forked from ~ljy/RK356X_SDK_RELEASE

blame | history | patch | commit | commitdiff
hc
2024-10-22 8ac6c7a54ed1b98d142dce24b11c6de6a1e239a5 
 kernel/net/ipv6/icmp.c
....@@ -1,3 +1,4 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23  *	Internet Control Message Protocol (ICMPv6)
34  *	Linux INET6 implementation
....@@ -8,11 +9,6 @@
89  *	Based on net/ipv4/icmp.c
910  *
1011  *	RFC 1885
11
- *
12
- *	This program is free software; you can redistribute it and/or
13
- *      modify it under the terms of the GNU General Public License
14
- *      as published by the Free Software Foundation; either version
15
- *      2 of the License, or (at your option) any later version.
1612  */
1713 
1814 /*
....@@ -79,12 +75,12 @@
7975  *
8076  *	On SMP we have one ICMP socket per-cpu.
8177  */
82
-static inline struct sock *icmpv6_sk(struct net *net)
78
+static struct sock *icmpv6_sk(struct net *net)
8379 {
84
-	return net->ipv6.icmp_sk[smp_processor_id()];
80
+	return this_cpu_read(*net->ipv6.icmp_sk);
8581 }
8682 
87
-static void icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
83
+static int icmpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
8884 		       u8 type, u8 code, int offset, __be32 info)
8985 {
9086 	/* icmpv6_notify checks 8 bytes can be pulled, icmp6hdr is 8 bytes */
....@@ -100,6 +96,8 @@
10096 	if (!(type & ICMPV6_INFOMSG_MASK))
10197 		if (icmp6->icmp6_type == ICMPV6_ECHO_REQUEST)
10298 			ping_err(skb, offset, ntohl(info));
99
+
100
+	return 0;
103101 }
104102 
105103 static int icmpv6_rcv(struct sk_buff *skb);
....@@ -160,28 +158,33 @@
160158 		tp = skb_header_pointer(skb,
161159 			ptr+offsetof(struct icmp6hdr, icmp6_type),
162160 			sizeof(_type), &_type);
163
-		if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
161
+
162
+		/* Based on RFC 8200, Section 4.5 Fragment Header, return
163
+		 * false if this is a fragment packet with no icmp header info.
164
+		 */
165
+		if (!tp && frag_off != 0)
166
+			return false;
167
+		else if (!tp || !(*tp & ICMPV6_INFOMSG_MASK))
164168 			return true;
165169 	}
166170 	return false;
167171 }
168172 
169
-static bool icmpv6_mask_allow(int type)
173
+static bool icmpv6_mask_allow(struct net *net, int type)
170174 {
171
-	/* Informational messages are not limited. */
172
-	if (type & ICMPV6_INFOMSG_MASK)
175
+	if (type > ICMPV6_MSG_MAX)
173176 		return true;
174177 
175
-	/* Do not limit pmtu discovery, it would break it. */
176
-	if (type == ICMPV6_PKT_TOOBIG)
178
+	/* Limit if icmp type is set in ratemask. */
179
+	if (!test_bit(type, net->ipv6.sysctl.icmpv6_ratemask))
177180 		return true;
178181 
179182 	return false;
180183 }
181184 
182
-static bool icmpv6_global_allow(int type)
185
+static bool icmpv6_global_allow(struct net *net, int type)
183186 {
184
-	if (icmpv6_mask_allow(type))
187
+	if (icmpv6_mask_allow(net, type))
185188 		return true;
186189 
187190 	if (icmp_global_allow())
....@@ -200,7 +203,7 @@
200203 	struct dst_entry *dst;
201204 	bool res = false;
202205 
203
-	if (icmpv6_mask_allow(type))
206
+	if (icmpv6_mask_allow(net, type))
204207 		return true;
205208 
206209 	/*
....@@ -227,6 +230,25 @@
227230 		res = inet_peer_xrlim_allow(peer, tmo);
228231 		if (peer)
229232 			inet_putpeer(peer);
233
+	}
234
+	dst_release(dst);
235
+	return res;
236
+}
237
+
238
+static bool icmpv6_rt_has_prefsrc(struct sock *sk, u8 type,
239
+				  struct flowi6 *fl6)
240
+{
241
+	struct net *net = sock_net(sk);
242
+	struct dst_entry *dst;
243
+	bool res = false;
