fib_rules.c 18.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12
/*
 * net/core/fib_rules.c		Generic Routing Rules
 *
 *	This program is free software; you can redistribute it and/or
 *	modify it under the terms of the GNU General Public License as
 *	published by the Free Software Foundation, version 2.
 *
 * Authors:	Thomas Graf <tgraf@suug.ch>
 */

#include <linux/types.h>
#include <linux/kernel.h>
13
#include <linux/slab.h>
14
#include <linux/list.h>
15
#include <linux/module.h>
16
#include <net/net_namespace.h>
17
#include <net/sock.h>
18
#include <net/fib_rules.h>
19
#include <net/ip_tunnels.h>
20

21 22 23 24 25 26 27 28 29 30 31 32 33 34
int fib_default_rule_add(struct fib_rules_ops *ops,
			 u32 pref, u32 table, u32 flags)
{
	struct fib_rule *r;

	r = kzalloc(ops->rule_size, GFP_KERNEL);
	if (r == NULL)
		return -ENOMEM;

	atomic_set(&r->refcnt, 1);
	r->action = FR_ACT_TO_TBL;
	r->pref = pref;
	r->table = table;
	r->flags = flags;
35
	r->fr_net = ops->fro_net;
36

37 38 39
	r->suppress_prefixlen = -1;
	r->suppress_ifgroup = -1;

40 41 42 43 44 45 46
	/* The lock is not required here, the list in unreacheable
	 * at the moment this function is called */
	list_add_tail(&r->list, &ops->rules_list);
	return 0;
}
EXPORT_SYMBOL(fib_default_rule_add);

47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64
u32 fib_default_rule_pref(struct fib_rules_ops *ops)
{
	struct list_head *pos;
	struct fib_rule *rule;

	if (!list_empty(&ops->rules_list)) {
		pos = ops->rules_list.next;
		if (pos->next != &ops->rules_list) {
			rule = list_entry(pos->next, struct fib_rule, list);
			if (rule->pref)
				return rule->pref - 1;
		}
	}

	return 0;
}
EXPORT_SYMBOL(fib_default_rule_pref);

D
Denis V. Lunev 已提交
65
static void notify_rule_change(int event, struct fib_rule *rule,
66 67
			       struct fib_rules_ops *ops, struct nlmsghdr *nlh,
			       u32 pid);
68

69
static struct fib_rules_ops *lookup_rules_ops(struct net *net, int family)
70 71 72 73
{
	struct fib_rules_ops *ops;

	rcu_read_lock();
74
	list_for_each_entry_rcu(ops, &net->rules_ops, list) {
75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92
		if (ops->family == family) {
			if (!try_module_get(ops->owner))
				ops = NULL;
			rcu_read_unlock();
			return ops;
		}
	}
	rcu_read_unlock();

	return NULL;
}

static void rules_ops_put(struct fib_rules_ops *ops)
{
	if (ops)
		module_put(ops->owner);
}

93 94 95
static void flush_route_cache(struct fib_rules_ops *ops)
{
	if (ops->flush_cache)
96
		ops->flush_cache(ops);
97 98
}

99
static int __fib_rules_register(struct fib_rules_ops *ops)
100 101 102
{
	int err = -EEXIST;
	struct fib_rules_ops *o;
D
Denis V. Lunev 已提交
103 104 105
	struct net *net;

	net = ops->fro_net;
106 107 108 109 110 111 112 113 114

	if (ops->rule_size < sizeof(struct fib_rule))
		return -EINVAL;

	if (ops->match == NULL || ops->configure == NULL ||
	    ops->compare == NULL || ops->fill == NULL ||
	    ops->action == NULL)
		return -EINVAL;

115 116
	spin_lock(&net->rules_mod_lock);
	list_for_each_entry(o, &net->rules_ops, list)
117 118 119
		if (ops->family == o->family)
			goto errout;

120
	list_add_tail_rcu(&ops->list, &net->rules_ops);
121 122
	err = 0;
errout:
123
	spin_unlock(&net->rules_mod_lock);
124 125 126 127

	return err;
}

128
struct fib_rules_ops *
129
fib_rules_register(const struct fib_rules_ops *tmpl, struct net *net)
130 131 132 133
{
	struct fib_rules_ops *ops;
	int err;

134
	ops = kmemdup(tmpl, sizeof(*ops), GFP_KERNEL);
135 136 137 138 139 140 141 142 143 144 145 146 147 148
	if (ops == NULL)
		return ERR_PTR(-ENOMEM);

