1#!/bin/bash 2# SPDX-License-Identifier: GPL-2.0 3# 4# This tests basic flowtable functionality. 5# Creates following default topology: 6# 7# Originator (MTU 9000) <-Router1-> MTU 1500 <-Router2-> Responder (MTU 2000) 8# Router1 is the one doing flow offloading, Router2 has no special 9# purpose other than having a link that is smaller than either Originator 10# and responder, i.e. TCPMSS announced values are too large and will still 11# result in fragmentation and/or PMTU discovery. 12# 13# You can check with different Orgininator/Link/Responder MTU eg: 14# nft_flowtable.sh -o8000 -l1500 -r2000 15# 16 17 18# Kselftest framework requirement - SKIP code is 4. 19ksft_skip=4 20ret=0 21 22ns1in="" 23ns2in="" 24ns1out="" 25ns2out="" 26 27log_netns=$(sysctl -n net.netfilter.nf_log_all_netns) 28 29checktool (){ 30 if ! $1 > /dev/null 2>&1; then 31 echo "SKIP: Could not $2" 32 exit $ksft_skip 33 fi 34} 35 36checktool "nft --version" "run test without nft tool" 37checktool "ip -Version" "run test without ip tool" 38checktool "which nc" "run test without nc (netcat)" 39checktool "ip netns add nsr1" "create net namespace" 40 41ip netns add ns1 42ip netns add ns2 43 44ip netns add nsr2 45 46cleanup() { 47 for i in 1 2; do 48 ip netns del ns$i 49 ip netns del nsr$i 50 done 51 52 rm -f "$ns1in" "$ns1out" 53 rm -f "$ns2in" "$ns2out" 54 55 [ $log_netns -eq 0 ] && sysctl -q net.netfilter.nf_log_all_netns=$log_netns 56} 57 58trap cleanup EXIT 59 60sysctl -q net.netfilter.nf_log_all_netns=1 61 62ip link add veth0 netns nsr1 type veth peer name eth0 netns ns1 63ip link add veth1 netns nsr1 type veth peer name veth0 netns nsr2 64 65ip link add veth1 netns nsr2 type veth peer name eth0 netns ns2 66 67for dev in lo veth0 veth1; do 68 for i in 1 2; do 69 ip -net nsr$i link set $dev up 70 done 71done 72 73ip -net nsr1 addr add 10.0.1.1/24 dev veth0 74ip -net nsr1 addr add dead:1::1/64 dev veth0 75 76ip -net nsr2 addr add 10.0.2.1/24 dev veth1 77ip -net nsr2 addr add dead:2::1/64 dev veth1 78 79# set different MTUs so we need to push packets coming from ns1 (large MTU) 80# to ns2 (smaller MTU) to stack either to perform fragmentation (ip_no_pmtu_disc=1), 81# or to do PTMU discovery (send ICMP error back to originator). 82# ns2 is going via nsr2 with a smaller mtu, so that TCPMSS announced by both peers 83# is NOT the lowest link mtu. 84 85omtu=9000 86lmtu=1500 87rmtu=2000 88 89usage(){ 90 echo "nft_flowtable.sh [OPTIONS]" 91 echo 92 echo "MTU options" 93 echo " -o originator" 94 echo " -l link" 95 echo " -r responder" 96 exit 1 97} 98 99while getopts "o:l:r:" o 100do 101 case $o in 102 o) omtu=$OPTARG;; 103 l) lmtu=$OPTARG;; 104 r) rmtu=$OPTARG;; 105 *) usage;; 106 esac 107done 108 109if ! ip -net nsr1 link set veth0 mtu $omtu; then 110 exit 1 111fi 112 113ip -net ns1 link set eth0 mtu $omtu 114 115if ! ip -net nsr2 link set veth1 mtu $rmtu; then 116 exit 1 117fi 118 119ip -net ns2 link set eth0 mtu $rmtu 120 121# transfer-net between nsr1 and nsr2. 122# these addresses are not used for connections. 123ip -net nsr1 addr add 192.168.10.1/24 dev veth1 124ip -net nsr1 addr add fee1:2::1/64 dev veth1 125 126ip -net nsr2 addr add 192.168.10.2/24 dev veth0 127ip -net nsr2 addr add fee1:2::2/64 dev veth0 128 129for i in 1 2; do 130 ip netns exec nsr$i sysctl net.ipv4.conf.veth0.forwarding=1 > /dev/null 131 ip netns exec nsr$i sysctl net.ipv4.conf.veth1.forwarding=1 > /dev/null 132 133 ip -net ns$i link set lo up 134 ip -net ns$i link set eth0 up 135 ip -net ns$i addr add 10.0.$i.99/24 dev eth0 136 ip -net ns$i route add default via 10.0.