fib_rules.c

   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * net/core/fib_rules.c		Generic Routing Rules
   4 *
   5 * Authors:	Thomas Graf <tgraf@suug.ch>
   6 */
   7
   8#include <linux/types.h>
   9#include <linux/kernel.h>
  10#include <linux/slab.h>
  11#include <linux/list.h>
  12#include <linux/module.h>
  13#include <net/net_namespace.h>
  14#include <net/inet_dscp.h>
  15#include <net/sock.h>
  16#include <net/fib_rules.h>
  17#include <net/ip_tunnels.h>
  18#include <linux/indirect_call_wrapper.h>
  19
  20#if defined(CONFIG_IPV6) && defined(CONFIG_IPV6_MULTIPLE_TABLES)
  21#ifdef CONFIG_IP_MULTIPLE_TABLES
  22#define INDIRECT_CALL_MT(f, f2, f1, ...) \
  23	INDIRECT_CALL_INET(f, f2, f1, __VA_ARGS__)
  24#else
  25#define INDIRECT_CALL_MT(f, f2, f1, ...) INDIRECT_CALL_1(f, f2, __VA_ARGS__)
  26#endif
  27#elif defined(CONFIG_IP_MULTIPLE_TABLES)
  28#define INDIRECT_CALL_MT(f, f2, f1, ...) INDIRECT_CALL_1(f, f1, __VA_ARGS__)
  29#else
  30#define INDIRECT_CALL_MT(f, f2, f1, ...) f(__VA_ARGS__)
  31#endif
  32
  33static const struct fib_kuid_range fib_kuid_range_unset = {
  34	KUIDT_INIT(0),
  35	KUIDT_INIT(~0),
  36};
  37
  38bool fib_rule_matchall(const struct fib_rule *rule)
  39{
  40	if (READ_ONCE(rule->iifindex) || READ_ONCE(rule->oifindex) ||
  41	    rule->mark || rule->tun_id || rule->flags)
  42		return false;
  43	if (rule->suppress_ifgroup != -1 || rule->suppress_prefixlen != -1)
  44		return false;
  45	if (!uid_eq(rule->uid_range.start, fib_kuid_range_unset.start) ||
  46	    !uid_eq(rule->uid_range.end, fib_kuid_range_unset.end))
  47		return false;
  48	if (fib_rule_port_range_set(&rule->sport_range))
  49		return false;
  50	if (fib_rule_port_range_set(&rule->dport_range))
  51		return false;
  52	return true;
  53}
  54EXPORT_SYMBOL_GPL(fib_rule_matchall);
  55
  56int fib_default_rule_add(struct fib_rules_ops *ops,
  57			 u32 pref, u32 table)
  58{
  59	struct fib_rule *r;
  60
  61	r = kzalloc(ops->rule_size, GFP_KERNEL_ACCOUNT);
  62	if (r == NULL)
  63		return -ENOMEM;
  64
  65	refcount_set(&r->refcnt, 1);
  66	r->action = FR_ACT_TO_TBL;
  67	r->pref = pref;
  68	r->table = table;
  69	r->proto = RTPROT_KERNEL;
  70	r->fr_net = ops->fro_net;
  71	r->uid_range = fib_kuid_range_unset;
  72
  73	r->suppress_prefixlen = -1;
  74	r->suppress_ifgroup = -1;
  75
  76	/* The lock is not required here, the list in unreachable
  77	 * at the moment this function is called */
  78	list_add_tail(&r->list, &ops->rules_list);
  79	return 0;
  80}
  81EXPORT_SYMBOL(fib_default_rule_add);
  82
  83static u32 fib_default_rule_pref(struct fib_rules_ops *ops)
  84{
  85	struct list_head *pos;
  86	struct fib_rule *rule;
  87
  88	if (!list_empty(&ops->rules_list)) {
  89		pos = ops->rules_list.next;
  90		if (pos->next != &ops->rules_list) {
  91			rule = list_entry(pos->next, struct fib_rule, list);
  92			if (rule->pref)
  93				return rule->pref - 1;
  94		}
  95	}
  96
  97	return 0;
  98}
  99
 100static void notify_rule_change(int event, struct fib_rule *rule,
 101			       struct fib_rules_ops *ops, struct nlmsghdr *nlh,
 102			       u32 pid);
 103
 104static struct fib_rules_ops *lookup_rules_ops(const struct net *net,
 105					      int family)
 106{
 107	struct fib_rules_ops *ops;
 108
 109	rcu_read_lock();
 110	list_for_each_entry_rcu(ops, &net->rules_ops, list) {
 111		if (ops->family == family) {
 112			if (!try_module_get(ops->owner))
 113				ops = NULL;
 114			rcu_read_unlock();
 115			return ops;
 116		}
 117	}
 118	rcu_read_unlock();
 119
 120	return NULL;
 121}
 122
 123static void rules_ops_put(struct fib_rules_ops *ops)
 124{
 125	if (ops)
 126		module_put(ops->owner);
 127}
 128
 129static void flush_route_cache(struct fib_rules_ops *ops)
 130{
 131	if (ops->flush_cache)
 132		ops->flush_cache(ops);
 133}
 134
 135static int __fib_rules_register(struct fib_rules_ops *ops)
 136{
 137	int err = -EEXIST;
 138	struct fib_rules_ops *o;
 139	struct net *net;
 140
 141	net = ops->fro_net;
 142
 143	if (ops->rule_size < sizeof(struct fib_rule))
 144		return -EINVAL;
 145
 146	if (ops->match == NULL || ops->configure == NULL ||
 147	    ops->compare == NULL || ops->fill == NULL ||
 148	    ops->action == NULL)
 149		return -EINVAL;
 150
 151	spin_lock(&net->rules_mod_lock);
 152	list_for_each_entry(o, &net->rules_ops, list)
 153		if (ops->family == o->family)
 154			goto errout;
 155
 156	list_add_tail_rcu(&ops->list, &net->rules_ops);
 157	err = 0;
 158errout:
 159	spin_unlock(&net->rules_mod_lock);
 160
 161	return err;
 162}
 163
 164struct fib_rules_ops *
 165fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net)
 166{
 167	struct fib_rules_ops *ops;
 168	int err;
 169
 170	ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL);
 171	if (ops == NULL)
 172		return ERR_PTR(-ENOMEM);
 173
 174	INIT_LIST_HEAD(&ops->rules_list);
 175	ops->fro_net = net;
 176
 177	err = __fib_rules_register(ops);
 178	if (err) {
 179		kfree(ops);
 180		ops = ERR_PTR(err);
 181	}
 182
 183	return ops;
 184}
 185EXPORT_SYMBOL_GPL(fib_rules_register);
 186
 187static void fib_rules_cleanup_ops(struct fib_rules_ops *ops)
 188{
 189	struct fib_rule *rule, *tmp;
 190
 191	list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) {
 192		list_del_rcu(&rule->list);
 193		if (ops->delete)
 194			ops->delete(rule);
 195		fib_rule_put(rule);
 196	}
 197}
 198