	INIT_LIST_HEAD(&ops->rules_list);
	ops->fro_net = net;

	err = __fib_rules_register(ops);
	if (err) {
		kfree(ops);
		ops = ERR_PTR(err);
	}

	return ops;
}
149 150
EXPORT_SYMBOL_GPL(fib_rules_register);

151
static void fib_rules_cleanup_ops(struct fib_rules_ops *ops)
152 153 154
{
	struct fib_rule *rule, *tmp;

155
	list_for_each_entry_safe(rule, tmp, &ops->rules_list, list) {
156
		list_del_rcu(&rule->list);
157 158
		if (ops->delete)
			ops->delete(rule);
159 160 161 162
		fib_rule_put(rule);
	}
}

D
Denis V. Lunev 已提交
163
void fib_rules_unregister(struct fib_rules_ops *ops)
164
{
D
Denis V. Lunev 已提交
165
	struct net *net = ops->fro_net;
166

167
	spin_lock(&net->rules_mod_lock);
168
	list_del_rcu(&ops->list);
169
	spin_unlock(&net->rules_mod_lock);
170

171
	fib_rules_cleanup_ops(ops);
172
	kfree_rcu(ops, rcu);
173 174 175
}
EXPORT_SYMBOL_GPL(fib_rules_unregister);

176 177 178 179 180
static int fib_rule_match(struct fib_rule *rule, struct fib_rules_ops *ops,
			  struct flowi *fl, int flags)
{
	int ret = 0;

181
	if (rule->iifindex && (rule->iifindex != fl->flowi_iif))
182 183
		goto out;

184
	if (rule->oifindex && (rule->oifindex != fl->flowi_oif))
185 186
		goto out;

187
	if ((rule->mark ^ fl->flowi_mark) & rule->mark_mask)
188 189
		goto out;

190 191 192
	if (rule->tun_id && (rule->tun_id != fl->flowi_tun_key.tun_id))
		goto out;

193 194 195 196 197
	ret = ops->match(rule, fl, flags);
out:
	return (rule->flags & FIB_RULE_INVERT) ? !ret : ret;
}

198 199 200 201 202 203 204 205
int fib_rules_lookup(struct fib_rules_ops *ops, struct flowi *fl,
		     int flags, struct fib_lookup_arg *arg)
{
	struct fib_rule *rule;
	int err;

	rcu_read_lock();

206
	list_for_each_entry_rcu(rule, &ops->rules_list, list) {
T
Thomas Graf 已提交
207
jumped:
208
		if (!fib_rule_match(rule, ops, fl, flags))
209 210
			continue;

T
Thomas Graf 已提交
211 212 213 214 215 216 217 218 219 220
		if (rule->action == FR_ACT_GOTO) {
			struct fib_rule *target;

			target = rcu_dereference(rule->ctarget);
			if (target == NULL) {
				continue;
			} else {
				rule = target;
				goto jumped;
			}
221 222 223
		} else if (rule->action == FR_ACT_NOP)
			continue;
		else
T
Thomas Graf 已提交
224 225
			err = ops->action(rule, fl, flags, arg);

226 227 228
		if (!err && ops->suppress && ops->suppress(rule, arg))
			continue;

229
		if (err != -EAGAIN) {
E
Eric Dumazet 已提交
230 231
			if ((arg->flags & FIB_LOOKUP_NOREF) ||
			    likely(atomic_inc_not_zero(&rule->refcnt))) {
232 233 234 235
				arg->rule = rule;
				goto out;
			}
			break;
236 237 238
		}
	}

239
	err = -ESRCH;
240 241 242 243 244 245 246
out:
	rcu_read_unlock();

	return err;
}
EXPORT_SYMBOL_GPL(fib_rules_lookup);

