* [PATCHv2 nf 1/2] netfilter: handle the connecting collision properly in nf_conntrack_proto_sctp
2023-10-03 17:17 [PATCHv2 nf 0/2] netfilter: handle the sctp collision properly and add selftest Xin Long
@ 2023-10-03 17:17 ` Xin Long
2023-10-03 17:17 ` [PATCHv2 nf 2/2] selftests: netfilter: test for sctp collision processing in nf_conntrack Xin Long
2023-10-07 14:47 ` [PATCHv2 nf 0/2] netfilter: handle the sctp collision properly and add selftest Simon Horman
2 siblings, 0 replies; 4+ messages in thread
From: Xin Long @ 2023-10-03 17:17 UTC (permalink / raw)
To: network dev, netfilter-devel, linux-sctp
Cc: davem, kuba, Eric Dumazet, Paolo Abeni, Pablo Neira Ayuso,
Jozsef Kadlecsik, Florian Westphal, Marcelo Ricardo Leitner,
Simon Horman
In Scenario A and B below, as the delayed INIT_ACK always changes the peer
vtag, SCTP ct with the incorrect vtag may cause packet loss.
Scenario A: INIT_ACK is delayed until the peer receives its own INIT_ACK
192.168.1.2 > 192.168.1.1: [INIT] [init tag: 1328086772]
192.168.1.1 > 192.168.1.2: [INIT] [init tag: 1414468151]
192.168.1.2 > 192.168.1.1: [INIT ACK] [init tag: 1328086772]
192.168.1.1 > 192.168.1.2: [INIT ACK] [init tag: 1650211246] *
192.168.1.2 > 192.168.1.1: [COOKIE ECHO]
192.168.1.1 > 192.168.1.2: [COOKIE ECHO]
192.168.1.2 > 192.168.1.1: [COOKIE ACK]
Scenario B: INIT_ACK is delayed until the peer completes its own handshake
192.168.1.2 > 192.168.1.1: sctp (1) [INIT] [init tag: 3922216408]
192.168.1.1 > 192.168.1.2: sctp (1) [INIT] [init tag: 144230885]
192.168.1.2 > 192.168.1.1: sctp (1) [INIT ACK] [init tag: 3922216408]
192.168.1.1 > 192.168.1.2: sctp (1) [COOKIE ECHO]
192.168.1.2 > 192.168.1.1: sctp (1) [COOKIE ACK]
192.168.1.1 > 192.168.1.2: sctp (1) [INIT ACK] [init tag: 3914796021] *
This patch fixes it as below:
In SCTP_CID_INIT processing:
- clear ct->proto.sctp.init[!dir] if ct->proto.sctp.init[dir] &&
ct->proto.sctp.init[!dir]. (Scenario E)
- set ct->proto.sctp.init[dir].
In SCTP_CID_INIT_ACK processing:
- drop it if !ct->proto.sctp.init[!dir] && ct->proto.sctp.vtag[!dir] &&
ct->proto.sctp.vtag[!dir] != ih->init_tag. (Scenario B, Scenario C)
- drop it if ct->proto.sctp.init[dir] && ct->proto.sctp.init[!dir] &&
ct->proto.sctp.vtag[!dir] != ih->init_tag. (Scenario A)
In SCTP_CID_COOKIE_ACK processing:
- clear ct->proto.sctp.init[dir] and ct->proto.sctp.init[!dir].
(Scenario D)
Also, it's important to allow the ct state to move forward with cookie_echo
and cookie_ack from the opposite dir for the collision scenarios.
There are also other Scenarios where it should allow the packet through,
addressed by the processing above:
Scenario C: new CT is created by INIT_ACK.
Scenario D: start INIT on the existing ESTABLISHED ct.
Scenario E: start INIT after the old collision on the existing ESTABLISHED
ct.