 199void fib_rules_unregister(struct fib_rules_ops *ops)
 200{
 201	struct net *net = ops->fro_net;
 202
 203	spin_lock(&net->rules_mod_lock);
 204	list_del_rcu(&ops->list);
 205	spin_unlock(&net->rules_mod_lock);
 206
 207	fib_rules_cleanup_ops(ops);
 208	kfree_rcu(ops, rcu);
 209}
 210EXPORT_SYMBOL_GPL(fib_rules_unregister);
 211
 212static int uid_range_set(struct fib_kuid_range *range)
 213{
 214	return uid_valid(range->start) && uid_valid(range->end);
 215}
 216
 217static struct fib_kuid_range nla_get_kuid_range(struct nlattr **tb)
 218{
 219	struct fib_rule_uid_range *in;
 220	struct fib_kuid_range out;
 221
 222	in = (struct fib_rule_uid_range *)nla_data(tb[FRA_UID_RANGE]);
 223
 224	out.start = make_kuid(current_user_ns(), in->start);
 225	out.end = make_kuid(current_user_ns(), in->end);
 226
 227	return out;
 228}
 229
 230static int nla_put_uid_range(struct sk_buff *skb, struct fib_kuid_range *range)
 231{
 232	struct fib_rule_uid_range out = {
 233		from_kuid_munged(current_user_ns(), range->start),
 234		from_kuid_munged(current_user_ns(), range->end)
 235	};
 236
 237	return nla_put(skb, FRA_UID_RANGE, sizeof(out), &out);
 238}
 239
 240static int nla_get_port_range(struct nlattr *pattr,
 241			      struct fib_rule_port_range *port_range)
 242{
 243	const struct fib_rule_port_range *pr = nla_data(pattr);
 244
 245	if (!fib_rule_port_range_valid(pr))
 246		return -EINVAL;
 247
 248	port_range->start = pr->start;
 249	port_range->end = pr->end;
 250
 251	return 0;
 252}
 253
 254static int nla_put_port_range(struct sk_buff *skb, int attrtype,
 255			      struct fib_rule_port_range *range)
 256{
 257	return nla_put(skb, attrtype, sizeof(*range), range);
 258}
 259
 260static bool fib_rule_iif_match(const struct fib_rule *rule, int iifindex,
 261			       const struct flowi *fl)
 262{
 263	u8 iif_is_l3_master = READ_ONCE(rule->iif_is_l3_master);
 264
 265	return iif_is_l3_master ? l3mdev_fib_rule_iif_match(fl, iifindex) :
 266				  fl->flowi_iif == iifindex;
 267}
 268
 269static bool fib_rule_oif_match(const struct fib_rule *rule, int oifindex,
 270			       const struct flowi *fl)
 271{
 272	u8 oif_is_l3_master = READ_ONCE(rule->oif_is_l3_master);
 273
 274	return oif_is_l3_master ? l3mdev_fib_rule_oif_match(fl, oifindex) :
 275				  fl->flowi_oif == oifindex;
 276}
 277
 278static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops,
 279			  struct flowi *fl, int flags,
 280			  struct fib_lookup_arg *arg)
 281{
 282	int iifindex, oifindex, ret = 0;
 283
 284	iifindex = READ_ONCE(rule->iifindex);
 285	if (iifindex && !fib_rule_iif_match(rule, iifindex, fl))
 286		goto out;
 287
 288	oifindex = READ_ONCE(rule->oifindex);
 289	if (oifindex && !fib_rule_oif_match(rule, oifindex, fl))
 290		goto out;
 291
 292	if ((rule->mark ^ fl->flowi_mark) & rule->mark_mask)
 293		goto out;
 294
 295	if (rule->tun_id && (rule->tun_id != fl->flowi_tun_key.tun_id))
 296		goto out;
 297
 298	if (rule->l3mdev && !l3mdev_fib_rule_match(rule->fr_net, fl, arg))
 299		goto out;
 300
 301	if (uid_lt(fl->flowi_uid, rule->uid_range.start) ||
 302	    uid_gt(fl->flowi_uid, rule->uid_range.end))
 303		goto out;
 304
 305	ret = INDIRECT_CALL_MT(ops->match,
 306			       fib6_rule_match,
 307			       fib4_rule_match,
 308			       rule, fl, flags);
 309out:
 310	return (rule->flags & FIB_RULE_INVERT) ? !ret : ret;
 311}
 312
 313int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl,
 314		     int flags, struct fib_lookup_arg *arg)
 315{
 316	struct fib_rule *rule;
 317	int err;
 318
 319	rcu_read_lock();
 320
 321	list_for_each_entry_rcu(rule, &ops->rules_list, list) {
 322jumped:
 323		if (!fib_rule_match(rule, ops, fl, flags, arg))
 324			continue;
 325
 326		if (rule->action == FR_ACT_GOTO) {
 327			struct fib_rule *target;
 328
 329			target = rcu_dereference(rule->ctarget);
 330			if (target == NULL) {
 331				continue;
 332			} else {
 333				rule = target;
 334				goto jumped;
 335			}
 336		} else if (rule->action == FR_ACT_NOP)
 337			continue;
 338		else
 339			err = INDIRECT_CALL_MT(ops->action,
 340					       fib6_rule_action,
 341					       fib4_rule_action,
 342					       rule, fl, flags, arg);
 343
 344		if (!err && ops->suppress && INDIRECT_CALL_MT(ops->suppress,
 345							      fib6_rule_suppress,
 346							      fib4_rule_suppress,
 347							      rule, flags, arg))
 348			continue;
 349
 350		if (err != -EAGAIN) {
 351			if ((arg->flags & FIB_LOOKUP_NOREF) ||
 352			    likely(refcount_inc_not_zero(&rule->refcnt))) {
 353				arg->rule = rule;
 354				goto out;
 355			}
 356			break;
 357		}
 358	}
 359
 360	err = -ESRCH;
 361out:
 362	rcu_read_unlock();
 363
 364	return err;
 365}
 366EXPORT_SYMBOL_GPL(fib_rules_lookup);
 367
 368static int call_fib_rule_notifier(struct notifier_block *nb,
 369				  enum fib_event_type event_type,
 370				  struct fib_rule *rule, int family,
 371				  struct netlink_ext_ack *extack)
 372{
 373	struct fib_rule_notifier_info info = {
 374		.info.family = family,
 375		.info.extack = extack,
 376		.rule = rule,
 377	};
 378
 379	return call_fib_notifier(nb, event_type, &info.info);
 380}
 381
 382static int call_fib_rule_notifiers(struct net *net,
 383				   enum fib_event_type event_type,
 384				   struct fib_rule *rule,
 385				   struct fib_rules_ops *ops,
 386				   struct netlink_ext_ack *extack)
 387{
