forked from ~ljy/RK356X_SDK_RELEASE

blame | history | patch | commit | commitdiff
hc
2024-01-05 071106ecf68c401173c58808b1cf5f68cc50d390 
 kernel/net/ipv4/fib_frontend.c
....@@ -1,3 +1,4 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23  * INET		An implementation of the TCP/IP protocol suite for the LINUX
34  *		operating system.  INET is implemented using the  BSD Socket
....@@ -6,11 +7,6 @@
67  *		IPv4 Forwarding Information Base: FIB frontend.
78  *
89  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9
- *
10
- *		This program is free software; you can redistribute it and/or
11
- *		modify it under the terms of the GNU General Public License
12
- *		as published by the Free Software Foundation; either version
13
- *		2 of the License, or (at your option) any later version.
1410  */
1511 
1612 #include <linux/module.h>
....@@ -43,6 +39,7 @@
4339 #include <net/sock.h>
4440 #include <net/arp.h>
4541 #include <net/ip_fib.h>
42
+#include <net/nexthop.h>
4643 #include <net/rtnetlink.h>
4744 #include <net/xfrm.h>
4845 #include <net/l3mdev.h>
....@@ -72,11 +69,6 @@
7269 fail:
7370 	fib_free_table(main_table);
7471 	return -ENOMEM;
75
-}
76
-
77
-static bool fib4_has_custom_rules(struct net *net)
78
-{
79
-	return false;
8072 }
8173 #else
8274 
....@@ -127,16 +119,12 @@
127119 	h = id & (FIB_TABLE_HASHSZ - 1);
128120 
129121 	head = &net->ipv4.fib_table_hash[h];
130
-	hlist_for_each_entry_rcu(tb, head, tb_hlist) {
122
+	hlist_for_each_entry_rcu(tb, head, tb_hlist,
123
+				 lockdep_rtnl_is_held()) {
131124 		if (tb->tb_id == id)
132125 			return tb;
133126 	}
134127 	return NULL;
135
-}
136
-
137
-static bool fib4_has_custom_rules(struct net *net)
138
-{
139
-	return net->ipv4.fib_has_custom_rules;
140128 }
141129 #endif /* CONFIG_IP_MULTIPLE_TABLES */
142130 
....@@ -192,7 +180,7 @@
192180 	return 0;
193181 }
194182 
195
-static void fib_flush(struct net *net)
183
+void fib_flush(struct net *net)
196184 {
197185 	int flushed = 0;
198186 	unsigned int h;
....@@ -234,7 +222,9 @@
234222 	if (table) {
235223 		ret = RTN_UNICAST;
236224 		if (!fib_table_lookup(table, &fl4, &res, FIB_LOOKUP_NOREF)) {
237
-			if (!dev || dev == res.fi->fib_dev)
225
+			struct fib_nh_common *nhc = fib_info_nhc(res.fi, 0);
226
+
227
+			if (!dev || dev == nhc->nhc_dev)
238228 				ret = res.type;
239229 		}
240230 	}
....@@ -307,13 +297,40 @@
307297 			.flowi4_mark = vmark ? skb->mark : 0,
308298 		};
309299 		if (!fib_lookup(net, &fl4, &res, 0))
310
-			return FIB_RES_PREFSRC(net, res);
300
+			return fib_result_prefsrc(net, &res);
311301 	} else {
312302 		scope = RT_SCOPE_LINK;
313303 	}
314304 
315305 	return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
316306 }
307
+
308
+bool fib_info_nh_uses_dev(struct fib_info *fi, const struct net_device *dev)
309
+{
310
+	bool dev_match = false;
311
+#ifdef CONFIG_IP_ROUTE_MULTIPATH
312
+	if (unlikely(fi->nh)) {
313
+		dev_match = nexthop_uses_dev(fi->nh, dev);
314
+	} else {
315
+		int ret;
316
+
317
+		for (ret = 0; ret < fib_info_num_path(fi); ret++) {
318
+			const struct fib_nh_common *nhc = fib_info_nhc(fi, ret);
319
+
320
+			if (nhc_l3mdev_matches_dev(nhc, dev)) {
321
+				dev_match = true;
322
+				break;
323
+			}
324
+		}
325
+	}
326
+#else
327
+	if (fib_info_nhc(fi, 0)->nhc_dev == dev)
328
+		dev_match = true;
329
+#endif
330
+
331
+	return dev_match;
332
+}
333
+EXPORT_SYMBOL_GPL(fib_info_nh_uses_dev);
317334 
