Quelle  mptcp_connect.sh   Sprache: Shell

#!/bin/bash
# SPDX-License-Identifier: GPL-2.0

# Double quotes to prevent globbing and word splitting is recommended in new
# code but we accept it, especially because there were too many before having
# address all other issues detected by shellcheck.
#shellcheck disable=SC2086

. "$(dirname "${0}")/mptcp_lib.sh"

time_start=$(date +%s)

optstring="S:R:d:e:l:r:h4cm:f:tC"
ret=0
final_ret=0
sin=""
sout=""
cin_disconnect=""
cin=""
cout=""
capture=false
timeout_poll=30
timeout_test=$((timeout_poll * 2 + 1))
ipv6=true
ethtool_random_on=true
tc_delay="$((RANDOM%50))"
tc_loss=$((RANDOM%101))
testmode=""
sndbuf=0
rcvbuf=0
options_log=true
do_tcp=0
checksum=false
filesize=0
connect_per_transfer=1
port=$((10000 - 1))

if [ $tc_loss -eq 100 ];then
 tc_loss=1%
elif [ $tc_loss -ge 10 ]; then
 tc_loss=0.$tc_loss%
elif [ $tc_loss -ge 1 ]; then
 tc_loss=0.0$tc_loss%
else
 tc_loss=""
fi

usage() {
 echo "Usage: $0 [ -a ]"
 echo -e "\t-d: tc/netem delay in milliseconds, e.g. \"-d 10\" (default random)"
 echo -e "\t-l: tc/netem loss percentage, e.g. \"-l 0.02\" (default random)"
 echo -e "\t-r: tc/netem reorder mode, e.g. \"-r 25% 50% gap 5\", use "-r 0" to disable reordering (default random)"
 echo -e "\t-e: ethtool features to disable, e.g.: \"-e tso -e gso\" (default: randomly disable any of tso/gso/gro)"
 echo -e "\t-4: IPv4 only: disable IPv6 tests (default: test both IPv4 and IPv6)"
 echo -e "\t-c: capture packets for each test using tcpdump (default: no capture)"
 echo -e "\t-f: size of file to transfer in bytes (default random)"
 echo -e "\t-S: set sndbuf value (default: use kernel default)"
 echo -e "\t-R: set rcvbuf value (default: use kernel default)"
 echo -e "\t-m: test mode (poll, sendfile; default: poll)"
 echo -e "\t-t: also run tests with TCP (use twice to non-fallback tcp)"
 echo -e "\t-C: enable the MPTCP data checksum"
}

while getopts "$optstring" option;do
 case "$option" in
 "h")
  usage $0
  exit ${KSFT_PASS}
  ;;
 "d")
  if [ $OPTARG -ge 0 ];then
   tc_delay="$OPTARG"
  else
   echo "-d requires numeric argument, got \"$OPTARG\"" 1>&2
   exit ${KSFT_FAIL}
  fi
  ;;
 "e")
  ethtool_args="$ethtool_args $OPTARG off"
  ethtool_random_on=false
  ;;
 "l")
  tc_loss="$OPTARG"
  ;;
 "r")
  tc_reorder="$OPTARG"
  ;;
 "4")
  ipv6=false
  ;;
 "c")
  capture=true
  ;;
 "S")
  if [ $OPTARG -ge 0 ];then
   sndbuf="$OPTARG"
  else
   echo "-S requires numeric argument, got \"$OPTARG\"" 1>&2
   exit ${KSFT_FAIL}
  fi
  ;;
 "R")
  if [ $OPTARG -ge 0 ];then
   rcvbuf="$OPTARG"
  else
   echo "-R requires numeric argument, got \"$OPTARG\"" 1>&2
   exit ${KSFT_FAIL}
  fi
  ;;
 "m")
  testmode="$OPTARG"
  ;;
 "f")
  filesize="$OPTARG"
  ;;
 "t")
  do_tcp=$((do_tcp+1))
  ;;
 "C")
  checksum=true
  ;;
 "?")
  usage $0
  exit ${KSFT_FAIL}
  ;;
 esac
done

ns1=""
ns2=""
ns3=""
ns4=""

TEST_GROUP=""