 388	struct fib_rule_notifier_info info = {
 389		.info.family = ops->family,
 390		.info.extack = extack,
 391		.rule = rule,
 392	};
 393
 394	ASSERT_RTNL_NET(net);
 395
 396	/* Paired with READ_ONCE() in fib_rules_seq() */
 397	WRITE_ONCE(ops->fib_rules_seq, ops->fib_rules_seq + 1);
 398	return call_fib_notifiers(net, event_type, &info.info);
 399}
 400
 401/* Called with rcu_read_lock() */
 402int fib_rules_dump(struct net *net, struct notifier_block *nb, int family,
 403		   struct netlink_ext_ack *extack)
 404{
 405	struct fib_rules_ops *ops;
 406	struct fib_rule *rule;
 407	int err = 0;
 408
 409	ops = lookup_rules_ops(net, family);
 410	if (!ops)
 411		return -EAFNOSUPPORT;
 412	list_for_each_entry_rcu(rule, &ops->rules_list, list) {
 413		err = call_fib_rule_notifier(nb, FIB_EVENT_RULE_ADD,
 414					     rule, family, extack);
 415		if (err)
 416			break;
 417	}
 418	rules_ops_put(ops);
 419
 420	return err;
 421}
 422EXPORT_SYMBOL_GPL(fib_rules_dump);
 423
 424unsigned int fib_rules_seq_read(const struct net *net, int family)
 425{
 426	unsigned int fib_rules_seq;
 427	struct fib_rules_ops *ops;
 428
 429	ops = lookup_rules_ops(net, family);
 430	if (!ops)
 431		return 0;
 432	/* Paired with WRITE_ONCE() in call_fib_rule_notifiers() */
 433	fib_rules_seq = READ_ONCE(ops->fib_rules_seq);
 434	rules_ops_put(ops);
 435
 436	return fib_rules_seq;
 437}
 438EXPORT_SYMBOL_GPL(fib_rules_seq_read);
 439
 440static struct fib_rule *rule_find(struct fib_rules_ops *ops,
 441				  struct fib_rule_hdr *frh,
 442				  struct nlattr **tb,
 443				  struct fib_rule *rule,
 444				  bool user_priority)
 445{
 446	struct fib_rule *r;
 447
 448	list_for_each_entry(r, &ops->rules_list, list) {
 449		if (rule->action && r->action != rule->action)
 450			continue;
 451
 452		if (rule->table && r->table != rule->table)
 453			continue;
 454
 455		if (user_priority && r->pref != rule->pref)
 456			continue;
 457
 458		if (rule->iifname[0] &&
 459		    memcmp(r->iifname, rule->iifname, IFNAMSIZ))
 460			continue;
 461
 462		if (rule->oifname[0] &&
 463		    memcmp(r->oifname, rule->oifname, IFNAMSIZ))
 464			continue;
 465
 466		if (rule->mark && r->mark != rule->mark)
 467			continue;
 468
 469		if (rule->suppress_ifgroup != -1 &&
 470		    r->suppress_ifgroup != rule->suppress_ifgroup)
 471			continue;
 472
 473		if (rule->suppress_prefixlen != -1 &&
 474		    r->suppress_prefixlen != rule->suppress_prefixlen)
 475			continue;
 476
 477		if (rule->mark_mask && r->mark_mask != rule->mark_mask)
 478			continue;
 479
 480		if (rule->tun_id && r->tun_id != rule->tun_id)
 481			continue;
 482
 483		if (rule->l3mdev && r->l3mdev != rule->l3mdev)
 484			continue;
 485
 486		if (uid_range_set(&rule->uid_range) &&
 487		    (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
 488		    !uid_eq(r->uid_range.end, rule->uid_range.end)))
 489			continue;
 490
 491		if (rule->ip_proto && r->ip_proto != rule->ip_proto)
 492			continue;
 493
 494		if (rule->proto && r->proto != rule->proto)
 495			continue;
 496
 497		if (fib_rule_port_range_set(&rule->sport_range) &&
 498		    !fib_rule_port_range_compare(&r->sport_range,
 499						 &rule->sport_range))
 500			continue;
 501
 502		if (rule->sport_mask && r->sport_mask != rule->sport_mask)
 503			continue;
 504
 505		if (fib_rule_port_range_set(&rule->dport_range) &&
 506		    !fib_rule_port_range_compare(&r->dport_range,
 507						 &rule->dport_range))
 508			continue;
 509
 510		if (rule->dport_mask && r->dport_mask != rule->dport_mask)
 511			continue;
 512
 513		if (!ops->compare(r, frh, tb))
 514			continue;
 515		return r;
 516	}
 517
 518	return NULL;
 519}
 520
 521#ifdef CONFIG_NET_L3_MASTER_DEV
 522static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
 523			      struct netlink_ext_ack *extack)
 524{
 525	nlrule->l3mdev = nla_get_u8(nla);
 526	if (nlrule->l3mdev != 1) {
 527		NL_SET_ERR_MSG(extack, "Invalid l3mdev attribute");
 528		return -1;
 529	}
 530
 531	return 0;
 532}
 533#else
 534static int fib_nl2rule_l3mdev(struct nlattr *nla, struct fib_rule *nlrule,
 535			      struct netlink_ext_ack *extack)
 536{
 537	NL_SET_ERR_MSG(extack, "l3mdev support is not enabled in kernel");
 538	return -1;
 539}
 540#endif
 541
 542static int fib_nl2rule_port_mask(const struct nlattr *mask_attr,
 543				 const struct fib_rule_port_range *range,
 544				 u16 *port_mask,
 545				 struct netlink_ext_ack *extack)
 546{
 547	if (!fib_rule_port_range_valid(range)) {
 548		NL_SET_ERR_MSG_ATTR(extack, mask_attr,
 549				    "Cannot specify port mask without port value");
 550		return -EINVAL;
 551	}
 552
 553	if (fib_rule_port_is_range(range)) {
 554		NL_SET_ERR_MSG_ATTR(extack, mask_attr,
 555				    "Cannot specify port mask for port range");
 556		return -EINVAL;
 557	}
 558
 559	if (range->start & ~nla_get_u16(mask_attr)) {
 560		NL_SET_ERR_MSG_ATTR(extack, mask_attr, "Invalid port mask");
 561		return -EINVAL;
 562	}
 563
 564	*port_mask = nla_get_u16(mask_attr);
 565
 566	return 0;
 567}
 568
 569static int fib_nl2rule(struct net *net, struct nlmsghdr *nlh,
 570		       struct netlink_ext_ack *extack,
 571		       struct fib_rules_ops *ops,
 572		       struct nlattr *tb[],
 573		       struct fib_rule **rule,
 574		       bool *user_priority)
 575{
 576	struct fib_rule_hdr *frh = nlmsg_data(nlh);
 577	struct fib_rule *nlrule = NULL;
 578	int err = -EINVAL;
 579