318335 /* Given (packet source, input interface) and optional (dst, oif, tos):
319336  * - (main) check, that source is valid i.e. not broadcast or our local
....@@ -345,6 +362,7 @@
345362 	fl4.flowi4_tun_key.tun_id = 0;
346363 	fl4.flowi4_flags = 0;
347364 	fl4.flowi4_uid = sock_net_uid(net, NULL);
365
+	fl4.flowi4_multipath_hash = 0;
348366 
349367 	no_addr = idev->ifa_list == NULL;
350368 
....@@ -363,26 +381,15 @@
363381 	    (res.type != RTN_LOCAL || !IN_DEV_ACCEPT_LOCAL(idev)))
364382 		goto e_inval;
365383 	fib_combine_itag(itag, &res);
366
-	dev_match = false;
367384 
368
-#ifdef CONFIG_IP_ROUTE_MULTIPATH
369
-	for (ret = 0; ret < res.fi->fib_nhs; ret++) {
370
-		struct fib_nh *nh = &res.fi->fib_nh[ret];
371
-
372
-		if (nh->nh_dev == dev) {
373
-			dev_match = true;
374
-			break;
375
-		} else if (l3mdev_master_ifindex_rcu(nh->nh_dev) == dev->ifindex) {
376
-			dev_match = true;
377
-			break;
378
-		}
379
-	}
380
-#else
381
-	if (FIB_RES_DEV(res) == dev)
382
-		dev_match = true;
383
-#endif
385
+	dev_match = fib_info_nh_uses_dev(res.fi, dev);
386
+	/* This is not common, loopback packets retain skb_dst so normally they
387
+	 * would not even hit this slow path.
388
+	 */
389
+	dev_match = dev_match || (res.type == RTN_LOCAL &&
390
+				  dev == net->loopback_dev);
384391 	if (dev_match) {
385
-		ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
392
+		ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_LINK;
386393 		return ret;
387394 	}
388395 	if (no_addr)
....@@ -394,7 +401,7 @@
394401 	ret = 0;
395402 	if (fib_lookup(net, &fl4, &res, FIB_LOOKUP_IGNORE_LINKSTATE) == 0) {
396403 		if (res.type == RTN_UNICAST)
397
-			ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
404
+			ret = FIB_RES_NHC(res)->nhc_scope >= RT_SCOPE_LINK;
398405 	}
399406 	return ret;
400407 
....@@ -532,14 +539,22 @@
532539 		cfg->fc_oif = dev->ifindex;
533540 		cfg->fc_table = l3mdev_fib_table(dev);
534541 		if (colon) {
535
-			struct in_ifaddr *ifa;
536
-			struct in_device *in_dev = __in_dev_get_rtnl(dev);
542
+			const struct in_ifaddr *ifa;
543
+			struct in_device *in_dev;
544
+
545
+			in_dev = __in_dev_get_rtnl(dev);
537546 			if (!in_dev)
538547 				return -ENODEV;
548
+
539549 			*colon = ':';
540
-			for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
550
+
551
+			rcu_read_lock();
552
+			in_dev_for_each_ifa_rcu(ifa, in_dev) {
541553 				if (strcmp(ifa->ifa_label, devname) == 0)
542554 					break;
555
+			}
556
+			rcu_read_unlock();
557
+
543558 			if (!ifa)
544559 				return -ENODEV;
545560 			cfg->fc_prefsrc = ifa->ifa_local;
....@@ -550,17 +565,21 @@
550565 	if (rt->rt_gateway.sa_family == AF_INET && addr) {
551566 		unsigned int addr_type;
552567 
553
-		cfg->fc_gw = addr;
568
+		cfg->fc_gw4 = addr;
569
+		cfg->fc_gw_family = AF_INET;
554570 		addr_type = inet_addr_type_table(net, addr, cfg->fc_table);
555571 		if (rt->rt_flags & RTF_GATEWAY &&
556572 		    addr_type == RTN_UNICAST)
557573 			cfg->fc_scope = RT_SCOPE_UNIVERSE;
558574 	}
559575 
576
+	if (!cfg->fc_table)
577
+		cfg->fc_table = RT_TABLE_MAIN;
578
+
560579 	if (cmd == SIOCDELRT)
561580 		return 0;
562581 
563
-	if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
582
+	if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw_family)
564583 		return -EINVAL;
565584 
566585 	if (cfg->fc_scope == RT_SCOPE_NOWHERE)
....@@ -636,6 +655,7 @@
636655 }
637656 
638657 const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