244
+
245
+	dst = ip6_route_output(net, sk, fl6);
246
+	if (!dst->error) {
247
+		struct rt6_info *rt = (struct rt6_info *)dst;
248
+		struct in6_addr prefsrc;
249
+
250
+		rt6_get_prefsrc(rt, &prefsrc);
251
+		res = !ipv6_addr_any(&prefsrc);
230252 	}
231253 	dst_release(dst);
232254 	return res;
....@@ -298,10 +320,10 @@
298320 {
299321 	struct icmpv6_msg *msg = (struct icmpv6_msg *) from;
300322 	struct sk_buff *org_skb = msg->skb;
301
-	__wsum csum = 0;
323
+	__wsum csum;
302324 
303325 	csum = skb_copy_and_csum_bits(org_skb, msg->offset + offset,
304
-				      to, len, csum);
326
+				      to, len);
305327 	skb->csum = csum_block_add(skb->csum, csum, odd);
306328 	if (!(msg->type & ICMPV6_INFOMSG_MASK))
307329 		nf_ct_attach(skb, org_skb);
....@@ -395,23 +417,31 @@
395417 	return ERR_PTR(err);
396418 }
397419 
398
-static int icmp6_iif(const struct sk_buff *skb)
420
+static struct net_device *icmp6_dev(const struct sk_buff *skb)
399421 {
400
-	int iif = skb->dev->ifindex;
422
+	struct net_device *dev = skb->dev;
401423 
402424 	/* for local traffic to local address, skb dev is the loopback
403425 	 * device. Check if there is a dst attached to the skb and if so
404426 	 * get the real device index. Same is needed for replies to a link
405427 	 * local address on a device enslaved to an L3 master device
406428 	 */
407
-	if (unlikely(iif == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) {
429
+	if (unlikely(dev->ifindex == LOOPBACK_IFINDEX || netif_is_l3_master(skb->dev))) {
408430 		const struct rt6_info *rt6 = skb_rt6_info(skb);
409431 
410
-		if (rt6)
411
-			iif = rt6->rt6i_idev->dev->ifindex;
432
+		/* The destination could be an external IP in Ext Hdr (SRv6, RPL, etc.),
433
+		 * and ip6_null_entry could be set to skb if no route is found.
434
+		 */
435
+		if (rt6 && rt6->rt6i_idev)
436
+			dev = rt6->rt6i_idev->dev;
412437 	}
413438 
414
-	return iif;
439
+	return dev;
440
+}
441
+
442
+static int icmp6_iif(const struct sk_buff *skb)
443
+{
444
+	return icmp6_dev(skb)->ifindex;
415445 }
416446 
417447 /*
....@@ -480,8 +510,11 @@
480510 	if (__ipv6_addr_needs_scope_id(addr_type)) {
481511 		iif = icmp6_iif(skb);
482512 	} else {
483
-		dst = skb_dst(skb);
484
-		iif = l3mdev_master_ifindex(dst ? dst->dev : skb->dev);
513
+		/*
514
+		 * The source device is used for looking up which routing table
515
+		 * to use for sending an ICMP error.
516
+		 */
517
+		iif = l3mdev_master_ifindex(skb->dev);
485518 	}
486519 
487520 	/*
....@@ -509,31 +542,42 @@
509542 	local_bh_disable();
510543 
511544 	/* Check global sysctl_icmp_msgs_per_sec ratelimit */
512
-	if (!(skb->dev->flags&IFF_LOOPBACK) && !icmpv6_global_allow(type))
545
+	if (!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, type))
513546 		goto out_bh_enable;
514547 
515548 	mip6_addr_swap(skb, parm);
549
+
550
+	sk = icmpv6_xmit_lock(net);
551
+	if (!sk)
552
+		goto out_bh_enable;
516553 
517554 	memset(&fl6, 0, sizeof(fl6));
518555 	fl6.flowi6_proto = IPPROTO_ICMPV6;
519556 	fl6.daddr = hdr->saddr;
520557 	if (force_saddr)
521558 		saddr = force_saddr;
522
-	if (saddr)
559
+	if (saddr) {
523560 		fl6.saddr = *saddr;
561
+	} else if (!icmpv6_rt_has_prefsrc(sk, type, &fl6)) {
562
+		/* select a more meaningful saddr from input if */
563
+		struct net_device *in_netdev;
564
+
565
+		in_netdev = dev_get_by_index(net, parm->iif);
566
+		if (in_netdev) {
567
+			ipv6_dev_get_saddr(net, in_netdev, &fl6.daddr,
568
+					   inet6_sk(sk)->srcprefs,
569