 580	if (frh->src_len)
 581		if (!tb[FRA_SRC] ||
 582		    frh->src_len > (ops->addr_size * 8) ||
 583		    nla_len(tb[FRA_SRC]) != ops->addr_size) {
 584			NL_SET_ERR_MSG(extack, "Invalid source address");
 585			goto errout;
 586	}
 587
 588	if (frh->dst_len)
 589		if (!tb[FRA_DST] ||
 590		    frh->dst_len > (ops->addr_size * 8) ||
 591		    nla_len(tb[FRA_DST]) != ops->addr_size) {
 592			NL_SET_ERR_MSG(extack, "Invalid dst address");
 593			goto errout;
 594	}
 595
 596	nlrule = kzalloc(ops->rule_size, GFP_KERNEL_ACCOUNT);
 597	if (!nlrule) {
 598		err = -ENOMEM;
 599		goto errout;
 600	}
 601	refcount_set(&nlrule->refcnt, 1);
 602	nlrule->fr_net = net;
 603
 604	if (tb[FRA_PRIORITY]) {
 605		nlrule->pref = nla_get_u32(tb[FRA_PRIORITY]);
 606		*user_priority = true;
 607	}
 608
 609	nlrule->proto = nla_get_u8_default(tb[FRA_PROTOCOL], RTPROT_UNSPEC);
 610
 611	if (tb[FRA_IIFNAME]) {
 612		nlrule->iifindex = -1;
 613		nla_strscpy(nlrule->iifname, tb[FRA_IIFNAME], IFNAMSIZ);
 614	}
 615
 616	if (tb[FRA_OIFNAME]) {
 617		nlrule->oifindex = -1;
 618		nla_strscpy(nlrule->oifname, tb[FRA_OIFNAME], IFNAMSIZ);
 619	}
 620
 621	if (tb[FRA_FWMARK]) {
 622		nlrule->mark = nla_get_u32(tb[FRA_FWMARK]);
 623		if (nlrule->mark)
 624			/* compatibility: if the mark value is non-zero all bits
 625			 * are compared unless a mask is explicitly specified.
 626			 */
 627			nlrule->mark_mask = 0xFFFFFFFF;
 628	}
 629
 630	if (tb[FRA_FWMASK])
 631		nlrule->mark_mask = nla_get_u32(tb[FRA_FWMASK]);
 632
 633	if (tb[FRA_TUN_ID])
 634		nlrule->tun_id = nla_get_be64(tb[FRA_TUN_ID]);
 635
 636	if (tb[FRA_L3MDEV] &&
 637	    fib_nl2rule_l3mdev(tb[FRA_L3MDEV], nlrule, extack) < 0)
 638		goto errout_free;
 639
 640	nlrule->action = frh->action;
 641	nlrule->flags = frh->flags;
 642	nlrule->table = frh_get_table(frh, tb);
 643	if (tb[FRA_SUPPRESS_PREFIXLEN])
 644		nlrule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]);
 645	else
 646		nlrule->suppress_prefixlen = -1;
 647
 648	if (tb[FRA_SUPPRESS_IFGROUP])
 649		nlrule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]);
 650	else
 651		nlrule->suppress_ifgroup = -1;
 652
 653	if (tb[FRA_GOTO]) {
 654		if (nlrule->action != FR_ACT_GOTO) {
 655			NL_SET_ERR_MSG(extack, "Unexpected goto");
 656			goto errout_free;
 657		}
 658
 659		nlrule->target = nla_get_u32(tb[FRA_GOTO]);
 660	} else if (nlrule->action == FR_ACT_GOTO) {
 661		NL_SET_ERR_MSG(extack, "Missing goto target for action goto");
 662		goto errout_free;
 663	}
 664
 665	if (nlrule->l3mdev && nlrule->table) {
 666		NL_SET_ERR_MSG(extack, "l3mdev and table are mutually exclusive");
 667		goto errout_free;
 668	}
 669
 670	if (tb[FRA_UID_RANGE]) {
 671		if (current_user_ns() != net->user_ns) {
 672			err = -EPERM;
 673			NL_SET_ERR_MSG(extack, "No permission to set uid");
 674			goto errout_free;
 675		}
 676
 677		nlrule->uid_range = nla_get_kuid_range(tb);
 678
 679		if (!uid_range_set(&nlrule->uid_range) ||
 680		    !uid_lte(nlrule->uid_range.start, nlrule->uid_range.end)) {
 681			NL_SET_ERR_MSG(extack, "Invalid uid range");
 682			goto errout_free;
 683		}
 684	} else {
 685		nlrule->uid_range = fib_kuid_range_unset;
 686	}
 687
 688	if (tb[FRA_IP_PROTO])
 689		nlrule->ip_proto = nla_get_u8(tb[FRA_IP_PROTO]);
 690
 691	if (tb[FRA_SPORT_RANGE]) {
 692		err = nla_get_port_range(tb[FRA_SPORT_RANGE],
 693					 &nlrule->sport_range);
 694		if (err) {
 695			NL_SET_ERR_MSG(extack, "Invalid sport range");
 696			goto errout_free;
 697		}
 698		if (!fib_rule_port_is_range(&nlrule->sport_range))
 699			nlrule->sport_mask = U16_MAX;
 700	}
 701
 702	if (tb[FRA_SPORT_MASK]) {
 703		err = fib_nl2rule_port_mask(tb[FRA_SPORT_MASK],
 704					    &nlrule->sport_range,
 705					    &nlrule->sport_mask, extack);
 706		if (err)
 707			goto errout_free;
 708	}
 709
 710	if (tb[FRA_DPORT_RANGE]) {
 711		err = nla_get_port_range(tb[FRA_DPORT_RANGE],
 712					 &nlrule->dport_range);
 713		if (err) {
 714			NL_SET_ERR_MSG(extack, "Invalid dport range");
 715			goto errout_free;
 716		}
 717		if (!fib_rule_port_is_range(&nlrule->dport_range))
 718			nlrule->dport_mask = U16_MAX;
 719	}
 720
 721	if (tb[FRA_DPORT_MASK]) {
 722		err = fib_nl2rule_port_mask(tb[FRA_DPORT_MASK],
 723					    &nlrule->dport_range,
 724					    &nlrule->dport_mask, extack);
 725		if (err)
 726			goto errout_free;
 727	}
 728
 729	*rule = nlrule;
 730
 731	return 0;
 732
 733errout_free:
 734	kfree(nlrule);
 735errout:
 736	return err;
 737}
 738
 739static int fib_nl2rule_rtnl(struct fib_rule *nlrule,
 740			    struct fib_rules_ops *ops,
 741			    struct nlattr *tb[],
 742			    struct netlink_ext_ack *extack)
 743{
 744	if (!tb[FRA_PRIORITY])
 745		nlrule->pref = fib_default_rule_pref(ops);
 746
 747	/* Backward jumps are prohibited to avoid endless loops */
 748	if (tb[FRA_GOTO] && nlrule->target <= nlrule->pref) {
 749		NL_SET_ERR_MSG(extack, "Backward goto not supported");
 750		return -EINVAL;
 751	}
 752
 753	if (tb[FRA_IIFNAME]) {
 754		struct net_device *dev;
 755
 756		dev = __dev_get_by_name(nlrule->fr_net, nlrule->iifname);
 757		if (dev) {
 758			nlrule->iifindex = dev->ifindex;
 759			nlrule->iif_is_l3_master = netif_is_l3_master(dev);
 760		}
 761	}
 762
 763	if (tb[FRA_OIFNAME]) {
 764		struct net_device *dev;
 765
 766		dev = __dev_get_by_name(nlrule->fr_net, nlrule->oifname);
 767		if (dev) {