658
+	[RTA_UNSPEC]		= { .strict_start_type = RTA_DPORT + 1 },
639659 	[RTA_DST]		= { .type = NLA_U32 },
640660 	[RTA_SRC]		= { .type = NLA_U32 },
641661 	[RTA_IIF]		= { .type = NLA_U32 },
....@@ -654,18 +674,64 @@
654674 	[RTA_IP_PROTO]		= { .type = NLA_U8 },
655675 	[RTA_SPORT]		= { .type = NLA_U16 },
656676 	[RTA_DPORT]		= { .type = NLA_U16 },
677
+	[RTA_NH_ID]		= { .type = NLA_U32 },
657678 };
679
+
680
+int fib_gw_from_via(struct fib_config *cfg, struct nlattr *nla,
681
+		    struct netlink_ext_ack *extack)
682
+{
683
+	struct rtvia *via;
684
+	int alen;
685
+
686
+	if (nla_len(nla) < offsetof(struct rtvia, rtvia_addr)) {
687
+		NL_SET_ERR_MSG(extack, "Invalid attribute length for RTA_VIA");
688
+		return -EINVAL;
689
+	}
690
+
691
+	via = nla_data(nla);
692
+	alen = nla_len(nla) - offsetof(struct rtvia, rtvia_addr);
693
+
694
+	switch (via->rtvia_family) {
695
+	case AF_INET:
696
+		if (alen != sizeof(__be32)) {
697
+			NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_VIA");
698
+			return -EINVAL;
699
+		}
700
+		cfg->fc_gw_family = AF_INET;
701
+		cfg->fc_gw4 = *((__be32 *)via->rtvia_addr);
702
+		break;
703
+	case AF_INET6:
704
+#if IS_ENABLED(CONFIG_IPV6)
705
+		if (alen != sizeof(struct in6_addr)) {
706
+			NL_SET_ERR_MSG(extack, "Invalid IPv6 address in RTA_VIA");
707
+			return -EINVAL;
708
+		}
709
+		cfg->fc_gw_family = AF_INET6;
710
+		cfg->fc_gw6 = *((struct in6_addr *)via->rtvia_addr);
711
+#else
712
+		NL_SET_ERR_MSG(extack, "IPv6 support not enabled in kernel");
713
+		return -EINVAL;
714
+#endif
715
+		break;
716
+	default:
717
+		NL_SET_ERR_MSG(extack, "Unsupported address family in RTA_VIA");
718
+		return -EINVAL;
719
+	}
720
+
721
+	return 0;
722
+}
658723 
659724 static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
660725 			     struct nlmsghdr *nlh, struct fib_config *cfg,
661726 			     struct netlink_ext_ack *extack)
662727 {
728
+	bool has_gw = false, has_via = false;
663729 	struct nlattr *attr;
664730 	int err, remaining;
665731 	struct rtmsg *rtm;
666732 
667
-	err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy,
668
-			     extack);
733
+	err = nlmsg_validate_deprecated(nlh, sizeof(*rtm), RTA_MAX,
734
+					rtm_ipv4_policy, extack);
669735 	if (err < 0)
670736 		goto errout;
671737 
....@@ -700,12 +766,17 @@
700766 			cfg->fc_oif = nla_get_u32(attr);
701767 			break;
702768 		case RTA_GATEWAY:
703
-			cfg->fc_gw = nla_get_be32(attr);
769
+			has_gw = true;
770
+			cfg->fc_gw4 = nla_get_be32(attr);
771
+			if (cfg->fc_gw4)
772
+				cfg->fc_gw_family = AF_INET;
704773 			break;
705774 		case RTA_VIA:
706
-			NL_SET_ERR_MSG(extack, "IPv4 does not support RTA_VIA attribute");
707
-			err = -EINVAL;
708
-			goto errout;
775
+			has_via = true;
776
+			err = fib_gw_from_via(cfg, attr, extack);
777
+			if (err)
778
+				goto errout;
779
+			break;
709780 		case RTA_PRIORITY:
710781 			cfg->fc_priority = nla_get_u32(attr);
711782 			break;
....@@ -741,8 +812,29 @@
741812 			if (err < 0)
742813 				goto errout;
743814 			break;
815
+		case RTA_NH_ID:
816
+			cfg->fc_nh_id = nla_get_u32(attr);
817
+			break;
744818 		}
745819 	}
820
+
821
+	if (cfg->fc_nh_id) {
822
+		if (cfg->fc_oif || cfg->fc_gw_family ||
823
+		    cfg->fc_encap || cfg->fc_mp) {
824
+			NL_SET_ERR_MSG(extack,
825
+				       "Nexthop specification and nexthop id are mutually exclusive");
826
+			return -EINVAL;
827
+		}
828
+	}
829
+
830