# This function is used in the cleanup trap
#shellcheck disable=SC2317,SC2329
cleanup()
{
 rm -f "$cin_disconnect"
 rm -f "$cin" "$cout"
 rm -f "$sin" "$sout"
 rm -f "$capout"

 mptcp_lib_ns_exit "${ns1}" "${ns2}" "${ns3}" "${ns4}"
}

mptcp_lib_check_mptcp
mptcp_lib_check_kallsyms
mptcp_lib_check_tools ip tc

sin=$(mktemp)
sout=$(mktemp)
cin=$(mktemp)
cout=$(mktemp)
capout=$(mktemp)
cin_disconnect="$cin".disconnect
trap cleanup EXIT

mptcp_lib_ns_init ns1 ns2 ns3 ns4

#  "$ns1"              ns2                    ns3                     ns4
# ns1eth2    ns2eth1   ns2eth3      ns3eth2   ns3eth4       ns4eth3
#                           - drop 1% ->            reorder 25%
#                           <- TSO off -

ip link add ns1eth2 netns "$ns1" type veth peer name ns2eth1 netns "$ns2"
ip link add ns2eth3 netns "$ns2" type veth peer name ns3eth2 netns "$ns3"
ip link add ns3eth4 netns "$ns3" type veth peer name ns4eth3 netns "$ns4"

ip -net "$ns1" addr add 10.0.1.1/24 dev ns1eth2
ip -net "$ns1" addr add dead:beef:1::1/64 dev ns1eth2 nodad

ip -net "$ns1" link set ns1eth2 up
ip -net "$ns1" route add default via 10.0.1.2
ip -net "$ns1" route add default via dead:beef:1::2

ip -net "$ns2" addr add 10.0.1.2/24 dev ns2eth1
ip -net "$ns2" addr add dead:beef:1::2/64 dev ns2eth1 nodad
ip -net "$ns2" link set ns2eth1 up

ip -net "$ns2" addr add 10.0.2.1/24 dev ns2eth3
ip -net "$ns2" addr add dead:beef:2::1/64 dev ns2eth3 nodad
ip -net "$ns2" link set ns2eth3 up
ip -net "$ns2" route add default via 10.0.2.2
ip -net "$ns2" route add default via dead:beef:2::2
ip netns exec "$ns2" sysctl -q net.ipv4.ip_forward=1
ip netns exec "$ns2" sysctl -q net.ipv6.conf.all.forwarding=1

ip -net "$ns3" addr add 10.0.2.2/24 dev ns3eth2
ip -net "$ns3" addr add dead:beef:2::2/64 dev ns3eth2 nodad
ip -net "$ns3" link set ns3eth2 up

ip -net "$ns3" addr add 10.0.3.2/24 dev ns3eth4
ip -net "$ns3" addr add dead:beef:3::2/64 dev ns3eth4 nodad
ip -net "$ns3" link set ns3eth4 up
ip -net "$ns3" route add default via 10.0.2.1
ip -net "$ns3" route add default via dead:beef:2::1
ip netns exec "$ns3" sysctl -q net.ipv4.ip_forward=1
ip netns exec "$ns3" sysctl -q net.ipv6.conf.all.forwarding=1

ip -net "$ns4" addr add 10.0.3.1/24 dev ns4eth3
ip -net "$ns4" addr add dead:beef:3::1/64 dev ns4eth3 nodad
ip -net "$ns4" link set ns4eth3 up
ip -net "$ns4" route add default via 10.0.3.2
ip -net "$ns4" route add default via dead:beef:3::2

if $checksum; then
 for i in "$ns1" "$ns2" "$ns3" "$ns4";do
  ip netns exec $i sysctl -q net.mptcp.checksum_enabled=1
 done
fi

if $capture; then
 rndh="${ns1:4}"
 mptcp_lib_pr_info "Packet capture files will have this prefix: ${rndh}-"
fi

set_ethtool_flags() {
 local ns="$1"
 local dev="$2"
 local flags="$3"

 if ip netns exec $ns ethtool -K $dev $flags 2>/dev/null; then
  mptcp_lib_pr_info "set $ns dev $dev: ethtool -K $flags"
 fi
}

set_random_ethtool_flags() {
 local flags=""
 local r=$RANDOM

 local pick1=$((r & 1))
 local pick2=$((r & 2))
 local pick3=$((r & 4))

 [ $pick1 -ne 0 ] && flags="tso off"
 [ $pick2 -ne 0 ] && flags="$flags gso off"
 [ $pick3 -ne 0 ] && flags="$flags gro off"