192.168.1.2 > 192.168.1.1: sctp (1) [INIT] [init tag: 3922216408]
192.168.1.1 > 192.168.1.2: sctp (1) [INIT] [init tag: 144230885]
(both side are stopped, then start new connection again in hours)
192.168.1.2 > 192.168.1.1: sctp (1) [INIT] [init tag: 242308742]
Fixes: 9fb9cbb1082d ("[NETFILTER]: Add nf_conntrack subsystem.")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
---
v2:
- use __be32 for vtag variable to make Sparse unhappy, as Simon noticed.
- add Fixes tag, as Simon suggested.
- replace "ih == NULL" with "!ih", as checkpatch.pl suggested.
---
include/linux/netfilter/nf_conntrack_sctp.h | 1 +
net/netfilter/nf_conntrack_proto_sctp.c | 43 ++++++++++++++++-----
2 files changed, 34 insertions(+), 10 deletions(-)
diff --git a/include/linux/netfilter/nf_conntrack_sctp.h b/include/linux/netfilter/nf_conntrack_sctp.h
index 625f491b95de..fb31312825ae 100644
--- a/include/linux/netfilter/nf_conntrack_sctp.h
+++ b/include/linux/netfilter/nf_conntrack_sctp.h
@@ -9,6 +9,7 @@ struct ip_ct_sctp {
enum sctp_conntrack state;
__be32 vtag[IP_CT_DIR_MAX];
+ u8 init[IP_CT_DIR_MAX];
u8 last_dir;
u8 flags;
};
diff --git a/net/netfilter/nf_conntrack_proto_sctp.c b/net/netfilter/nf_conntrack_proto_sctp.c
index b6bcc8f2f46b..c6bd533983c1 100644
--- a/net/netfilter/nf_conntrack_proto_sctp.c
+++ b/net/netfilter/nf_conntrack_proto_sctp.c
@@ -112,7 +112,7 @@ static const u8 sctp_conntracks[2][11][SCTP_CONNTRACK_MAX] = {
/* shutdown_ack */ {sSA, sCL, sCW, sCE, sES, sSA, sSA, sSA, sSA},
/* error */ {sCL, sCL, sCW, sCE, sES, sSS, sSR, sSA, sCL},/* Can't have Stale cookie*/
/* cookie_echo */ {sCL, sCL, sCE, sCE, sES, sSS, sSR, sSA, sCL},/* 5.2.4 - Big TODO */
-/* cookie_ack */ {sCL, sCL, sCW, sCE, sES, sSS, sSR, sSA, sCL},/* Can't come in orig dir */
+/* cookie_ack */ {sCL, sCL, sCW, sES, sES, sSS, sSR, sSA, sCL},/* Can't come in orig dir */
/* shutdown_comp*/ {sCL, sCL, sCW, sCE, sES, sSS, sSR, sCL, sCL},
/* heartbeat */ {sHS, sCL, sCW, sCE, sES, sSS, sSR, sSA, sHS},
/* heartbeat_ack*/ {sCL, sCL, sCW, sCE, sES, sSS, sSR, sSA, sHS},
@@ -126,7 +126,7 @@ static const u8 sctp_conntracks[2][11][SCTP_CONNTRACK_MAX] = {
/* shutdown */ {sIV, sCL, sCW, sCE, sSR, sSS, sSR, sSA, sIV},
/* shutdown_ack */ {sIV, sCL, sCW, sCE, sES, sSA, sSA, sSA, sIV},
/* error */ {sIV, sCL, sCW, sCL, sES, sSS, sSR, sSA, sIV},
-/* cookie_echo */ {sIV, sCL, sCW, sCE, sES, sSS, sSR, sSA, sIV},/* Can't come in reply dir */
+/* cookie_echo */ {sIV, sCL, sCE, sCE, sES, sSS, sSR, sSA, sIV},/* Can't come in reply dir */
/* cookie_ack */ {sIV, sCL, sCW, sES, sES, sSS, sSR, sSA, sIV},
/* shutdown_comp*/ {sIV, sCL, sCW, sCE, sES, sSS, sSR, sCL, sIV},
/* heartbeat */ {sIV, sCL, sCW, sCE, sES, sSS, sSR, sSA, sHS},
@@ -412,6 +412,9 @@ int nf_conntrack_sctp_packet(struct nf_conn *ct,
/* (D) vtag must be same as init_vtag as found in INIT_ACK */
if (sh->vtag != ct->proto.sctp.vtag[dir])