+	if (has_gw && has_via) {
831
+		NL_SET_ERR_MSG(extack,
832
+			       "Nexthop configuration can not contain both GATEWAY and VIA");
833
+		return -EINVAL;
834
+	}
835
+
836
+	if (!cfg->fc_table)
837
+		cfg->fc_table = RT_TABLE_MAIN;
746838 
747839 	return 0;
748840 errout:
....@@ -760,6 +852,12 @@
760852 	err = rtm_to_fib_config(net, skb, nlh, &cfg, extack);
761853 	if (err < 0)
762854 		goto errout;
855
+
856
+	if (cfg.fc_nh_id && !nexthop_find_by_id(net, cfg.fc_nh_id)) {
857
+		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
858
+		err = -EINVAL;
859
+		goto errout;
860
+	}
763861 
764862 	tb = fib_get_table(net, cfg.fc_table);
765863 	if (!tb) {
....@@ -798,8 +896,85 @@
798896 	return err;
799897 }
800898 
899
+int ip_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
900
+			  struct fib_dump_filter *filter,
901
+			  struct netlink_callback *cb)
902
+{
903
+	struct netlink_ext_ack *extack = cb->extack;
904
+	struct nlattr *tb[RTA_MAX + 1];
905
+	struct rtmsg *rtm;
906
+	int err, i;
907
+
908
+	ASSERT_RTNL();
909
+
910
+	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
911
+		NL_SET_ERR_MSG(extack, "Invalid header for FIB dump request");
912
+		return -EINVAL;
913
+	}
914
+
915
+	rtm = nlmsg_data(nlh);
916
+	if (rtm->rtm_dst_len || rtm->rtm_src_len  || rtm->rtm_tos   ||
917
+	    rtm->rtm_scope) {
918
+		NL_SET_ERR_MSG(extack, "Invalid values in header for FIB dump request");
919
+		return -EINVAL;
920
+	}
921
+
922
+	if (rtm->rtm_flags & ~(RTM_F_CLONED | RTM_F_PREFIX)) {
923
+		NL_SET_ERR_MSG(extack, "Invalid flags for FIB dump request");
924
+		return -EINVAL;
925
+	}
926
+	if (rtm->rtm_flags & RTM_F_CLONED)
927
+		filter->dump_routes = false;
928
+	else
929
+		filter->dump_exceptions = false;
930
+
931
+	filter->flags    = rtm->rtm_flags;
932
+	filter->protocol = rtm->rtm_protocol;
933
+	filter->rt_type  = rtm->rtm_type;
934
+	filter->table_id = rtm->rtm_table;
935
+
936
+	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
937
+					    rtm_ipv4_policy, extack);
938
+	if (err < 0)
939
+		return err;
940
+
941
+	for (i = 0; i <= RTA_MAX; ++i) {
942
+		int ifindex;
943
+
944
+		if (!tb[i])
945
+			continue;
946
+
947
+		switch (i) {
948
+		case RTA_TABLE:
949
+			filter->table_id = nla_get_u32(tb[i]);
950
+			break;
951
+		case RTA_OIF:
952
+			ifindex = nla_get_u32(tb[i]);
953
+			filter->dev = __dev_get_by_index(net, ifindex);
954
+			if (!filter->dev)
955
+				return -ENODEV;
956
+			break;
957
+		default:
958
+			NL_SET_ERR_MSG(extack, "Unsupported attribute in dump request");
959
+			return -EINVAL;
960
+		}
961
+	}
962
+
963
+	if (filter->flags || filter->protocol || filter->rt_type ||
964
+	    filter->table_id || filter->dev) {
965
+		filter->filter_set = 1;
966
+		cb->answer_flags = NLM_F_DUMP_FILTERED;
967
+	}
968
+
969
+	return 0;
970
+}
971
+EXPORT_SYMBOL_GPL(ip_valid_fib_dump_req);
972
+
801973 static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
802974 {
975
+	struct fib_dump_filter filter = { .dump_routes = true,
976
+					  .dump_exceptions = true };
977
+	const struct nlmsghdr *nlh = cb->nlh;
803978 	struct net *net = sock_net(skb->sk);
804979 	unsigned int h, s_h;
805980 	unsigned int e = 0, s_e;
....@@ -807,9 +982,35 @@
807982 	struct hlist_head *head;
808983 	int dumped = 0, err;
809984 
810
-	if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
811
-	    ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
985