+					   &fl6.saddr);
570
+			dev_put(in_netdev);
571
+		}
572
+	}
524573 	fl6.flowi6_mark = mark;
525574 	fl6.flowi6_oif = iif;
526575 	fl6.fl6_icmp_type = type;
527576 	fl6.fl6_icmp_code = code;
528577 	fl6.flowi6_uid = sock_net_uid(net, NULL);
529578 	fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, NULL);
530
-	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
579
+	security_skb_classify_flow(skb, flowi6_to_flowi_common(&fl6));
531580 
532
-	sk = icmpv6_xmit_lock(net);
533
-	if (!sk)
534
-		goto out_bh_enable;
535
-
536
-	sk->sk_mark = mark;
537581 	np = inet6_sk(sk);
538582 
539583 	if (!icmpv6_xrlim_allow(sk, type, &fl6))
....@@ -550,6 +594,7 @@
550594 		fl6.flowi6_oif = np->ucast_oif;
551595 
552596 	ipcm6_init_sk(&ipc6, np);
597
+	ipc6.sockc.mark = mark;
553598 	fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
554599 
555600 	dst = icmpv6_route_lookup(net, skb, sk, &fl6);
....@@ -682,18 +727,29 @@
682727 	struct dst_entry *dst;
683728 	struct ipcm6_cookie ipc6;
684729 	u32 mark = IP6_REPLY_MARK(net, skb->mark);
730
+	bool acast;
731
+
732
+	if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) &&
733
+	    net->ipv6.sysctl.icmpv6_echo_ignore_multicast)
734
+		return;
685735 
686736 	saddr = &ipv6_hdr(skb)->daddr;
687737 
738
+	acast = ipv6_anycast_destination(skb_dst(skb), saddr);
739
+	if (acast && net->ipv6.sysctl.icmpv6_echo_ignore_anycast)
740
+		return;
741
+
688742 	if (!ipv6_unicast_destination(skb) &&
689
-	    !(net->ipv6.sysctl.anycast_src_echo_reply &&
690
-	      ipv6_anycast_destination(skb_dst(skb), saddr)))
743
+	    !(net->ipv6.sysctl.anycast_src_echo_reply && acast))
691744 		saddr = NULL;
692745 
693746 	memcpy(&tmp_hdr, icmph, sizeof(tmp_hdr));
694747 	tmp_hdr.icmp6_type = ICMPV6_ECHO_REPLY;
695748 
696749 	memset(&fl6, 0, sizeof(fl6));
750
+	if (net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES)
751
+		fl6.flowlabel = ip6_flowlabel(ipv6_hdr(skb));
752
+
697753 	fl6.flowi6_proto = IPPROTO_ICMPV6;
698754 	fl6.daddr = ipv6_hdr(skb)->saddr;
699755 	if (saddr)
....@@ -702,13 +758,12 @@
702758 	fl6.fl6_icmp_type = ICMPV6_ECHO_REPLY;
703759 	fl6.flowi6_mark = mark;
704760 	fl6.flowi6_uid = sock_net_uid(net, NULL);
705
-	security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
761
+	security_skb_classify_flow(skb, flowi6_to_flowi_common(&fl6));
706762 
707763 	local_bh_disable();
708764 	sk = icmpv6_xmit_lock(net);
709765 	if (!sk)
710766 		goto out_bh_enable;
711
-	sk->sk_mark = mark;
712767 	np = inet6_sk(sk);
713768 
714769 	if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
....@@ -722,6 +777,11 @@
722777 	if (IS_ERR(dst))
723778 		goto out;
724779 
780
+	/* Check the ratelimit */
781
+	if ((!(skb->dev->flags & IFF_LOOPBACK) && !icmpv6_global_allow(net, ICMPV6_ECHO_REPLY)) ||
782
+	    !icmpv6_xrlim_allow(sk, ICMPV6_ECHO_REPLY, &fl6))
783
+		goto out_dst_release;
784
+
725785 	idev = __in6_dev_get(skb->dev);
726786 
727787 	msg.skb = skb;
....@@ -731,6 +791,7 @@
731791 	ipcm6_init_sk(&ipc6, np);
732792 	ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
733793 	ipc6.tclass = ipv6_get_dsfield(ipv6_hdr(skb));
794
+	ipc6.sockc.mark = mark;
734795 
735796 	if (ip6_append_data(sk, icmpv6_getfrag, &msg,
736797 			    skb->len + sizeof(struct icmp6hdr),
....@@ -742,6 +803,7 @@
742803 		icmpv6_push_pending_frames(sk, &fl6, &tmp_hdr,
743804 					   skb->len + sizeof(struct icmp6hdr));
744805 	}
806
+out_dst_release:
745807 	dst_release(dst);
746808 out:
747809 	icmpv6_xmit_unlock(sk);