 768			nlrule->oifindex = dev->ifindex;
 769			nlrule->oif_is_l3_master = netif_is_l3_master(dev);
 770		}
 771	}
 772
 773	return 0;
 774}
 775
 776static int rule_exists(struct fib_rules_ops *ops, struct fib_rule_hdr *frh,
 777		       struct nlattr **tb, struct fib_rule *rule)
 778{
 779	struct fib_rule *r;
 780
 781	list_for_each_entry(r, &ops->rules_list, list) {
 782		if (r->action != rule->action)
 783			continue;
 784
 785		if (r->table != rule->table)
 786			continue;
 787
 788		if (r->pref != rule->pref)
 789			continue;
 790
 791		if (memcmp(r->iifname, rule->iifname, IFNAMSIZ))
 792			continue;
 793
 794		if (memcmp(r->oifname, rule->oifname, IFNAMSIZ))
 795			continue;
 796
 797		if (r->mark != rule->mark)
 798			continue;
 799
 800		if (r->suppress_ifgroup != rule->suppress_ifgroup)
 801			continue;
 802
 803		if (r->suppress_prefixlen != rule->suppress_prefixlen)
 804			continue;
 805
 806		if (r->mark_mask != rule->mark_mask)
 807			continue;
 808
 809		if (r->tun_id != rule->tun_id)
 810			continue;
 811
 812		if (r->l3mdev != rule->l3mdev)
 813			continue;
 814
 815		if (!uid_eq(r->uid_range.start, rule->uid_range.start) ||
 816		    !uid_eq(r->uid_range.end, rule->uid_range.end))
 817			continue;
 818
 819		if (r->ip_proto != rule->ip_proto)
 820			continue;
 821
 822		if (r->proto != rule->proto)
 823			continue;
 824
 825		if (!fib_rule_port_range_compare(&r->sport_range,
 826						 &rule->sport_range))
 827			continue;
 828
 829		if (r->sport_mask != rule->sport_mask)
 830			continue;
 831
 832		if (!fib_rule_port_range_compare(&r->dport_range,
 833						 &rule->dport_range))
 834			continue;
 835
 836		if (r->dport_mask != rule->dport_mask)
 837			continue;
 838
 839		if (!ops->compare(r, frh, tb))
 840			continue;
 841		return 1;
 842	}
 843	return 0;
 844}
 845
 846static const struct nla_policy fib_rule_policy[FRA_MAX + 1] = {
 847	[FRA_UNSPEC]	= { .strict_start_type = FRA_DPORT_RANGE + 1 },
 848	[FRA_IIFNAME]	= { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
 849	[FRA_OIFNAME]	= { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
 850	[FRA_PRIORITY]	= { .type = NLA_U32 },
 851	[FRA_FWMARK]	= { .type = NLA_U32 },
 852	[FRA_FLOW]	= { .type = NLA_U32 },
 853	[FRA_TUN_ID]	= { .type = NLA_U64 },
 854	[FRA_FWMASK]	= { .type = NLA_U32 },
 855	[FRA_TABLE]     = { .type = NLA_U32 },
 856	[FRA_SUPPRESS_PREFIXLEN] = { .type = NLA_U32 },
 857	[FRA_SUPPRESS_IFGROUP] = { .type = NLA_U32 },
 858	[FRA_GOTO]	= { .type = NLA_U32 },
 859	[FRA_L3MDEV]	= { .type = NLA_U8 },
 860	[FRA_UID_RANGE]	= { .len = sizeof(struct fib_rule_uid_range) },
 861	[FRA_PROTOCOL]  = { .type = NLA_U8 },
 862	[FRA_IP_PROTO]  = { .type = NLA_U8 },
 863	[FRA_SPORT_RANGE] = { .len = sizeof(struct fib_rule_port_range) },
 864	[FRA_DPORT_RANGE] = { .len = sizeof(struct fib_rule_port_range) },
 865	[FRA_DSCP]	= NLA_POLICY_MAX(NLA_U8, INET_DSCP_MASK >> 2),
 866	[FRA_FLOWLABEL] = { .type = NLA_BE32 },
 867	[FRA_FLOWLABEL_MASK] = { .type = NLA_BE32 },
 868	[FRA_SPORT_MASK] = { .type = NLA_U16 },
 869	[FRA_DPORT_MASK] = { .type = NLA_U16 },
 870	[FRA_DSCP_MASK] = NLA_POLICY_MASK(NLA_U8, INET_DSCP_MASK >> 2),
 871};
 872
 873int fib_newrule(struct net *net, struct sk_buff *skb, struct nlmsghdr *nlh,
 874		struct netlink_ext_ack *extack, bool rtnl_held)
 875{
 876	struct fib_rule *rule = NULL, *r, *last = NULL;
 877	int err = -EINVAL, unresolved = 0;
 878	struct fib_rules_ops *ops = NULL;
 879	struct nlattr *tb[FRA_MAX + 1];
 880	bool user_priority = false;
 881	struct fib_rule_hdr *frh;
 882
 883	frh = nlmsg_payload(nlh, sizeof(*frh));
 884	if (!frh) {
 885		NL_SET_ERR_MSG(extack, "Invalid msg length");
 886		goto errout;
 887	}
 888
 889	ops = lookup_rules_ops(net, frh->family);
 890	if (!ops) {
 891		err = -EAFNOSUPPORT;
 892		NL_SET_ERR_MSG(extack, "Rule family not supported");
 893		goto errout;
 894	}
 895
 896	err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
 897				     fib_rule_policy, extack);
 898	if (err < 0) {
 899		NL_SET_ERR_MSG(extack, "Error parsing msg");
 900		goto errout;
 901	}
 902
 903	err = fib_nl2rule(net, nlh, extack, ops, tb, &rule, &user_priority);
 904	if (err)
 905		goto errout;
 906
 907	if (!rtnl_held)
 908		rtnl_net_lock(net);
 909
 910	err = fib_nl2rule_rtnl(rule, ops, tb, extack);
 911	if (err)
 912		goto errout_free;
 913
 914	if ((nlh->nlmsg_flags & NLM_F_EXCL) &&
 915	    rule_exists(ops, frh, tb, rule)) {
 916		err = -EEXIST;
 917		goto errout_free;
 918	}
 919
 920	err = ops->configure(rule, skb, frh, tb, extack);
 921	if (err < 0)
 922		goto errout_free;
 923
 924	err = call_fib_rule_notifiers(net, FIB_EVENT_RULE_ADD, rule, ops,
 925				      extack);
 926	if (err < 0)
 927		goto errout_free;
 928
 929	list_for_each_entry(r, &ops->rules_list, list) {
 930		if (r->pref == rule->target) {
 931			RCU_INIT_POINTER(rule->ctarget, r);
 932			break;
 933		}
 934	}
 935
 936	if (rcu_dereference_protected(rule->ctarget, 1) == NULL)
 937		unresolved = 1;
 938
 939	list_for_each_entry(r, &ops->rules_list, list) {
 940		if (r->pref > rule->pref)
 941			break;
 942		last = r;
 943	}
 944
 945	if (last)
 946		list_add_rcu(&rule->list, &last->list);
 947	else
 948		list_add_rcu(&rule->list, &ops->rules_list);
 949
 950	if (ops->unresolved_rules) {
 951		/*
 952		 * There are unresolved goto rules in the list, check if
 953		 * any of them are pointing to this new rule.