 [ -z "$flags" ] && return

 set_ethtool_flags "$1" "$2" "$flags"
}

if $ethtool_random_on;then
 set_random_ethtool_flags "$ns3" ns3eth2
 set_random_ethtool_flags "$ns4" ns4eth3
else
 set_ethtool_flags "$ns3" ns3eth2 "$ethtool_args"
 set_ethtool_flags "$ns4" ns4eth3 "$ethtool_args"
fi

print_larger_title() {
 # here we don't have the time, a bit longer for the alignment
 MPTCP_LIB_TEST_FORMAT="%02u %-69s" \
  mptcp_lib_print_title "${@}"
}

check_mptcp_disabled()
{
 local disabled_ns
 mptcp_lib_ns_init disabled_ns

 print_larger_title "New MPTCP socket can be blocked via sysctl"

 # mainly to cover more code
 if ! ip netns exec ${disabled_ns} sysctl net.mptcp >/dev/null; then
  mptcp_lib_pr_fail "not able to list net.mptcp sysctl knobs"
  mptcp_lib_result_fail "not able to list net.mptcp sysctl knobs"
  ret=${KSFT_FAIL}
  return 1
 fi

 # net.mptcp.enabled should be enabled by default
 if [ "$(ip netns exec ${disabled_ns} sysctl net.mptcp.enabled | awk '{ print $3 }')" -ne 1 ]; then
  mptcp_lib_pr_fail "net.mptcp.enabled sysctl is not 1 by default"
  mptcp_lib_result_fail "net.mptcp.enabled sysctl is not 1 by default"
  ret=${KSFT_FAIL}
  return 1
 fi
 ip netns exec ${disabled_ns} sysctl -q net.mptcp.enabled=0

 local err=0
 LC_ALL=C ip netns exec ${disabled_ns} ./mptcp_connect -p 10000 -s MPTCP 127.0.0.1 < "$cin" 2>&1&nbsp;| \
  grep -q "^socket: Protocol not available$" && err=1
 mptcp_lib_ns_exit "${disabled_ns}"

 if [ ${err} -eq 0 ]; then
  mptcp_lib_pr_fail "New MPTCP socket cannot be blocked via sysctl"
  mptcp_lib_result_fail "New MPTCP socket cannot be blocked via sysctl"
  ret=${KSFT_FAIL}
  return 1
 fi

 mptcp_lib_pr_ok
 mptcp_lib_result_pass "New MPTCP socket can be blocked via sysctl"
 return 0
}

do_ping()
{
 local listener_ns="$1"
 local connector_ns="$2"
 local connect_addr="$3"
 local ping_args="-q -c 1"
 local rc=0

 if mptcp_lib_is_v6 "${connect_addr}"; then
  $ipv6 || return 0
  ping_args="${ping_args} -6"
 fi

 ip netns exec ${connector_ns} ping ${ping_args} $connect_addr >/dev/null || rc=1

 if [ $rc -ne 0 ] ; then
  mptcp_lib_pr_fail "$listener_ns -> $connect_addr connectivity"
  ret=${KSFT_FAIL}

  return 1
 fi

 return 0
}

do_transfer()
{
 local listener_ns="$1"
 local connector_ns="$2"
 local cl_proto="$3"
 local srv_proto="$4"
 local connect_addr="$5"
 local local_addr="$6"
 local extra_args="$7"

 port=$((port + 1))

 if [ "$rcvbuf" -gt 0 ]; then
  extra_args+=" -R $rcvbuf"
 fi

 if [ "$sndbuf" -gt 0 ]; then
  extra_args+=" -S $sndbuf"
 fi

 if [ -n "$testmode" ]; then
  extra_args+=" -m $testmode"
 fi

 if [ -n "$extra_args" ] && $options_log; then
  mptcp_lib_pr_info "extra options: $extra_args"
 fi
 options_log=false

 :> "$cout"
 :> "$sout"
 :> "$capout"

 local addr_port
 addr_port=$(printf "%s:%d" ${connect_addr} ${port})
 local pretty_title
 pretty_title="$(printf "%.3s %-5s -> %.3s (%-20s) %-5s" ${connector_ns} ${cl_proto} ${listener_ns} ${addr_port} ${srv_proto})"
 mptcp_lib_print_title "${pretty_title}"

 local tap_title="${connector_ns:0:3} ${cl_proto} -> ${listener_ns:0:3} (${addr_port}) ${srv_proto}"

 if $capture; then
  local capuser
  if [ -z $SUDO_USER ] ; then
   capuser=""
  else
   capuser="-Z $SUDO_USER"
  fi

  local capfile="${rndh}-${connector_ns:0:3}-${listener_ns:0:3}-${cl_proto}-${srv_proto}-${connect_addr}-${port}"
  local capopt="-i any -s 65535 -B 32768 ${capuser}"

  ip netns exec ${listener_ns}  tcpdump ${capopt} -w "${capfile}-listener.pcap"  >> "${capout}" 2>&1 &
  local cappid_listener=$!