247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268
static int validate_rulemsg(struct fib_rule_hdr *frh, struct nlattr **tb,
			    struct fib_rules_ops *ops)
{
	int err = -EINVAL;

	if (frh->src_len)
		if (tb[FRA_SRC] == NULL ||
		    frh->src_len > (ops->addr_size * 8) ||
		    nla_len(tb[FRA_SRC]) != ops->addr_size)
			goto errout;

	if (frh->dst_len)
		if (tb[FRA_DST] == NULL ||
		    frh->dst_len > (ops->addr_size * 8) ||
		    nla_len(tb[FRA_DST]) != ops->addr_size)
			goto errout;

	err = 0;
errout:
	return err;
}

269
static int fib_nl_newrule(struct sk_buff *skb, struct nlmsghdr* nlh)
270
{
271
	struct net *net = sock_net(skb->sk);
272 273 274 275
	struct fib_rule_hdr *frh = nlmsg_data(nlh);
	struct fib_rules_ops *ops = NULL;
	struct fib_rule *rule, *r, *last = NULL;
	struct nlattr *tb[FRA_MAX+1];
T
Thomas Graf 已提交
276
	int err = -EINVAL, unresolved = 0;
277 278 279 280

	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh)))
		goto errout;

281
	ops = lookup_rules_ops(net, frh->family);
282
	if (ops == NULL) {
283
		err = -EAFNOSUPPORT;
284 285 286 287 288 289 290
		goto errout;
	}

	err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy);
	if (err < 0)
		goto errout;

291 292 293 294
	err = validate_rulemsg(frh, tb, ops);
	if (err < 0)
		goto errout;

295 296 297 298 299
	rule = kzalloc(ops->rule_size, GFP_KERNEL);
	if (rule == NULL) {
		err = -ENOMEM;
		goto errout;
	}
300
	rule->fr_net = net;
301 302 303 304

	if (tb[FRA_PRIORITY])
		rule->pref = nla_get_u32(tb[FRA_PRIORITY]);

305
	if (tb[FRA_IIFNAME]) {
306 307
		struct net_device *dev;

308 309 310
		rule->iifindex = -1;
		nla_strlcpy(rule->iifname, tb[FRA_IIFNAME], IFNAMSIZ);
		dev = __dev_get_by_name(net, rule->iifname);
311
		if (dev)
312
			rule->iifindex = dev->ifindex;
313 314
	}

315 316 317 318 319 320 321 322 323 324
	if (tb[FRA_OIFNAME]) {
		struct net_device *dev;

		rule->oifindex = -1;
		nla_strlcpy(rule->oifname, tb[FRA_OIFNAME], IFNAMSIZ);
		dev = __dev_get_by_name(net, rule->oifname);
		if (dev)
			rule->oifindex = dev->ifindex;
	}

325 326 327 328 329 330 331 332 333 334 335 336
	if (tb[FRA_FWMARK]) {
		rule->mark = nla_get_u32(tb[FRA_FWMARK]);
		if (rule->mark)
			/* compatibility: if the mark value is non-zero all bits
			 * are compared unless a mask is explicitly specified.
			 */
			rule->mark_mask = 0xFFFFFFFF;
	}

	if (tb[FRA_FWMASK])
		rule->mark_mask = nla_get_u32(tb[FRA_FWMASK]);

337 338 339
	if (tb[FRA_TUN_ID])
		rule->tun_id = nla_get_be64(tb[FRA_TUN_ID]);

340 341
	rule->action = frh->action;
	rule->flags = frh->flags;
342
	rule->table = frh_get_table(frh, tb);
343 344 345 346
	if (tb[FRA_SUPPRESS_PREFIXLEN])
		rule->suppress_prefixlen = nla_get_u32(tb[FRA_SUPPRESS_PREFIXLEN]);
	else
		rule->suppress_prefixlen = -1;
347

348 349
	if (tb[FRA_SUPPRESS_IFGROUP])
		rule->suppress_ifgroup = nla_get_u32(tb[FRA_SUPPRESS_IFGROUP]);
350 351
	else
		rule->suppress_ifgroup = -1;
352

353
	if (!tb[FRA_PRIORITY] && ops->default_pref)
354
		rule->pref = ops->default_pref(ops);
355

T
Thomas Graf 已提交
356 357 358 359 360 361 362 363 364 365
	err = -EINVAL;
	if (tb[FRA_GOTO]) {
		if (rule->action != FR_ACT_GOTO)
			goto errout_free;

		rule->target = nla_get_u32(tb[FRA_GOTO]);
		/* Backward jumps are prohibited to avoid endless loops */
		if (rule->target <= rule->pref)
			goto errout_free;

366
		list_for_each_entry(r, &ops->rules_list, list) {
T
Thomas Graf 已提交
367
			if (r->pref == rule->target) {
E
Eric Dumazet 已提交
368
				RCU_INIT_POINTER(rule->ctarget, r);
T
Thomas Graf 已提交
369 370 371 372
				break;
			}
		}

