Skip to contentSkip to conversation

netfilter: nat: force port remap to prevent shadowing well-known ports

PATCH Superseded Florian Westphal 3 participants 3 messages Dec 16, 2021

Subsystems (3)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

PATCH v3 1/1 netfilter: nat: force port remap to prevent shadowing well-known ports

If destination port is above 32k and source port below 16k assume this might cause 'port shadowing' where a 'new' inbound connection matches an existing one, e.g.

inbound X:41234 -> Y:53 matches existing conntrack entry

        Z:53 -> X:4123, where Z got natted to X.

In this case, new packet is natted to Z:53 which is likely unwanted.

We avoid the rewrite for connections that originate from local host:
port-shadowing is only possible with forwarded connections.

Also adjust test case.

v3: no need to call tuple_force_port_remap if already in random mode

---
 net/netfilter/nf_nat_core.c                  | 43 ++++++++++++++++++--
 tools/testing/selftests/netfilter/nft_nat.sh |  5 ++-
 2 files changed, 43 insertions(+), 5 deletions(-)
net/netfilter/nf_nat_core.c +40 -3
index 4d50d51db796..29d26bc977dc 100644
@@ -494,6 +494,38 @@ static void nf_nat_l4proto_unique_tuple(struct nf_conntrack_tuple *tuple,
 	goto another_round;
 }
 
+static bool tuple_force_port_remap(const struct nf_conntrack_tuple *tuple)
+{
+	u16 sp, dp;
+
+	switch (tuple->dst.protonum) {
+	case IPPROTO_TCP:
+		sp = ntohs(tuple->src.u.tcp.port);
+		dp = ntohs(tuple->dst.u.tcp.port);
+		break;
+	case IPPROTO_UDP:
+	case IPPROTO_UDPLITE:
+		sp = ntohs(tuple->src.u.udp.port);
+		dp = ntohs(tuple->dst.u.udp.port);
+		break;
+	default:
+		return false;
+	}
+
+	/* IANA: System port range: 1-1023,
+	 *         user port range: 1024-49151,
+	 *      private port range: 49152-65535.
+	 *
+	 * Linux default ephemeral port range is 32768-60999.
+	 *
+	 * Enforce port remapping if sport is significantly lower
+	 * than dport to prevent NAT port shadowing, i.e.
+	 * accidental match of 'new' inbound connection vs.
+	 * existing outbound one.
+	 */
+	return sp < 16384 && dp >= 32768;
+}
+
 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
  * we change the source to map into the range. For NF_INET_PRE_ROUTING
  * and NF_INET_LOCAL_OUT, we change the destination to map into the
@@ -507,11 +539,17 @@ get_unique_tuple(struct nf_conntrack_tuple *tuple,
 		 struct nf_conn *ct,
 		 enum nf_nat_manip_type maniptype)
 {
+	bool random_port = range->flags & NF_NAT_RANGE_PROTO_RANDOM_ALL;
 	const struct nf_conntrack_zone *zone;
 	struct net *net = nf_ct_net(ct);
 
 	zone = nf_ct_zone(ct);
 
+	if (maniptype == NF_NAT_MANIP_SRC &&
+	    !random_port &&
+	    !ct->local_origin)
+		random_port = tuple_force_port_remap(orig_tuple);
+
 	/* 1) If this srcip/proto/src-proto-part is currently mapped,
 	 * and that same mapping gives a unique tuple within the given
 	 * range, use that.
@@ -520,8 +558,7 @@ get_unique_tuple(struct nf_conntrack_tuple *tuple,
 	 * So far, we don't do local source mappings, so multiple
 	 * manips not an issue.
 	 */
-	if (maniptype == NF_NAT_MANIP_SRC &&
-	    !(range->flags & NF_NAT_RANGE_PROTO_RANDOM_ALL)) {
+	if (maniptype == NF_NAT_MANIP_SRC && !random_port) {
 		/* try the original tuple first */
 		if (in_range(orig_tuple, range)) {
 			if (!nf_nat_used_tuple(orig_tuple, ct)) {
@@ -545,7 +582,7 @@ get_unique_tuple(struct nf_conntrack_tuple *tuple,
 	 */
 
 	/* Only bother mapping if it's not already in range and unique */
-	if (!(range->flags & NF_NAT_RANGE_PROTO_RANDOM_ALL)) {
+	if (!random_port) {
 		if (range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) {
 			if (!(range->flags & NF_NAT_RANGE_PROTO_OFFSET) &&
 			    l4proto_in_range(tuple, maniptype,
tools/testing/selftests/netfilter/nft_nat.sh +3 -2
index d88867d2fed7..349a319a9e51 100755
@@ -880,8 +880,9 @@ EOF
 		return $ksft_skip
 	fi
 
-	# test default behaviour. Packet from ns1 to ns0 is redirected to ns2.
-	test_port_shadow "default" "CLIENT"
+	# test default behaviour. Packet from ns1 to ns0 is not redirected
+	# due to automatic port translation.
+	test_port_shadow "default" "ROUTER"
 
 	# test packet filter based mitigation: prevent forwarding of
 	# packets claiming to come from the service port.
Signature · 1 line
2.32.0
reviewer Acked-by on v3 1/1
#1 lore raw
Quoted 13 lines · Florian Westphal

If destination port is above 32k and source port below 16k assume this might cause 'port shadowing' where a 'new' inbound connection matches an existing one, e.g.

inbound X:41234 -> Y:53 matches existing conntrack entry

        Z:53 -> X:4123, where Z got natted to X.