$i.1 137 ip -net ns$i addr add dead:$i::99/64 dev eth0 138 ip -net ns$i route add default via dead:$i::1 139 if ! ip netns exec ns$i sysctl net.ipv4.tcp_no_metrics_save=1 > /dev/null; then 140 echo "ERROR: Check Originator/Responder values (problem during address addition)" 141 exit 1 142 fi 143 144 # don't set ip DF bit for first two tests 145 ip netns exec ns$i sysctl net.ipv4.ip_no_pmtu_disc=1 > /dev/null 146done 147 148ip -net nsr1 route add default via 192.168.10.2 149ip -net nsr2 route add default via 192.168.10.1 150 151ip netns exec nsr1 nft -f - <<EOF 152table inet filter { 153 flowtable f1 { 154 hook ingress priority 0 155 devices = { veth0, veth1 } 156 } 157 158 chain forward { 159 type filter hook forward priority 0; policy drop; 160 161 # flow offloaded? Tag ct with mark 1, so we can detect when it fails. 162 meta oif "veth1" tcp dport 12345 flow offload @f1 counter 163 164 # use packet size to trigger 'should be offloaded by now'. 165 # otherwise, if 'flow offload' expression never offloads, the 166 # test will pass. 167 tcp dport 12345 meta length gt 200 ct mark set 1 counter 168 169 # this turns off flow offloading internally, so expect packets again 170 tcp flags fin,rst ct mark set 0 accept 171 172 # this allows large packets from responder, we need this as long 173 # as PMTUd is off. 174 # This rule is deleted for the last test, when we expect PMTUd 175 # to kick in and ensure all packets meet mtu requirements. 176 meta length gt $lmtu accept comment something-to-grep-for 177 178 # next line blocks connection w.o. working offload. 179 # we only do this for reverse dir, because we expect packets to 180 # enter slow path due to MTU mismatch of veth0 and veth1. 181 tcp sport 12345 ct mark 1 counter log prefix "mark failure " drop 182 183 ct state established,related accept 184 185 # for packets that we can't offload yet, i.e. SYN (any ct that is not confirmed) 186 meta length lt 200 oif "veth1" tcp dport 12345 counter accept 187 188 meta nfproto ipv4 meta l4proto icmp accept 189 meta nfproto ipv6 meta l4proto icmpv6 accept 190 } 191} 192EOF 193 194if [ $? -ne 0 ]; then 195 echo "SKIP: Could not load nft ruleset" 196 exit $ksft_skip 197fi 198 199# test basic connectivity 200if ! ip netns exec ns1 ping -c 1 -q 10.0.2.99 > /dev/null; then 201 echo "ERROR: ns1 cannot reach ns2" 1>&2 202 exit 1 203fi 204 205if ! ip netns exec ns2 ping -c 1 -q 10.0.1.99 > /dev/null; then 206 echo "ERROR: ns2 cannot reach ns1" 1>&2 207 exit 1 208fi 209 210if [ $ret -eq 0 ];then 211 echo "PASS: netns routing/connectivity: ns1 can reach ns2" 212fi 213 214ns1in=$(mktemp) 215ns1out=$(mktemp) 216ns2in=$(mktemp) 217ns2out=$(mktemp) 218 219make_file() 220{ 221 name=$1 222 223 SIZE=$((RANDOM % (1024 * 8))) 224 TSIZE=$((SIZE * 1024)) 225 226 dd if=/dev/urandom of="$name" bs=1024 count=$SIZE 2> /dev/null 227 228 SIZE=$((RANDOM % 1024)) 229 SIZE=$((SIZE + 128)) 230 TSIZE=$((TSIZE + SIZE)) 231 dd if=/dev/urandom conf=notrunc of="$name" bs=1 count=$SIZE 2> /dev/null 232} 233 234check_transfer() 235{ 236 in=$1 237 out=$2 238 what=$3 239 240 if ! cmp "$in" "$out" > /dev/null 2>&1; then 241 echo "FAIL: file mismatch for $what" 1>&2 242 ls -l "$in" 243 ls -l "$out" 244 return 1 245 fi 246 247 return 0 248} 249 250test_tcp_forwarding_ip() 251{ 252 local nsa=$1 253 local nsb=$2 254 local dstip=$3 255 local dstport=$4 256 local lret=0 257 258 ip netns exec $nsb nc -w 5 -l -p 12345 < "$ns2in" > "$ns2out" & 259 lpid=$! 260 261 sleep 1 262 ip netns exec $nsa nc -w 4 "$dstip" "$dstport" < "$ns1in" > "$ns1out" & 263 cpid=$! 264 265 sleep 3 266 267 if ps -p $lpid > /dev/null;then 268 kill $lpid 269 fi 270 271 if ps -p $cpid > /dev/null;then 272 kill $cpid 273 fi 274 275 wait 276 277 if ! check_transfer "$ns1in" "$ns2out" "ns1 -> ns2"; then 278 lret=1 279 fi 280 281 if ! check_transfer "$ns2in" "$ns1out" "ns1 <- ns2"; then 282 lret=1 283 fi 284 285 return $lret 286} 287 288test_tcp_forwarding() 289{ 290 test_tcp_forwarding_ip "$1" "$2" 10.0.2.99 12345 291 292 return $? 293} 294 295test_tcp_forwarding_nat() 296{ 297 local lret 298 299 test_tcp_forwarding_ip "$1" "$2" 10.0.2.99 12345 300 lret=$? 