 954		 */
 955		list_for_each_entry(r, &ops->rules_list, list) {
 956			if (r->action == FR_ACT_GOTO &&
 957			    r->target == rule->pref &&
 958			    rtnl_dereference(r->ctarget) == NULL) {
 959				rcu_assign_pointer(r->ctarget, rule);
 960				if (--ops->unresolved_rules == 0)
 961					break;
 962			}
 963		}
 964	}
 965
 966	if (rule->action == FR_ACT_GOTO)
 967		ops->nr_goto_rules++;
 968
 969	if (unresolved)
 970		ops->unresolved_rules++;
 971
 972	if (rule->tun_id)
 973		ip_tunnel_need_metadata();
 974
 975	fib_rule_get(rule);
 976
 977	if (!rtnl_held)
 978		rtnl_net_unlock(net);
 979
 980	notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid);
 981	fib_rule_put(rule);
 982	flush_route_cache(ops);
 983	rules_ops_put(ops);
 984	return 0;
 985
 986errout_free:
 987	if (!rtnl_held)
 988		rtnl_net_unlock(net);
 989	kfree(rule);
 990errout:
 991	rules_ops_put(ops);
 992	return err;
 993}
 994EXPORT_SYMBOL_GPL(fib_newrule);
 995
 996static int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr *nlh,
 997			  struct netlink_ext_ack *extack)
 998{
 999	return fib_newrule(sock_net(skb->sk), skb, nlh, extack, false);
1000}
1001
1002int fib_delrule(struct net *net, struct sk_buff *skb, struct nlmsghdr *nlh,
1003		struct netlink_ext_ack *extack, bool rtnl_held)
1004{
1005	struct fib_rule *rule = NULL, *nlrule = NULL;
1006	struct fib_rules_ops *ops = NULL;
1007	struct nlattr *tb[FRA_MAX+1];
1008	bool user_priority = false;
1009	struct fib_rule_hdr *frh;
1010	int err = -EINVAL;
1011
1012	frh = nlmsg_payload(nlh, sizeof(*frh));
1013	if (!frh) {
1014		NL_SET_ERR_MSG(extack, "Invalid msg length");
1015		goto errout;
1016	}
1017
1018	ops = lookup_rules_ops(net, frh->family);
1019	if (ops == NULL) {
1020		err = -EAFNOSUPPORT;
1021		NL_SET_ERR_MSG(extack, "Rule family not supported");
1022		goto errout;
1023	}
1024
1025	err = nlmsg_parse_deprecated(nlh, sizeof(*frh), tb, FRA_MAX,
1026				     fib_rule_policy, extack);
1027	if (err < 0) {
1028		NL_SET_ERR_MSG(extack, "Error parsing msg");
1029		goto errout;
1030	}
1031
1032	err = fib_nl2rule(net, nlh, extack, ops, tb, &nlrule, &user_priority);
1033	if (err)
1034		goto errout;
1035
1036	if (!rtnl_held)
1037		rtnl_net_lock(net);
1038
1039	err = fib_nl2rule_rtnl(nlrule, ops, tb, extack);
1040	if (err)
1041		goto errout_free;
1042
1043	rule = rule_find(ops, frh, tb, nlrule, user_priority);
1044	if (!rule) {
1045		err = -ENOENT;
1046		goto errout_free;
1047	}
1048
1049	if (rule->flags & FIB_RULE_PERMANENT) {
1050		err = -EPERM;
1051		goto errout_free;
1052	}
1053
1054	if (ops->delete) {
1055		err = ops->delete(rule);
1056		if (err)
1057			goto errout_free;
1058	}
1059
1060	if (rule->tun_id)
1061		ip_tunnel_unneed_metadata();
1062
1063	list_del_rcu(&rule->list);
1064
1065	if (rule->action == FR_ACT_GOTO) {
1066		ops->nr_goto_rules--;
1067		if (rtnl_dereference(rule->ctarget) == NULL)
1068			ops->unresolved_rules--;
1069	}
1070
1071	/*
1072	 * Check if this rule is a target to any of them. If so,
1073	 * adjust to the next one with the same preference or
1074	 * disable them. As this operation is eventually very
1075	 * expensive, it is only performed if goto rules, except
1076	 * current if it is goto rule, have actually been added.