In this case, new packet is natted to Z:53 which is likely unwanted.

We avoid the rewrite for connections that originate from local host:
port-shadowing is only possible with forwarded connections.

Quoted 7 lines

Also adjust test case.

v3: no need to call tuple_force_port_remap if already in random mode

Thanks for the quick follow-up!

#2 lore raw
In reply to Florian Westphal
I love your patch! Yet something to improve:

Hi Florian,

I love your patch! Yet something to improve:

[auto build test ERROR on nf/master]

url:    https://github.com/0day-ci/linux/commits/Florian-Westphal/netfilter-nat-force-port-remap-to-prevent-shadowing-well-known-ports/20211216-232930
base:   https://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf.git master
config: arm-randconfig-r005-20211216 (https://download.01.org/0day-ci/archive/20211217/202112170757.knetsZWh-lkp(a)intel.com/config)
compiler: clang version 14.0.0 (https://github.com/llvm/llvm-project dd245bab9fbb364faa1581e4f92ba3119a872fba)

reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# install arm cross compiling tool for clang build
# apt-get install binutils-arm-linux-gnueabi
# https://github.com/0day-ci/linux/commit/cc216934b951862fcd3ea10c9bef2eecd84d8e6f
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Florian-Westphal/netfilter-nat-force-port-remap-to-prevent-shadowing-well-known-ports/20211216-232930 git checkout cc216934b951862fcd3ea10c9bef2eecd84d8e6f
# save the config file to linux build tree
mkdir build_dir
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=arm SHELL=/bin/bash net/netfilter/

If you fix the issue, kindly add following tag as appropriate

All errors (new ones prefixed by >>):

>> net/netfilter/nf_nat_core.c:550:11: error: no member named 'local_origin' in 'struct nf_conn'
               !ct->local_origin)
                ~~  ^

1 error generated.

vim +550 net/netfilter/nf_nat_core.c

528
529 /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING,
530 * we change the source to map into the range. For NF_INET_PRE_ROUTING
531 * and NF_INET_LOCAL_OUT, we change the destination to map into the
532 * range. It might not be possible to get a unique tuple, but we try.
533 * At worst (or if we race), we will end up with a final duplicate in
534 * __nf_conntrack_confirm and drop the packet. */
535 static void
536 get_unique_tuple(struct nf_conntrack_tuple *tuple,
537 const struct nf_conntrack_tuple *orig_tuple,
538 const struct nf_nat_range2 *range,
539 struct nf_conn *ct,
540 enum nf_nat_manip_type maniptype)
541 {
542 bool random_port = range->flags & NF_NAT_RANGE_PROTO_RANDOM_ALL;
543 const struct nf_conntrack_zone *zone;
544 struct net *net = nf_ct_net(ct);
545
546 zone = nf_ct_zone(ct);
547
548 if (maniptype == NF_NAT_MANIP_SRC &&
549 !random_port &&

 > 550		    !ct->local_origin)

551 random_port = tuple_force_port_remap(orig_tuple);
552
553 /* 1) If this srcip/proto/src-proto-part is currently mapped,
554 * and that same mapping gives a unique tuple within the given
555 * range, use that.
556 *
557 * This is only required for source (ie. NAT/masq) mappings.
558 * So far, we don't do local source mappings, so multiple
559 * manips not an issue.
560 */
561 if (maniptype == NF_NAT_MANIP_SRC && !random_port) {
562 /* try the original tuple first */
563 if (in_range(orig_tuple, range)) {
564 if (!nf_nat_used_tuple(orig_tuple, ct)) {
565 *tuple = *orig_tuple;
566 return;
567 }
568 } else if (find_appropriate_src(net, zone,
569 orig_tuple, tuple, range)) {
570 pr_debug("get_unique_tuple: Found current src map\n");
571 if (!nf_nat_used_tuple(tuple, ct))
572 return;
573 }
574 }
575
576 /* 2) Select the least-used IP/proto combination in the given range */
577 *tuple = *orig_tuple;
578 find_best_ips_proto(zone, tuple, range, ct, maniptype);
579
580 /* 3) The per-protocol part of the manip is made to map into
581 * the range to make a unique tuple.
582 */
583
584 /* Only bother mapping if it's not already in range and unique */
585 if (!random_port) {
586 if (range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) {
587 if (!(range->flags & NF_NAT_RANGE_PROTO_OFFSET) &&
588 l4proto_in_range(tuple, maniptype,
589 &range->min_proto,
590 &range->max_proto) &&
591 (range->min_proto.all == range->max_proto.all ||
592 !nf_nat_used_tuple(tuple, ct)))
593 return;
594 } else if (!nf_nat_used_tuple(tuple, ct)) {
595 return;
596 }
597 }
598
599 /* Last chance: get protocol to try to obtain unique tuple. */
600 nf_nat_l4proto_unique_tuple(tuple, range, maniptype, ct);
601 }
602

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org