....@@ -800,7 +862,7 @@
800862 static int icmpv6_rcv(struct sk_buff *skb)
801863 {
802864 	struct net *net = dev_net(skb->dev);
803
-	struct net_device *dev = skb->dev;
865
+	struct net_device *dev = icmp6_dev(skb);
804866 	struct inet6_dev *idev = __in6_dev_get(dev);
805867 	const struct in6_addr *saddr, *daddr;
806868 	struct icmp6hdr *hdr;
....@@ -868,7 +930,7 @@
868930 		hdr = icmp6_hdr(skb);
869931 
870932 		/* to notify */
871
-		/* fall through */
933
+		fallthrough;
872934 	case ICMPV6_DEST_UNREACH:
873935 	case ICMPV6_TIME_EXCEED:
874936 	case ICMPV6_PARAMPROB:
....@@ -949,16 +1011,24 @@
9491011 	fl6->fl6_icmp_type	= type;
9501012 	fl6->fl6_icmp_code	= 0;
9511013 	fl6->flowi6_oif		= oif;
952
-	security_sk_classify_flow(sk, flowi6_to_flowi(fl6));
1014
+	security_sk_classify_flow(sk, flowi6_to_flowi_common(fl6));
1015
+}
1016
+
1017
+static void __net_exit icmpv6_sk_exit(struct net *net)
1018
+{
1019
+	int i;
1020
+
1021
+	for_each_possible_cpu(i)
1022
+		inet_ctl_sock_destroy(*per_cpu_ptr(net->ipv6.icmp_sk, i));
1023
+	free_percpu(net->ipv6.icmp_sk);
9531024 }
9541025 
9551026 static int __net_init icmpv6_sk_init(struct net *net)
9561027 {
9571028 	struct sock *sk;
958
-	int err, i, j;
1029
+	int err, i;
9591030 
960
-	net->ipv6.icmp_sk =
961
-		kcalloc(nr_cpu_ids, sizeof(struct sock *), GFP_KERNEL);
1031
+	net->ipv6.icmp_sk = alloc_percpu(struct sock *);
9621032 	if (!net->ipv6.icmp_sk)
9631033 		return -ENOMEM;
9641034 
....@@ -971,7 +1041,7 @@
9711041 			goto fail;
9721042 		}
9731043 
974
-		net->ipv6.icmp_sk[i] = sk;
1044
+		*per_cpu_ptr(net->ipv6.icmp_sk, i) = sk;
9751045 
9761046 		/* Enough space for 2 64K ICMP packets, including
9771047 		 * sk_buff struct overhead.
....@@ -981,20 +1051,8 @@
9811051 	return 0;
9821052 
9831053  fail:
984
-	for (j = 0; j < i; j++)
985
-		inet_ctl_sock_destroy(net->ipv6.icmp_sk[j]);
986
-	kfree(net->ipv6.icmp_sk);
1054
+	icmpv6_sk_exit(net);
9871055 	return err;
988
-}
989
-
990
-static void __net_exit icmpv6_sk_exit(struct net *net)
991
-{
992
-	int i;
993
-
994
-	for_each_possible_cpu(i) {
995
-		inet_ctl_sock_destroy(net->ipv6.icmp_sk[i]);
996
-	}
997
-	kfree(net->ipv6.icmp_sk);
9981056 }
9991057 
10001058 static struct pernet_operations icmpv6_sk_ops = {
....@@ -1118,6 +1176,27 @@
11181176 		.mode		= 0644,
11191177 		.proc_handler = proc_dointvec,
11201178 	},
1179
+	{
1180
+		.procname	= "echo_ignore_multicast",
1181
+		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_multicast,
1182
+		.maxlen		= sizeof(int),
1183
+		.mode		= 0644,
1184
+		.proc_handler = proc_dointvec,
1185
+	},
1186
+	{
1187
+		.procname	= "echo_ignore_anycast",
1188
+		.data		= &init_net.ipv6.sysctl.icmpv6_echo_ignore_anycast,
1189
+		.maxlen		= sizeof(int),
1190
+		.mode		= 0644,
1191
+		.proc_handler = proc_dointvec,
1192
+	},
1193
+	{
1194
+		.procname	= "ratemask",
1195
+		.data		= &init_net.ipv6.sysctl.icmpv6_ratemask_ptr,
1196
+		.maxlen		= ICMPV6_MSG_MAX + 1,
1197
+		.mode		= 0644,
1198
+		.proc_handler = proc_do_large_bitmap,
1199
+	},
11211200 	{ },
11221201 };
11231202 
....@@ -1132,6 +1211,9 @@
11321211 	if (table) {
11331212 		table[0].data = &net->ipv6.sysctl.icmpv6_time;
11341213 		table[1].data = &net->ipv6.sysctl.icmpv6_echo_ignore_all;
1214
+		table[2].data = &net->ipv6.sysctl.icmpv6_echo_ignore_multicast;
1215
+		table[3].data = &net->ipv6.sysctl.icmpv6_echo_ignore_anycast;
1216
+		table[4].data = &net->ipv6.sysctl.icmpv6_ratemask_ptr;
11351217 	}
11361218 	return table;
11371219 }