E
Eric Dumazet 已提交
373
		if (rcu_dereference_protected(rule->ctarget, 1) == NULL)
T
Thomas Graf 已提交
374 375 376 377
			unresolved = 1;
	} else if (rule->action == FR_ACT_GOTO)
		goto errout_free;

378
	err = ops->configure(rule, skb, frh, tb);
379 380 381
	if (err < 0)
		goto errout_free;

382
	list_for_each_entry(r, &ops->rules_list, list) {
383 384 385 386 387 388 389
		if (r->pref > rule->pref)
			break;
		last = r;
	}

	fib_rule_get(rule);

E
Eric Dumazet 已提交
390 391 392 393 394
	if (last)
		list_add_rcu(&rule->list, &last->list);
	else
		list_add_rcu(&rule->list, &ops->rules_list);

T
Thomas Graf 已提交
395 396 397 398 399
	if (ops->unresolved_rules) {
		/*
		 * There are unresolved goto rules in the list, check if
		 * any of them are pointing to this new rule.
		 */
400
		list_for_each_entry(r, &ops->rules_list, list) {
T
Thomas Graf 已提交
401
			if (r->action == FR_ACT_GOTO &&
402 403
			    r->target == rule->pref &&
			    rtnl_dereference(r->ctarget) == NULL) {
T
Thomas Graf 已提交
404 405 406 407 408 409 410 411 412 413 414 415 416
				rcu_assign_pointer(r->ctarget, rule);
				if (--ops->unresolved_rules == 0)
					break;
			}
		}
	}

	if (rule->action == FR_ACT_GOTO)
		ops->nr_goto_rules++;

	if (unresolved)
		ops->unresolved_rules++;

417 418 419
	if (rule->tun_id)
		ip_tunnel_need_metadata();

420
	notify_rule_change(RTM_NEWRULE, rule, ops, nlh, NETLINK_CB(skb).portid);
421
	flush_route_cache(ops);
422 423 424 425 426 427 428 429 430 431
	rules_ops_put(ops);
	return 0;

errout_free:
	kfree(rule);
errout:
	rules_ops_put(ops);
	return err;
}

432
static int fib_nl_delrule(struct sk_buff *skb, struct nlmsghdr* nlh)
433
{
434
	struct net *net = sock_net(skb->sk);
435 436
	struct fib_rule_hdr *frh = nlmsg_data(nlh);
	struct fib_rules_ops *ops = NULL;
T
Thomas Graf 已提交
437
	struct fib_rule *rule, *tmp;
438 439 440 441 442 443
	struct nlattr *tb[FRA_MAX+1];
	int err = -EINVAL;

	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*frh)))
		goto errout;

444
	ops = lookup_rules_ops(net, frh->family);
445
	if (ops == NULL) {
446
		err = -EAFNOSUPPORT;
447 448 449 450 451 452 453
		goto errout;
	}

	err = nlmsg_parse(nlh, sizeof(*frh), tb, FRA_MAX, ops->policy);
	if (err < 0)
		goto errout;

454 455 456 457
	err = validate_rulemsg(frh, tb, ops);
	if (err < 0)
		goto errout;

458
	list_for_each_entry(rule, &ops->rules_list, list) {
459 460 461
		if (frh->action && (frh->action != rule->action))
			continue;

462 463
		if (frh_get_table(frh, tb) &&
		    (frh_get_table(frh, tb) != rule->table))
464 465 466 467 468 469
			continue;

		if (tb[FRA_PRIORITY] &&
		    (rule->pref != nla_get_u32(tb[FRA_PRIORITY])))
			continue;

470 471
		if (tb[FRA_IIFNAME] &&
		    nla_strcmp(tb[FRA_IIFNAME], rule->iifname))
472 473
			continue;

474 475 476 477
		if (tb[FRA_OIFNAME] &&
		    nla_strcmp(tb[FRA_OIFNAME], rule->oifname))
			continue;

478 479 480 481 482 483 484 485
		if (tb[FRA_FWMARK] &&
		    (rule->mark != nla_get_u32(tb[FRA_FWMARK])))
			continue;

		if (tb[FRA_FWMASK] &&
		    (rule->mark_mask != nla_get_u32(tb[FRA_FWMASK])))
			continue;

486 487 488 489
		if (tb[FRA_TUN_ID] &&
		    (rule->tun_id != nla_get_be64(tb[FRA_TUN_ID])))
			continue;