goto out_unlock;
+ } else if (sch->type == SCTP_CID_COOKIE_ACK) {
+ ct->proto.sctp.init[dir] = 0;
+ ct->proto.sctp.init[!dir] = 0;
} else if (sch->type == SCTP_CID_HEARTBEAT) {
if (ct->proto.sctp.vtag[dir] == 0) {
pr_debug("Setting %d vtag %x for dir %d\n", sch->type, sh->vtag, dir);
@@ -461,16 +464,18 @@ int nf_conntrack_sctp_packet(struct nf_conn *ct,
}
/* If it is an INIT or an INIT ACK note down the vtag */
- if (sch->type == SCTP_CID_INIT ||
- sch->type == SCTP_CID_INIT_ACK) {
- struct sctp_inithdr _inithdr, *ih;
+ if (sch->type == SCTP_CID_INIT) {
+ struct sctp_inithdr _ih, *ih;
- ih = skb_header_pointer(skb, offset + sizeof(_sch),
- sizeof(_inithdr), &_inithdr);
- if (ih == NULL)
+ ih = skb_header_pointer(skb, offset + sizeof(_sch), sizeof(*ih), &_ih);
+ if (!ih)
goto out_unlock;
- pr_debug("Setting vtag %x for dir %d\n",
- ih->init_tag, !dir);
+
+ if (ct->proto.sctp.init[dir] && ct->proto.sctp.init[!dir])
+ ct->proto.sctp.init[!dir] = 0;
+ ct->proto.sctp.init[dir] = 1;
+
+ pr_debug("Setting vtag %x for dir %d\n", ih->init_tag, !dir);
ct->proto.sctp.vtag[!dir] = ih->init_tag;
/* don't renew timeout on init retransmit so
@@ -481,6 +486,24 @@ int nf_conntrack_sctp_packet(struct nf_conn *ct,
old_state == SCTP_CONNTRACK_CLOSED &&
nf_ct_is_confirmed(ct))
ignore = true;
+ } else if (sch->type == SCTP_CID_INIT_ACK) {
+ struct sctp_inithdr _ih, *ih;
+ __be32 vtag;
+
+ ih = skb_header_pointer(skb, offset + sizeof(_sch), sizeof(*ih), &_ih);
+ if (!ih)
+ goto out_unlock;
+
+ vtag = ct->proto.sctp.vtag[!dir];
+ if (!ct->proto.sctp.init[!dir] && vtag && vtag != ih->init_tag)
+ goto out_unlock;
+ /* collision */
+ if (ct->proto.sctp.init[dir] && ct->proto.sctp.init[!dir] &&
+ vtag != ih->init_tag)
+ goto out_unlock;
+
+ pr_debug("Setting vtag %x for dir %d\n", ih->init_tag, !dir);
+ ct->proto.sctp.vtag[!dir] = ih->init_tag;
}
ct->proto.sctp.state = new_state;
--
2.39.1
^ permalink raw reply related [flat|nested] 4+ messages in thread* [PATCHv2 nf 2/2] selftests: netfilter: test for sctp collision processing in nf_conntrack
2023-10-03 17:17 [PATCHv2 nf 0/2] netfilter: handle the sctp collision properly and add selftest Xin Long
2023-10-03 17:17 ` [PATCHv2 nf 1/2] netfilter: handle the connecting collision properly in nf_conntrack_proto_sctp Xin Long
@ 2023-10-03 17:17 ` Xin Long
2023-10-07 14:47 ` [PATCHv2 nf 0/2] netfilter: handle the sctp collision properly and add selftest Simon Horman
2 siblings, 0 replies; 4+ messages in thread
From: Xin Long @ 2023-10-03 17:17 UTC (permalink / raw)
To: network dev, netfilter-devel, linux-sctp
Cc: davem, kuba, Eric Dumazet, Paolo Abeni, Pablo Neira Ayuso,
Jozsef Kadlecsik, Florian Westphal, Marcelo Ricardo Leitner,
Simon Horman
This patch adds a test case to reproduce the SCTP DATA chunk retransmission
timeout issue caused by the improper SCTP collision processing in netfilter
nf_conntrack_proto_sctp.