1077	 */
1078	if (ops->nr_goto_rules > 0) {
1079		struct fib_rule *n, *r;
1080
1081		n = list_next_entry(rule, list);
1082		if (&n->list == &ops->rules_list || n->pref != rule->pref)
1083			n = NULL;
1084		list_for_each_entry(r, &ops->rules_list, list) {
1085			if (rtnl_dereference(r->ctarget) != rule)
1086				continue;
1087			rcu_assign_pointer(r->ctarget, n);
1088			if (!n)
1089				ops->unresolved_rules++;
1090		}
1091	}
1092
1093	call_fib_rule_notifiers(net, FIB_EVENT_RULE_DEL, rule, ops, NULL);
1094
1095	if (!rtnl_held)
1096		rtnl_net_unlock(net);
1097
1098	notify_rule_change(RTM_DELRULE, rule, ops, nlh, NETLINK_CB(skb).portid);
1099	fib_rule_put(rule);
1100	flush_route_cache(ops);
1101	rules_ops_put(ops);
1102	kfree(nlrule);
1103	return 0;
1104
1105errout_free:
1106	if (!rtnl_held)
1107		rtnl_net_unlock(net);
1108	kfree(nlrule);
1109errout:
1110	rules_ops_put(ops);
1111	return err;
1112}
1113EXPORT_SYMBOL_GPL(fib_delrule);
1114
1115static int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr *nlh,
1116			  struct netlink_ext_ack *extack)
1117{
1118	return fib_delrule(sock_net(skb->sk), skb, nlh, extack, false);
1119}
1120
1121static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops,
1122					 struct fib_rule *rule)
1123{
1124	size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr))
1125			 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */
1126			 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */
1127			 + nla_total_size(4) /* FRA_PRIORITY */
1128			 + nla_total_size(4) /* FRA_TABLE */
1129			 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */
1130			 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */
1131			 + nla_total_size(4) /* FRA_FWMARK */
1132			 + nla_total_size(4) /* FRA_FWMASK */
1133			 + nla_total_size_64bit(8) /* FRA_TUN_ID */
1134			 + nla_total_size(sizeof(struct fib_kuid_range))
1135			 + nla_total_size(1) /* FRA_PROTOCOL */
1136			 + nla_total_size(1) /* FRA_IP_PROTO */
1137			 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_SPORT_RANGE */
1138			 + nla_total_size(sizeof(struct fib_rule_port_range)) /* FRA_DPORT_RANGE */
1139			 + nla_total_size(2) /* FRA_SPORT_MASK */
1140			 + nla_total_size(2); /* FRA_DPORT_MASK */
1141
1142	if (ops->nlmsg_payload)
1143		payload += ops->nlmsg_payload(rule);
1144
1145	return payload;
1146}
1147
1148static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule,
1149			    u32 pid, u32 seq, int type, int flags,
1150			    struct fib_rules_ops *ops)
1151{
1152	struct nlmsghdr *nlh;
1153	struct fib_rule_hdr *frh;
1154
1155	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags);
1156	if (nlh == NULL)
1157		return -EMSGSIZE;
1158
1159	frh = nlmsg_data(nlh);
1160	frh->family = ops->family;
1161	frh->table = rule->table < 256 ? rule->table : RT_TABLE_COMPAT;
1162	if (nla_put_u32(skb, FRA_TABLE, rule->table))
1163		goto nla_put_failure;
1164	if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen))
1165		goto nla_put_failure;
1166	frh->res1 = 0;
1167	frh->res2 = 0;
1168	frh->action = rule->action;
1169	frh->flags = rule->flags;
1170
1171	if (nla_put_u8(skb, FRA_PROTOCOL, rule->proto))
1172		goto nla_put_failure;
1173
1174	if (rule->action == FR_ACT_GOTO &&
1175	    rcu_access_pointer(rule->ctarget) == NULL)
1176		frh->flags |= FIB_RULE_UNRESOLVED;
1177
1178	if (rule->iifname[0]) {
1179		if (nla_put_string(skb, FRA_IIFNAME, rule->iifname))
1180			goto nla_put_failure;
1181		if (READ_ONCE(rule->iifindex) == -1)
1182			frh->flags |= FIB_RULE_IIF_DETACHED;
1183	}
1184
1185	if (rule->oifname[0]) {
1186		if (nla_put_string(skb, FRA_OIFNAME, rule->oifname))
1187			goto nla_put_failure;
1188		if (READ_ONCE(rule->oifindex) == -1)
1189			frh->flags |= FIB_RULE_OIF_DETACHED;
1190	}
1191
1192	if ((rule->pref &&
1193	     nla_put_u32(skb, FRA_PRIORITY, rule->pref)) ||
1194	    (rule->mark &&
1195	     nla_put_u32(skb, FRA_FWMARK, rule->mark)) ||
1196	    ((rule->mark_mask || rule->mark) &&
1197	     nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) ||
1198	    (rule->target &&
1199	     nla_put_u32(skb, FRA_GOTO, rule->target)) ||
1200	    (rule->tun_id &&
1201	     nla_put_be64(skb, FRA_TUN_ID, rule->tun_id, FRA_PAD)) ||
1202	    (rule->l3mdev &&
1203	     nla_put_u8(skb, FRA_L3MDEV, rule->l3mdev)) ||
1204	    (uid_range_set(&rule->uid_range) &&
1205	     nla_put_uid_range(skb, &rule->uid_range)) ||
1206	    (fib_rule_port_range_set(&rule->sport_range) &&
1207	     nla_put_port_range(skb, FRA_SPORT_RANGE, &rule->sport_range)) ||
1208	    (rule->sport_mask && nla_put_u16(skb, FRA_SPORT_MASK,
1209					     rule->sport_mask)) ||
1210	    (fib_rule_port_range_set(&rule->dport_range) &&
1211	     nla_put_port_range(skb, FRA_DPORT_RANGE, &rule->dport_range)) ||
1212	    (rule->dport_mask && nla_put_u16(skb, FRA_DPORT_MASK,
1213					     rule->dport_mask)) ||
1214	    (rule->ip_proto && nla_put_u8(skb, FRA_IP_PROTO, rule->ip_proto)))
1215		goto nla_put_failure;
1216
1217	if (rule->suppress_ifgroup != -1) {
1218		if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup))
1219			goto nla_put_failure;
1220	}
1221
1222	if (ops->fill(rule, skb, frh) < 0)
1223		goto nla_put_failure;
1224
1225	nlmsg_end(skb, nlh);
1226	return 0;
1227
1228nla_put_failure:
1229	nlmsg_cancel(skb, nlh);
1230	return -EMSGSIZE;
1231}
1232
1233static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
1234		      struct fib_rules_ops *ops)
1235{
1236	int idx = 0;
1237	struct fib_rule *rule;
1238	int err = 0;
1239
1240	rcu_read_lock();
1241	list_for_each_entry_rcu(rule, &ops->rules_list, list) {
1242		if (idx < cb->args[1])
1243			goto skip;
1244
1245		err = fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid,
1246				       cb->nlh->nlmsg_seq, RTM_NEWRULE,
1247				       NLM_F_MULTI, ops);
1248		if (err)
1249			break;
1250skip:
1251		idx++;
1252	}
1253	rcu_read_unlock();
1254	cb->args[1] = idx;
1255	rules_ops_put(ops);
1256
1257	return err;
1258}
1259