490 491 492 493 494 495 496 497
		if (!ops->compare(rule, frh, tb))
			continue;

		if (rule->flags & FIB_RULE_PERMANENT) {
			err = -EPERM;
			goto errout;
		}

498 499 500 501 502 503
		if (ops->delete) {
			err = ops->delete(rule);
			if (err)
				goto errout;
		}

504 505 506
		if (rule->tun_id)
			ip_tunnel_unneed_metadata();

507
		list_del_rcu(&rule->list);
T
Thomas Graf 已提交
508

509
		if (rule->action == FR_ACT_GOTO) {
T
Thomas Graf 已提交
510
			ops->nr_goto_rules--;
511 512 513
			if (rtnl_dereference(rule->ctarget) == NULL)
				ops->unresolved_rules--;
		}
T
Thomas Graf 已提交
514 515 516 517 518 519 520 521

		/*
		 * Check if this rule is a target to any of them. If so,
		 * disable them. As this operation is eventually very
		 * expensive, it is only performed if goto rules have
		 * actually been added.
		 */
		if (ops->nr_goto_rules > 0) {
522
			list_for_each_entry(tmp, &ops->rules_list, list) {
E
Eric Dumazet 已提交
523
				if (rtnl_dereference(tmp->ctarget) == rule) {
524
					RCU_INIT_POINTER(tmp->ctarget, NULL);
T
Thomas Graf 已提交
525 526 527 528 529
					ops->unresolved_rules++;
				}
			}
		}

D
Denis V. Lunev 已提交
530
		notify_rule_change(RTM_DELRULE, rule, ops, nlh,
531
				   NETLINK_CB(skb).portid);
532
		fib_rule_put(rule);
533
		flush_route_cache(ops);
534 535 536 537 538 539 540 541 542 543
		rules_ops_put(ops);
		return 0;
	}

	err = -ENOENT;
errout:
	rules_ops_put(ops);
	return err;
}

544 545 546 547
static inline size_t fib_rule_nlmsg_size(struct fib_rules_ops *ops,
					 struct fib_rule *rule)
{
	size_t payload = NLMSG_ALIGN(sizeof(struct fib_rule_hdr))
548
			 + nla_total_size(IFNAMSIZ) /* FRA_IIFNAME */
549
			 + nla_total_size(IFNAMSIZ) /* FRA_OIFNAME */
550 551
			 + nla_total_size(4) /* FRA_PRIORITY */
			 + nla_total_size(4) /* FRA_TABLE */
552
			 + nla_total_size(4) /* FRA_SUPPRESS_PREFIXLEN */
553
			 + nla_total_size(4) /* FRA_SUPPRESS_IFGROUP */
554
			 + nla_total_size(4) /* FRA_FWMARK */
555 556
			 + nla_total_size(4) /* FRA_FWMASK */
			 + nla_total_size(8); /* FRA_TUN_ID */
557 558 559 560 561 562 563

	if (ops->nlmsg_payload)
		payload += ops->nlmsg_payload(rule);

	return payload;
}

564 565 566 567 568 569 570 571 572
static int fib_nl_fill_rule(struct sk_buff *skb, struct fib_rule *rule,
			    u32 pid, u32 seq, int type, int flags,
			    struct fib_rules_ops *ops)
{
	struct nlmsghdr *nlh;
	struct fib_rule_hdr *frh;

	nlh = nlmsg_put(skb, pid, seq, type, sizeof(*frh), flags);
	if (nlh == NULL)
573
		return -EMSGSIZE;
574 575

	frh = nlmsg_data(nlh);
576
	frh->family = ops->family;
577
	frh->table = rule->table;
578 579
	if (nla_put_u32(skb, FRA_TABLE, rule->table))
		goto nla_put_failure;
580
	if (nla_put_u32(skb, FRA_SUPPRESS_PREFIXLEN, rule->suppress_prefixlen))
581
		goto nla_put_failure;
582 583 584 585 586
	frh->res1 = 0;
	frh->res2 = 0;
	frh->action = rule->action;
	frh->flags = rule->flags;

E
Eric Dumazet 已提交
587
	if (rule->action == FR_ACT_GOTO &&
588
	    rcu_access_pointer(rule->ctarget) == NULL)
T
Thomas Graf 已提交
589 590
		frh->flags |= FIB_RULE_UNRESOLVED;

