ipv6.ip6_null_entry);
6578#ifdef CONFIG_IPV6_MULTIPLE_TABLES
6579	kfree(net->ipv6.ip6_prohibit_entry);
6580	kfree(net->ipv6.ip6_blk_hole_entry);
6581#endif
6582	dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6583}
6584
6585static int __net_init ip6_route_net_init_late(struct net *net)
6586{
6587#ifdef CONFIG_PROC_FS
6588	if (!proc_create_net("ipv6_route", 0, net->proc_net,
6589			     &ipv6_route_seq_ops,
6590			     sizeof(struct ipv6_route_iter)))
6591		return -ENOMEM;
6592
6593	if (!proc_create_net_single("rt6_stats", 0444, net->proc_net,
6594				    rt6_stats_seq_show, NULL)) {
6595		remove_proc_entry("ipv6_route", net->proc_net);
6596		return -ENOMEM;
6597	}
6598#endif
6599	return 0;
6600}
6601
6602static void __net_exit ip6_route_net_exit_late(struct net *net)
6603{
6604#ifdef CONFIG_PROC_FS
6605	remove_proc_entry("ipv6_route", net->proc_net);
6606	remove_proc_entry("rt6_stats", net->proc_net);
6607#endif
6608}
6609
6610static struct pernet_operations ip6_route_net_ops = {
6611	.init = ip6_route_net_init,
6612	.exit = ip6_route_net_exit,
6613};
6614
6615static int __net_init ipv6_inetpeer_init(struct net *net)
6616{
6617	struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6618
6619	if (!bp)
6620		return -ENOMEM;
6621	inet_peer_base_init(bp);
6622	net->ipv6.peers = bp;
6623	return 0;
6624}
6625
6626static void __net_exit ipv6_inetpeer_exit(struct net *net)
6627{
6628	struct inet_peer_base *bp = net->ipv6.peers;
6629
6630	net->ipv6.peers = NULL;
6631	inetpeer_invalidate_tree(bp);
6632	kfree(bp);
6633}
6634
6635static struct pernet_operations ipv6_inetpeer_ops = {
6636	.init	=	ipv6_inetpeer_init,
6637	.exit	=	ipv6_inetpeer_exit,
6638};
6639
6640static struct pernet_operations ip6_route_net_late_ops = {
6641	.init = ip6_route_net_init_late,
6642	.exit = ip6_route_net_exit_late,
6643};
6644
6645static struct notifier_block ip6_route_dev_notifier = {
6646	.notifier_call = ip6_route_dev_notify,
6647	.priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6648};
6649
6650void __init ip6_route_init_special_entries(void)
6651{
6652	/* Registering of the loopback is done before this portion of code,
6653	 * the loopback reference in rt6_info will not be taken, do it
6654	 * manually for init_net */
6655	init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6656	init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6657	init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6658  #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6659	init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6660	init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6661	init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6662	init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6663  #endif
6664}
6665
6666#if IS_BUILTIN(CONFIG_IPV6)
6667#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6668DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt)
6669
6670BTF_ID_LIST(btf_fib6_info_id)
6671BTF_ID(struct, fib6_info)
6672
6673static const struct bpf_iter_seq_info ipv6_route_seq_info = {
6674	.seq_ops		= &ipv6_route_seq_ops,
6675	.init_seq_private	= bpf_iter_init_seq_net,
6676	.fini_seq_private	= bpf_iter_fini_seq_net,
6677	.seq_priv_size		= sizeof(struct ipv6_route_iter),
6678};
6679
6680static struct bpf_iter_reg ipv6_route_reg_info = {
6681	.target			= "ipv6_route",
6682	.ctx_arg_info_size	= 1,
6683	.ctx_arg_info		= {
6684		{ offsetof(struct bpf_iter__ipv6_route, rt),
6685		  PTR_TO_BTF_ID_OR_NULL },
6686	},
6687	.seq_info		= &ipv6_route_seq_info,
6688};
6689
6690static int __init bpf_iter_register(void)
6691{
6692	ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id;
6693	return bpf_iter_reg_target(&ipv6_route_reg_info);
6694}
6695
6696static void bpf_iter_unregister(void)
6697{
6698	bpf_iter_unreg_target(&ipv6_route_reg_info);
6699}
6700#endif
6701#endif
6702
6703static const struct rtnl_msg_handler ip6_route_rtnl_msg_handlers[] __initconst_or_module = {
6704	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWROUTE,
6705	 .doit = inet6_rtm_newroute},
6706	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELROUTE,
6707	 .doit = inet6_rtm_delroute},
6708	{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE,
6709	 .doit = inet6_rtm_getroute, .flags = RTNL_FLAG_DOIT_UNLOCKED},
6710};
6711
6712int __init ip6_route_init(void)
6713{
6714	int ret;
6715	int cpu;
6716
6717	ret = -ENOMEM;
6718	ip6_dst_ops_template.kmem_cachep =
6719		kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6720				  SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL);
6721	if (!ip6_dst_ops_template.kmem_cachep)
6722		goto out;
6723
6724	ret = dst_entries_init(&ip6_dst_blackhole_ops);
6725	if (ret)
6726		goto out_kmem_cache;
6727
6728	ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6729	if (ret)
6730		goto out_dst_entries;
6731
6732	ret = register_pernet_subsys(&ip6_route_net_ops);
6733	if (ret)
6734		goto out_register_inetpeer;
6735
6736	ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6737
 