+	if (cb->strict_check) {
986
+		err = ip_valid_fib_dump_req(net, nlh, &filter, cb);
987
+		if (err < 0)
988
+			return err;
989
+	} else if (nlmsg_len(nlh) >= sizeof(struct rtmsg)) {
990
+		struct rtmsg *rtm = nlmsg_data(nlh);
991
+
992
+		filter.flags = rtm->rtm_flags & (RTM_F_PREFIX | RTM_F_CLONED);
993
+	}
994
+
995
+	/* ipv4 does not use prefix flag */
996
+	if (filter.flags & RTM_F_PREFIX)
812997 		return skb->len;
998
+
999
+	if (filter.table_id) {
1000
+		tb = fib_get_table(net, filter.table_id);
1001
+		if (!tb) {
1002
+			if (rtnl_msg_family(cb->nlh) != PF_INET)
1003
+				return skb->len;
1004
+
1005
+			NL_SET_ERR_MSG(cb->extack, "ipv4: FIB table does not exist");
1006
+			return -ENOENT;
1007
+		}
1008
+
1009
+		rcu_read_lock();
1010
+		err = fib_table_dump(tb, skb, cb, &filter);
1011
+		rcu_read_unlock();
1012
+		return skb->len ? : err;
1013
+	}
8131014 
8141015 	s_h = cb->args[0];
8151016 	s_e = cb->args[1];
....@@ -825,7 +1026,7 @@
8251026 			if (dumped)
8261027 				memset(&cb->args[2], 0, sizeof(cb->args) -
8271028 						 2 * sizeof(cb->args[0]));
828
-			err = fib_table_dump(tb, skb, cb);
1029
+			err = fib_table_dump(tb, skb, cb, &filter);
8291030 			if (err < 0) {
8301031 				if (likely(skb->len))
8311032 					goto out;
....@@ -917,9 +1118,11 @@
9171118 		return;
9181119 
9191120 	/* Add broadcast address, if it is explicitly assigned. */
920
-	if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
1121
+	if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF)) {
9211122 		fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32,
9221123 			  prim, 0);
1124
+		arp_invalidate(dev, ifa->ifa_broadcast, false);
1125
+	}
9231126 
9241127 	if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
9251128 	    (prefix != addr || ifa->ifa_prefixlen < 32)) {
....@@ -935,6 +1138,7 @@
9351138 				  prim, 0);
9361139 			fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
9371140 				  32, prim, 0);
1141
+			arp_invalidate(dev, prefix | ~mask, false);
9381142 		}
9391143 	}
9401144 }
....@@ -1015,8 +1219,8 @@
10151219 	 *
10161220 	 * Scan address list to be sure that addresses are really gone.
10171221 	 */
1018
-
1019
-	for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
1222
+	rcu_read_lock();
1223
+	in_dev_for_each_ifa_rcu(ifa1, in_dev) {
10201224 		if (ifa1 == ifa) {
10211225 			/* promotion, keep the IP */
10221226 			gone = 0;
....@@ -1084,6 +1288,7 @@
10841288 			}
10851289 		}
10861290 	}
1291
+	rcu_read_unlock();
10871292 
10881293 no_promotions:
10891294 	if (!(ok & BRD_OK))
....@@ -1253,6 +1458,7 @@
12531458 	struct netdev_notifier_info_ext *info_ext = ptr;
12541459 	struct in_device *in_dev;
12551460 	struct net *net = dev_net(dev);
1461
+	struct in_ifaddr *ifa;
12561462 	unsigned int flags;
12571463 
12581464 	if (event == NETDEV_UNREGISTER) {
....@@ -1267,9 +1473,9 @@
12671473 
12681474 	switch (event) {
12691475 	case NETDEV_UP:
1270
-		for_ifa(in_dev) {
1476
+		in_dev_for_each_ifa_rtnl(ifa, in_dev) {
12711477 			fib_add_ifaddr(ifa);
1272
-		} endfor_ifa(in_dev);
1478
+		}
12731479 #ifdef CONFIG_IP_ROUTE_MULTIPATH
12741480 		fib_sync_up(dev, RTNH_F_DEAD);
12751481 #endif
....@@ -1381,7 +1587,7 @@
13811587 	int error;
13821588 
13831589 #ifdef CONFIG_IP_ROUTE_CLASSID
1384
-	net->ipv4.fib_num_tclassid_users = 0;
1590
+	atomic_set(&net->ipv4.fib_num_tclassid_users, 0);
13851591 #endif
13861592 	error = ip_fib_net_init(net);
13871593 	if (error < 0)