591
	if (rule->iifname[0]) {
592 593
		if (nla_put_string(skb, FRA_IIFNAME, rule->iifname))
			goto nla_put_failure;
594 595
		if (rule->iifindex == -1)
			frh->flags |= FIB_RULE_IIF_DETACHED;
596 597
	}

598
	if (rule->oifname[0]) {
599 600
		if (nla_put_string(skb, FRA_OIFNAME, rule->oifname))
			goto nla_put_failure;
601 602 603 604
		if (rule->oifindex == -1)
			frh->flags |= FIB_RULE_OIF_DETACHED;
	}

605 606 607 608 609 610 611
	if ((rule->pref &&
	     nla_put_u32(skb, FRA_PRIORITY, rule->pref)) ||
	    (rule->mark &&
	     nla_put_u32(skb, FRA_FWMARK, rule->mark)) ||
	    ((rule->mark_mask || rule->mark) &&
	     nla_put_u32(skb, FRA_FWMASK, rule->mark_mask)) ||
	    (rule->target &&
612 613 614
	     nla_put_u32(skb, FRA_GOTO, rule->target)) ||
	    (rule->tun_id &&
	     nla_put_be64(skb, FRA_TUN_ID, rule->tun_id)))
615
		goto nla_put_failure;
616 617 618 619 620 621

	if (rule->suppress_ifgroup != -1) {
		if (nla_put_u32(skb, FRA_SUPPRESS_IFGROUP, rule->suppress_ifgroup))
			goto nla_put_failure;
	}

622
	if (ops->fill(rule, skb, frh) < 0)
623 624
		goto nla_put_failure;

625 626
	nlmsg_end(skb, nlh);
	return 0;
627 628

nla_put_failure:
629 630
	nlmsg_cancel(skb, nlh);
	return -EMSGSIZE;
631 632
}

T
Thomas Graf 已提交
633 634
static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
		      struct fib_rules_ops *ops)
635 636 637 638
{
	int idx = 0;
	struct fib_rule *rule;

639 640
	rcu_read_lock();
	list_for_each_entry_rcu(rule, &ops->rules_list, list) {
T
Thomas Graf 已提交
641
		if (idx < cb->args[1])
642 643
			goto skip;

644
		if (fib_nl_fill_rule(skb, rule, NETLINK_CB(cb->skb).portid,
645 646 647 648 649 650
				     cb->nlh->nlmsg_seq, RTM_NEWRULE,
				     NLM_F_MULTI, ops) < 0)
			break;
skip:
		idx++;
	}
651
	rcu_read_unlock();
T
Thomas Graf 已提交
652
	cb->args[1] = idx;
653 654 655 656 657
	rules_ops_put(ops);

	return skb->len;
}

T
Thomas Graf 已提交
658 659
static int fib_nl_dumprule(struct sk_buff *skb, struct netlink_callback *cb)
{
660
	struct net *net = sock_net(skb->sk);
T
Thomas Graf 已提交
661 662 663 664 665 666
	struct fib_rules_ops *ops;
	int idx = 0, family;

	family = rtnl_msg_family(cb->nlh);
	if (family != AF_UNSPEC) {
		/* Protocol specific dump request */
667
		ops = lookup_rules_ops(net, family);
T
Thomas Graf 已提交
668 669 670 671 672 673 674
		if (ops == NULL)
			return -EAFNOSUPPORT;

		return dump_rules(skb, cb, ops);
	}

	rcu_read_lock();
675
	list_for_each_entry_rcu(ops, &net->rules_ops, list) {
T
Thomas Graf 已提交
676 677 678 679 680 681 682
		if (idx < cb->args[0] || !try_module_get(ops->owner))
			goto skip;

		if (dump_rules(skb, cb, ops) < 0)
			break;

		cb->args[1] = 0;
683
skip:
T
Thomas Graf 已提交
684 685 686 687 688 689 690
		idx++;
	}
	rcu_read_unlock();
	cb->args[0] = idx;

	return skb->len;
}
691

D
Denis V. Lunev 已提交
692
static void notify_rule_change(int event, struct fib_rule *rule,
693 694
			       struct fib_rules_ops *ops, struct nlmsghdr *nlh,
			       u32 pid)
695
{
D
Denis V. Lunev 已提交
696
	struct net *net;
697 698
	struct sk_buff *skb;
	int err = -ENOBUFS;
699