 
 
 
 
 
 
 
 
 
 
6738	ret = fib6_init();
6739	if (ret)
6740		goto out_register_subsys;
6741
6742	ret = xfrm6_init();
6743	if (ret)
6744		goto out_fib6_init;
6745
6746	ret = fib6_rules_init();
6747	if (ret)
6748		goto xfrm6_init;
6749
6750	ret = register_pernet_subsys(&ip6_route_net_late_ops);
6751	if (ret)
6752		goto fib6_rules_init;
6753
6754	ret = rtnl_register_many(ip6_route_rtnl_msg_handlers);
6755	if (ret < 0)
 
 
6756		goto out_register_late_subsys;
6757
6758	ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6759	if (ret)
6760		goto out_register_late_subsys;
6761
6762#if IS_BUILTIN(CONFIG_IPV6)
6763#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6764	ret = bpf_iter_register();
6765	if (ret)
6766		goto out_register_late_subsys;
6767#endif
6768#endif
6769
6770	for_each_possible_cpu(cpu) {
6771		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6772
6773		INIT_LIST_HEAD(&ul->head);
6774		spin_lock_init(&ul->lock);
6775	}
6776
6777out:
6778	return ret;
6779
6780out_register_late_subsys:
6781	rtnl_unregister_all(PF_INET6);
6782	unregister_pernet_subsys(&ip6_route_net_late_ops);
6783fib6_rules_init:
6784	fib6_rules_cleanup();
6785xfrm6_init:
6786	xfrm6_fini();
6787out_fib6_init:
6788	fib6_gc_cleanup();
6789out_register_subsys:
6790	unregister_pernet_subsys(&ip6_route_net_ops);
6791out_register_inetpeer:
6792	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6793out_dst_entries:
6794	dst_entries_destroy(&ip6_dst_blackhole_ops);
6795out_kmem_cache:
6796	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6797	goto out;
6798}
6799
6800void ip6_route_cleanup(void)
6801{
6802#if IS_BUILTIN(CONFIG_IPV6)
6803#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS)
6804	bpf_iter_unregister();
6805#endif
6806#endif
6807	unregister_netdevice_notifier(&ip6_route_dev_notifier);
6808	unregister_pernet_subsys(&ip6_route_net_late_ops);
6809	fib6_rules_cleanup();
6810	xfrm6_fini();
6811	fib6_gc_cleanup();
6812	unregister_pernet_subsys(&ipv6_inetpeer_ops);
6813	unregister_pernet_subsys(&ip6_route_net_ops);
6814	dst_entries_destroy(&ip6_dst_blackhole_ops);
6815	kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6816}