In this test, client sends a INIT chunk, but the INIT_ACK replied from
server is delayed until the server sends a INIT chunk to start a new
connection from its side. After the connection is complete from server
side, the delayed INIT_ACK arrives in nf_conntrack_proto_sctp.
The delayed INIT_ACK should be dropped in nf_conntrack_proto_sctp instead
of updating the vtag with the out-of-date init_tag, otherwise, the vtag
in DATA chunks later sent by client don't match the vtag in the conntrack
entry and the DATA chunks get dropped.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
---
tools/testing/selftests/netfilter/Makefile | 5 +-
.../netfilter/conntrack_sctp_collision.sh | 89 +++++++++++++++++
.../selftests/netfilter/sctp_collision.c | 99 +++++++++++++++++++
3 files changed, 191 insertions(+), 2 deletions(-)
create mode 100755 tools/testing/selftests/netfilter/conntrack_sctp_collision.sh
create mode 100644 tools/testing/selftests/netfilter/sctp_collision.c
diff --git a/tools/testing/selftests/netfilter/Makefile b/tools/testing/selftests/netfilter/Makefile
index 321db8850da0..ef90aca4cc96 100644
--- a/tools/testing/selftests/netfilter/Makefile
+++ b/tools/testing/selftests/netfilter/Makefile
@@ -6,13 +6,14 @@ TEST_PROGS := nft_trans_stress.sh nft_fib.sh nft_nat.sh bridge_brouter.sh \
nft_concat_range.sh nft_conntrack_helper.sh \
nft_queue.sh nft_meta.sh nf_nat_edemux.sh \
ipip-conntrack-mtu.sh conntrack_tcp_unreplied.sh \
- conntrack_vrf.sh nft_synproxy.sh rpath.sh nft_audit.sh
+ conntrack_vrf.sh nft_synproxy.sh rpath.sh nft_audit.sh \
+ conntrack_sctp_collision.sh
HOSTPKG_CONFIG := pkg-config
CFLAGS += $(shell $(HOSTPKG_CONFIG) --cflags libmnl 2>/dev/null)
LDLIBS += $(shell $(HOSTPKG_CONFIG) --libs libmnl 2>/dev/null || echo -lmnl)
-TEST_GEN_FILES = nf-queue connect_close audit_logread
+TEST_GEN_FILES = nf-queue connect_close audit_logread sctp_collision
include ../lib.mk
diff --git a/tools/testing/selftests/netfilter/conntrack_sctp_collision.sh b/tools/testing/selftests/netfilter/conntrack_sctp_collision.sh
new file mode 100755
index 000000000000..a924e595cfd8
--- /dev/null
+++ b/tools/testing/selftests/netfilter/conntrack_sctp_collision.sh
@@ -0,0 +1,89 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+#
+# Testing For SCTP COLLISION SCENARIO as Below:
+#
+# 14:35:47.655279 IP CLIENT_IP.PORT > SERVER_IP.PORT: sctp (1) [INIT] [init tag: 2017837359]
+# 14:35:48.353250 IP SERVER_IP.PORT > CLIENT_IP.PORT: sctp (1) [INIT] [init tag: 1187206187]
+# 14:35:48.353275 IP CLIENT_IP.PORT > SERVER_IP.PORT: sctp (1) [INIT ACK] [init tag: 2017837359]
+# 14:35:48.353283 IP SERVER_IP.PORT > CLIENT_IP.PORT: sctp (1) [COOKIE ECHO]
+# 14:35:48.353977 IP CLIENT_IP.PORT > SERVER_IP.PORT: sctp (1) [COOKIE ACK]