D
Denis V. Lunev 已提交
700
	net = ops->fro_net;
701
	skb = nlmsg_new(fib_rule_nlmsg_size(ops, rule), GFP_KERNEL);
702
	if (skb == NULL)
703 704 705
		goto errout;

	err = fib_nl_fill_rule(skb, rule, pid, nlh->nlmsg_seq, event, 0, ops);
706 707 708 709 710 711
	if (err < 0) {
		/* -EMSGSIZE implies BUG in fib_rule_nlmsg_size() */
		WARN_ON(err == -EMSGSIZE);
		kfree_skb(skb);
		goto errout;
	}
D
Denis V. Lunev 已提交
712

713 714
	rtnl_notify(skb, net, pid, ops->nlgroup, nlh, GFP_KERNEL);
	return;
715 716
errout:
	if (err < 0)
717
		rtnl_set_sk_err(net, ops->nlgroup, err);
718 719 720 721 722 723 724
}

static void attach_rules(struct list_head *rules, struct net_device *dev)
{
	struct fib_rule *rule;

	list_for_each_entry(rule, rules, list) {
725 726 727
		if (rule->iifindex == -1 &&
		    strcmp(dev->name, rule->iifname) == 0)
			rule->iifindex = dev->ifindex;
728 729 730
		if (rule->oifindex == -1 &&
		    strcmp(dev->name, rule->oifname) == 0)
			rule->oifindex = dev->ifindex;
731 732 733 734 735 736 737
	}
}

static void detach_rules(struct list_head *rules, struct net_device *dev)
{
	struct fib_rule *rule;

738
	list_for_each_entry(rule, rules, list) {
739 740
		if (rule->iifindex == dev->ifindex)
			rule->iifindex = -1;
741 742 743
		if (rule->oifindex == dev->ifindex)
			rule->oifindex = -1;
	}
744 745 746 747
}


static int fib_rules_event(struct notifier_block *this, unsigned long event,
748
			   void *ptr)
749
{
750
	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
751
	struct net *net = dev_net(dev);
752 753
	struct fib_rules_ops *ops;

754
	ASSERT_RTNL();
755 756 757

	switch (event) {
	case NETDEV_REGISTER:
758
		list_for_each_entry(ops, &net->rules_ops, list)
759
			attach_rules(&ops->rules_list, dev);
760 761
		break;

762 763 764 765 766 767 768
	case NETDEV_CHANGENAME:
		list_for_each_entry(ops, &net->rules_ops, list) {
			detach_rules(&ops->rules_list, dev);
			attach_rules(&ops->rules_list, dev);
		}
		break;

769
	case NETDEV_UNREGISTER:
770
		list_for_each_entry(ops, &net->rules_ops, list)
771
			detach_rules(&ops->rules_list, dev);
772 773 774 775 776 777 778 779 780 781
		break;
	}

	return NOTIFY_DONE;
}

static struct notifier_block fib_rules_notifier = {
	.notifier_call = fib_rules_event,
};

782
static int __net_init fib_rules_net_init(struct net *net)
783 784 785 786 787 788 789 790 791 792
{
	INIT_LIST_HEAD(&net->rules_ops);
	spin_lock_init(&net->rules_mod_lock);
	return 0;
}

static struct pernet_operations fib_rules_net_ops = {
	.init = fib_rules_net_init,
};

793 794
static int __init fib_rules_init(void)
{
795
	int err;
796 797 798
	rtnl_register(PF_UNSPEC, RTM_NEWRULE, fib_nl_newrule, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_DELRULE, fib_nl_delrule, NULL, NULL);
	rtnl_register(PF_UNSPEC, RTM_GETRULE, NULL, fib_nl_dumprule, NULL);
799

E
Eric W. Biederman 已提交
800
	err = register_pernet_subsys(&fib_rules_net_ops);
801 802 803
	if (err < 0)
		goto fail;

E
Eric W. Biederman 已提交
804
	err = register_netdevice_notifier(&fib_rules_notifier);
805 806
	if (err < 0)
		goto fail_unregister;
E
Eric W. Biederman 已提交
807

808 809 810
	return 0;

fail_unregister:
E
Eric W. Biederman 已提交
811
	unregister_pernet_subsys(&fib_rules_net_ops);
812 813 814 815 816
fail:
	rtnl_unregister(PF_UNSPEC, RTM_NEWRULE);
	rtnl_unregister(PF_UNSPEC, RTM_DELRULE);
	rtnl_unregister(PF_UNSPEC, RTM_GETRULE);
	return err;
817 818 819
}

subsys_initcall(fib_rules_init);