301 302 if [ $lret -eq 0 ] ; then 303 test_tcp_forwarding_ip "$1" "$2" 10.6.6.6 1666 304 lret=$? 305 fi 306 307 return $lret 308} 309 310make_file "$ns1in" 311make_file "$ns2in" 312 313# First test: 314# No PMTU discovery, nsr1 is expected to fragment packets from ns1 to ns2 as needed. 315if test_tcp_forwarding ns1 ns2; then 316 echo "PASS: flow offloaded for ns1/ns2" 317else 318 echo "FAIL: flow offload for ns1/ns2:" 1>&2 319 ip netns exec nsr1 nft list ruleset 320 ret=1 321fi 322 323# delete default route, i.e. ns2 won't be able to reach ns1 and 324# will depend on ns1 being masqueraded in nsr1. 325# expect ns1 has nsr1 address. 326ip -net ns2 route del default via 10.0.2.1 327ip -net ns2 route del default via dead:2::1 328ip -net ns2 route add 192.168.10.1 via 10.0.2.1 329 330# Second test: 331# Same, but with NAT enabled. 332ip netns exec nsr1 nft -f - <<EOF 333table ip nat { 334 chain prerouting { 335 type nat hook prerouting priority 0; policy accept; 336 meta iif "veth0" ip daddr 10.6.6.6 tcp dport 1666 counter dnat ip to 10.0.2.99:12345 337 } 338 339 chain postrouting { 340 type nat hook postrouting priority 0; policy accept; 341 meta oifname "veth1" counter masquerade 342 } 343} 344EOF 345 346if test_tcp_forwarding_nat ns1 ns2; then 347 echo "PASS: flow offloaded for ns1/ns2 with NAT" 348else 349 echo "FAIL: flow offload for ns1/ns2 with NAT" 1>&2 350 ip netns exec nsr1 nft list ruleset 351 ret=1 352fi 353 354# Third test: 355# Same as second test, but with PMTU discovery enabled. 356handle=$(ip netns exec nsr1 nft -a list table inet filter | grep something-to-grep-for | cut -d \# -f 2) 357 358if ! ip netns exec nsr1 nft delete rule inet filter forward $handle; then 359 echo "FAIL: Could not delete large-packet accept rule" 360 exit 1 361fi 362 363ip netns exec ns1 sysctl net.ipv4.ip_no_pmtu_disc=0 > /dev/null 364ip netns exec ns2 sysctl net.ipv4.ip_no_pmtu_disc=0 > /dev/null 365 366if test_tcp_forwarding_nat ns1 ns2; then 367 echo "PASS: flow offloaded for ns1/ns2 with NAT and pmtu discovery" 368else 369 echo "FAIL: flow offload for ns1/ns2 with NAT and pmtu discovery" 1>&2 370 ip netns exec nsr1 nft list ruleset 371fi 372 373KEY_SHA="0x"$(ps -xaf | sha1sum | cut -d " " -f 1) 374KEY_AES="0x"$(ps -xaf | md5sum | cut -d " " -f 1) 375SPI1=$RANDOM 376SPI2=$RANDOM 377 378if [ $SPI1 -eq $SPI2 ]; then 379 SPI2=$((SPI2+1)) 380fi 381 382do_esp() { 383 local ns=$1 384 local me=$2 385 local remote=$3 386 local lnet=$4 387 local rnet=$5 388 local spi_out=$6 389 local spi_in=$7 390 391 ip -net $ns xfrm state add src $remote dst $me proto esp spi $spi_in enc aes $KEY_AES auth sha1 $KEY_SHA mode tunnel sel src $rnet dst $lnet 392 ip -net $ns xfrm state add src $me dst $remote proto esp spi $spi_out enc aes $KEY_AES auth sha1 $KEY_SHA mode tunnel sel src $lnet dst $rnet 393 394 # to encrypt packets as they go out (includes forwarded packets that need encapsulation) 395 ip -net $ns xfrm policy add src $lnet dst $rnet dir out tmpl src $me dst $remote proto esp mode tunnel priority 1 action allow 396 # to fwd decrypted packets after esp processing: 397 ip -net $ns xfrm policy add src $rnet dst $lnet dir fwd tmpl src $remote dst $me proto esp mode tunnel priority 1 action allow 398 399} 400 401do_esp nsr1 192.168.10.1 192.168.10.2 10.0.1.0/24 10.0.2.0/24 $SPI1 $SPI2 402 403do_esp nsr2 192.168.10.2 192.168.10.1 10.0.2.0/24 10.0.1.0/24 $SPI2 $SPI1 404 405ip netns exec nsr1 nft delete table ip nat 406 407# restore default routes 408ip -net ns2 route del 192.168.10.1 via 10.0.2.1 409ip -net ns2 route add default via 10.0.2.1 410ip -net ns2 route add default via dead:2::1 411 412if test_tcp_forwarding ns1 ns2; then 413 echo "PASS: ipsec tunnel mode for ns1/ns2" 414else 415 echo "FAIL: ipsec tunnel mode for ns1/ns2" 416 ip netns exec nsr1 nft list ruleset 1>&2 417 ip netns exec nsr1 cat /proc/net/xfrm_stat 1>&2 418fi 419 420exit $ret 421