+# 14:35:48.855335 IP SERVER_IP.PORT > CLIENT_IP.PORT: sctp (1) [INIT ACK] [init tag: 164579970]
+#
+# TOPO: SERVER_NS (link0)<--->(link1) ROUTER_NS (link2)<--->(link3) CLIENT_NS
+
+CLIENT_NS=$(mktemp -u client-XXXXXXXX)
+CLIENT_IP="198.51.200.1"
+CLIENT_PORT=1234
+
+SERVER_NS=$(mktemp -u server-XXXXXXXX)
+SERVER_IP="198.51.100.1"
+SERVER_PORT=1234
+
+ROUTER_NS=$(mktemp -u router-XXXXXXXX)
+CLIENT_GW="198.51.200.2"
+SERVER_GW="198.51.100.2"
+
+# setup the topo
+setup() {
+ ip net add $CLIENT_NS
+ ip net add $SERVER_NS
+ ip net add $ROUTER_NS
+ ip -n $SERVER_NS link add link0 type veth peer name link1 netns $ROUTER_NS
+ ip -n $CLIENT_NS link add link3 type veth peer name link2 netns $ROUTER_NS
+
+ ip -n $SERVER_NS link set link0 up
+ ip -n $SERVER_NS addr add $SERVER_IP/24 dev link0
+ ip -n $SERVER_NS route add $CLIENT_IP dev link0 via $SERVER_GW
+
+ ip -n $ROUTER_NS link set link1 up
+ ip -n $ROUTER_NS link set link2 up
+ ip -n $ROUTER_NS addr add $SERVER_GW/24 dev link1
+ ip -n $ROUTER_NS addr add $CLIENT_GW/24 dev link2
+ ip net exec $ROUTER_NS sysctl -wq net.ipv4.ip_forward=1
+
+ ip -n $CLIENT_NS link set link3 up
+ ip -n $CLIENT_NS addr add $CLIENT_IP/24 dev link3
+ ip -n $CLIENT_NS route add $SERVER_IP dev link3 via $CLIENT_GW
+
+ # simulate the delay on OVS upcall by setting up a delay for INIT_ACK with
+ # tc on $SERVER_NS side
+ tc -n $SERVER_NS qdisc add dev link0 root handle 1: htb
+ tc -n $SERVER_NS class add dev link0 parent 1: classid 1:1 htb rate 100mbit
+ tc -n $SERVER_NS filter add dev link0 parent 1: protocol ip u32 match ip protocol 132 \
+ 0xff match u8 2 0xff at 32 flowid 1:1
+ tc -n $SERVER_NS qdisc add dev link0 parent 1:1 handle 10: netem delay 1200ms
+
+ # simulate the ctstate check on OVS nf_conntrack
+ ip net exec $ROUTER_NS iptables -A FORWARD -m state --state INVALID,UNTRACKED -j DROP
+ ip net exec $ROUTER_NS iptables -A INPUT -p sctp -j DROP
+
+ # use a smaller number for assoc's max_retrans to reproduce the issue
+ modprobe sctp
+ ip net exec $CLIENT_NS sysctl -wq net.sctp.association_max_retrans=3
+}
+
+cleanup() {
+ ip net exec $CLIENT_NS pkill sctp_collision 2>&1 >/dev/null
+ ip net exec $SERVER_NS pkill sctp_collision 2>&1 >/dev/null
+ ip net del "$CLIENT_NS"
+ ip net del "$SERVER_NS"
+ ip net del "$ROUTER_NS"
+}
+
+do_test() {
+ ip net exec $SERVER_NS ./sctp_collision server \
+ $SERVER_IP $SERVER_PORT $CLIENT_IP $CLIENT_PORT &
+ ip net exec $CLIENT_NS ./sctp_collision client \
+ $CLIENT_IP $CLIENT_PORT $SERVER_IP $SERVER_PORT
+}
+
+# NOTE: one way to work around the issue is set a smaller hb_interval
+# ip net exec $CLIENT_NS sysctl -wq net.sctp.hb_interval=3500
+
+# run the test case
+trap cleanup EXIT
+setup && \
+echo "Test for SCTP Collision in nf_conntrack:" && \
+do_test && echo "PASS!"