1260static int fib_valid_dumprule_req(const struct nlmsghdr *nlh,
1261				   struct netlink_ext_ack *extack)
1262{
1263	struct fib_rule_hdr *frh;
1264
1265	frh = nlmsg_payload(nlh, sizeof(*frh));
1266	if (!frh) {
1267		NL_SET_ERR_MSG(extack, "Invalid header for fib rule dump request");
1268		return -EINVAL;
1269	}
1270
1271	if (frh->dst_len || frh->src_len || frh->tos || frh->table ||
1272	    frh->res1 || frh->res2 || frh->action || frh->flags) {
1273		NL_SET_ERR_MSG(extack,
1274			       "Invalid values in header for fib rule dump request");
1275		return -EINVAL;
1276	}
1277
1278	if (nlmsg_attrlen(nlh, sizeof(*frh))) {
1279		NL_SET_ERR_MSG(extack, "Invalid data after header in fib rule dump request");
1280		return -EINVAL;
1281	}
1282
1283	return 0;
1284}
1285
1286static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb)
1287{
1288	const struct nlmsghdr *nlh = cb->nlh;
1289	struct net *net = sock_net(skb->sk);
1290	struct fib_rules_ops *ops;
1291	int err, idx = 0, family;
1292
1293	if (cb->strict_check) {
1294		err = fib_valid_dumprule_req(nlh, cb->extack);
1295
1296		if (err < 0)
1297			return err;
1298	}
1299
1300	family = rtnl_msg_family(nlh);
1301	if (family != AF_UNSPEC) {
1302		/* Protocol specific dump request */
1303		ops = lookup_rules_ops(net, family);
1304		if (ops == NULL)
1305			return -EAFNOSUPPORT;
1306
1307		return dump_rules(skb, cb, ops);
1308	}
1309
1310	err = 0;
1311	rcu_read_lock();
1312	list_for_each_entry_rcu(ops, &net->rules_ops, list) {
1313		if (idx < cb->args[0] || !try_module_get(ops->owner))
1314			goto skip;
1315
1316		err = dump_rules(skb, cb, ops);
1317		if (err < 0)
1318			break;
1319
1320		cb->args[1] = 0;
1321skip:
1322		idx++;
1323	}
1324	rcu_read_unlock();
1325	cb->args[0] = idx;
1326
1327	return err;
1328}
1329
1330static void notify_rule_change(int event, struct fib_rule *rule,
1331			       struct fib_rules_ops *ops, struct nlmsghdr *nlh,
1332			       u32 pid)
1333{
1334	struct net *net;
1335	struct sk_buff *skb;
1336	int err = -ENOMEM;
1337
1338	net = ops->fro_net;
1339	skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL);
1340	if (skb == NULL)
1341		goto errout;
1342
1343	err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops);
1344	if (err < 0) {
1345		/* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */
1346		WARN_ON(err == -EMSGSIZE);
1347		kfree_skb(skb);
1348		goto errout;
1349	}
1350
1351	rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL);
1352	return;
1353errout:
1354	rtnl_set_sk_err(net, ops->nlgroup, err);
1355}
1356
1357static void attach_rules(struct list_head *rules, struct net_device *dev)
1358{
1359	struct fib_rule *rule;
1360
1361	list_for_each_entry(rule, rules, list) {
1362		if (rule->iifindex == -1 &&
1363		    strcmp(dev->name, rule->iifname) == 0) {
1364			WRITE_ONCE(rule->iifindex, dev->ifindex);
1365			WRITE_ONCE(rule->iif_is_l3_master,
1366				   netif_is_l3_master(dev));
1367		}
1368		if (rule->oifindex == -1 &&
1369		    strcmp(dev->name, rule->oifname) == 0) {
1370			WRITE_ONCE(rule->oifindex, dev->ifindex);
1371			WRITE_ONCE(rule->oif_is_l3_master,
1372				   netif_is_l3_master(dev));
1373		}
1374	}
1375}
1376
1377static void detach_rules(struct list_head *rules, struct net_device *dev)
1378{
1379	struct fib_rule *rule;
1380
1381	list_for_each_entry(rule, rules, list) {
1382		if (rule->iifindex == dev->ifindex) {
1383			WRITE_ONCE(rule->iifindex, -1);
1384			WRITE_ONCE(rule->iif_is_l3_master, false);
1385		}
1386		if (rule->oifindex == dev->ifindex) {
1387			WRITE_ONCE(rule->oifindex, -1);
1388			WRITE_ONCE(rule->oif_is_l3_master, false);
1389		}
1390	}
1391}
1392
1393
1394static int fib_rules_event(struct notifier_block *this, unsigned long event,
1395			   void *ptr)
1396{
1397	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1398	struct net *net = dev_net(dev);
1399	struct fib_rules_ops *ops;
1400
1401	ASSERT_RTNL();
1402
1403	switch (event) {
1404	case NETDEV_REGISTER:
1405		list_for_each_entry(ops, &net->rules_ops, list)
1406			attach_rules(&ops->rules_list, dev);
1407		break;
1408
1409	case NETDEV_CHANGENAME:
1410		list_for_each_entry(ops, &net->rules_ops, list) {
1411			detach_rules(&ops->rules_list, dev);
1412			attach_rules(&ops->rules_list, dev);
1413		}
1414		break;
1415
1416	case NETDEV_UNREGISTER:
1417		list_for_each_entry(ops, &net->rules_ops, list)
1418			detach_rules(&ops->rules_list, dev);
1419		break;
1420	}
1421
1422	return NOTIFY_DONE;
1423}
1424
1425static struct notifier_block fib_rules_notifier = {
1426	.notifier_call = fib_rules_event,
1427};
1428
1429static int __net_init fib_rules_net_init(struct net *net)
1430{
1431	INIT_LIST_HEAD(&net->rules_ops);
1432	spin_lock_init(&net->rules_mod_lock);
1433	return 0;
1434}
1435
1436static void __net_exit fib_rules_net_exit(struct net *net)
1437{
1438	WARN_ON_ONCE(!list_empty(&net->rules_ops));
1439}
1440
1441static struct pernet_operations fib_rules_net_ops = {
1442	.init = fib_rules_net_init,
1443	.exit = fib_rules_net_exit,
1444};
1445
1446static const struct rtnl_msg_handler fib_rules_rtnl_msg_handlers[] __initconst = {
1447	{.msgtype = RTM_NEWRULE, .doit = fib_nl_newrule,
1448	 .flags = RTNL_FLAG_DOIT_PERNET},
1449	{.msgtype = RTM_DELRULE, .doit = fib_nl_delrule,
1450	 .flags = RTNL_FLAG_DOIT_PERNET},
1451	{.msgtype = RTM_GETRULE, .dumpit = fib_nl_dumprule,
1452	 .flags = RTNL_FLAG_DUMP_UNLOCKED},
1453};
1454
1455static int __init fib_rules_init(void)
1456{
1457	int err;
1458
1459	rtnl_register_many(fib_rules_rtnl_msg_handlers);
1460
1461	err = register_pernet_subsys(&fib_rules_net_ops);
1462	if (err < 0)
1463		goto fail;
1464
1465	err = register_netdevice_notifier(&fib_rules_notifier);
1466	if (err < 0)
1467		goto fail_unregister;
1468
1469	return 0;
1470
1471fail_unregister:
1472	unregister_pernet_subsys(&fib_rules_net_ops);
1473fail:
1474	rtnl_unregister_many(fib_rules_rtnl_msg_handlers);
1475	return err;
1476}
1477
1478subsys_initcall(fib_rules_init);