+exit $?
diff --git a/tools/testing/selftests/netfilter/sctp_collision.c b/tools/testing/selftests/netfilter/sctp_collision.c
new file mode 100644
index 000000000000..21bb1cfd8a85
--- /dev/null
+++ b/tools/testing/selftests/netfilter/sctp_collision.c
@@ -0,0 +1,99 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <arpa/inet.h>
+
+int main(int argc, char *argv[])
+{
+ struct sockaddr_in saddr = {}, daddr = {};
+ int sd, ret, len = sizeof(daddr);
+ struct timeval tv = {25, 0};
+ char buf[] = "hello";
+
+ if (argc != 6 || (strcmp(argv[1], "server") && strcmp(argv[1], "client"))) {
+ printf("%s <server|client> <LOCAL_IP> <LOCAL_PORT> <REMOTE_IP> <REMOTE_PORT>\n",
+ argv[0]);
+ return -1;
+ }
+
+ sd = socket(AF_INET, SOCK_SEQPACKET, IPPROTO_SCTP);
+ if (sd < 0) {
+ printf("Failed to create sd\n");
+ return -1;
+ }
+
+ saddr.sin_family = AF_INET;
+ saddr.sin_addr.s_addr = inet_addr(argv[2]);
+ saddr.sin_port = htons(atoi(argv[3]));
+
+ ret = bind(sd, (struct sockaddr *)&saddr, sizeof(saddr));
+ if (ret < 0) {
+ printf("Failed to bind to address\n");
+ goto out;
+ }
+
+ ret = listen(sd, 5);
+ if (ret < 0) {
+ printf("Failed to listen on port\n");
+ goto out;
+ }
+
+ daddr.sin_family = AF_INET;
+ daddr.sin_addr.s_addr = inet_addr(argv[4]);
+ daddr.sin_port = htons(atoi(argv[5]));
+
+ /* make test shorter than 25s */
+ ret = setsockopt(sd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv));
+ if (ret < 0) {
+ printf("Failed to setsockopt SO_RCVTIMEO\n");
+ goto out;
+ }
+
+ if (!strcmp(argv[1], "server")) {
+ sleep(1); /* wait a bit for client's INIT */
+ ret = connect(sd, (struct sockaddr *)&daddr, len);
+ if (ret < 0) {
+ printf("Failed to connect to peer\n");
+ goto out;
+ }
+ ret = recvfrom(sd, buf, sizeof(buf), 0, (struct sockaddr *)&daddr, &len);
+ if (ret < 0) {
+ printf("Failed to recv msg %d\n", ret);
+ goto out;
+ }
+ ret = sendto(sd, buf, strlen(buf) + 1, 0, (struct sockaddr *)&daddr, len);
+ if (ret < 0) {
+ printf("Failed to send msg %d\n", ret);
+ goto out;
+ }
+ printf("Server: sent! %d\n", ret);
+ }
+
+ if (!strcmp(argv[1], "client")) {
+ usleep(300000); /* wait a bit for server's listening */
+ ret = connect(sd, (struct sockaddr *)&daddr, len);
+ if (ret < 0) {
+ printf("Failed to connect to peer\n");
+ goto out;
+ }
+ sleep(1); /* wait a bit for server's delayed INIT_ACK to reproduce the issue */
+ ret = sendto(sd, buf, strlen(buf) + 1, 0, (struct sockaddr *)&daddr, len);
+ if (ret < 0) {
+ printf("Failed to send msg %d\n", ret);
+ goto out;
+ }
+ ret = recvfrom(sd, buf, sizeof(buf), 0, (struct sockaddr *)&daddr, &len);
+ if (ret < 0) {
+ printf("Failed to recv msg %d\n", ret);
+ goto out;
+ }
+ printf("Client: rcvd! %d\n", ret);
+ }
+ ret = 0;
+out:
+ close(sd);
+ return ret;
+}
--
2.39.1
^ permalink raw reply related [flat|nested] 4+ messages in thread