  ip netns exec ${connector_ns} tcpdump ${capopt} -w "${capfile}-connector.pcap" >> "${capout}" 2>&1 &
  local cappid_connector=$!

  sleep 1
 fi

 NSTAT_HISTORY=/tmp/${listener_ns}.nstat ip netns exec ${listener_ns} \
  nstat -n
 if [ ${listener_ns} != ${connector_ns} ]; then
  NSTAT_HISTORY=/tmp/${connector_ns}.nstat ip netns exec ${connector_ns} \
   nstat -n
 fi

 local stat_synrx_last_l
 local stat_ackrx_last_l
 local stat_cookietx_last
 local stat_cookierx_last
 local stat_csum_err_s
 local stat_csum_err_c
 local stat_tcpfb_last_l
 stat_synrx_last_l=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtMPCapableSYNRX")
 stat_ackrx_last_l=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtMPCapableACKRX")
 stat_cookietx_last=$(mptcp_lib_get_counter "${listener_ns}" "TcpExtSyncookiesSent")
 stat_cookierx_last=$(mptcp_lib_get_counter "${listener_ns}" "TcpExtSyncookiesRecv")
 stat_csum_err_s=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtDataCsumErr")
 stat_csum_err_c=$(mptcp_lib_get_counter "${connector_ns}" "MPTcpExtDataCsumErr")
 stat_tcpfb_last_l=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtMPCapableFallbackACK")

 timeout ${timeout_test} \
  ip netns exec ${listener_ns} \
   ./mptcp_connect -t ${timeout_poll} -l -p $port -s ${srv_proto} \
    $extra_args $local_addr < "$sin" > "$sout" &
 local spid=$!

 mptcp_lib_wait_local_port_listen "${listener_ns}" "${port}"

 local start
 start=$(date +%s%3N)
 timeout ${timeout_test} \
  ip netns exec ${connector_ns} \
   ./mptcp_connect -t ${timeout_poll} -p $port -s ${cl_proto} \
    $extra_args $connect_addr < "$cin" > "$cout" &
 local cpid=$!

 wait $cpid
 local retc=$?
 wait $spid
 local rets=$?

 local stop
 stop=$(date +%s%3N)

 if $capture; then
  sleep 1
  kill ${cappid_listener}
  kill ${cappid_connector}
 fi

 NSTAT_HISTORY=/tmp/${listener_ns}.nstat ip netns exec ${listener_ns} \
  nstat | grep Tcp > /tmp/${listener_ns}.out
 if [ ${listener_ns} != ${connector_ns} ]; then
  NSTAT_HISTORY=/tmp/${connector_ns}.nstat ip netns exec ${connector_ns} \
   nstat | grep Tcp > /tmp/${connector_ns}.out
 fi

 local duration
 duration=$((stop-start))
 printf "(duration %05sms) " "${duration}"
 if [ ${rets} -ne 0 ] || [ ${retc} -ne 0 ]; then
  mptcp_lib_pr_fail "client exit code $retc, server $rets"
  mptcp_lib_pr_err_stats "${listener_ns}" "${connector_ns}" "${port}" \
   "/tmp/${listener_ns}.out" "/tmp/${connector_ns}.out"

  echo
  cat "$capout"
  mptcp_lib_result_fail "${TEST_GROUP}: ${tap_title}"
  return 1
 fi

 mptcp_lib_check_transfer $sin $cout "file received by client"
 retc=$?
 mptcp_lib_check_transfer $cin $sout "file received by server"
 rets=$?

 local extra=""
 local stat_synrx_now_l
 local stat_ackrx_now_l
 local stat_cookietx_now
 local stat_cookierx_now
 local stat_ooo_now
 local stat_tcpfb_now_l
 stat_synrx_now_l=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtMPCapableSYNRX")
 stat_ackrx_now_l=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtMPCapableACKRX")
 stat_cookietx_now=$(mptcp_lib_get_counter "${listener_ns}" "TcpExtSyncookiesSent")
 stat_cookierx_now=$(mptcp_lib_get_counter "${listener_ns}" "TcpExtSyncookiesRecv")
 stat_ooo_now=$(mptcp_lib_get_counter "${listener_ns}" "TcpExtTCPOFOQueue")
 stat_tcpfb_now_l=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtMPCapableFallbackACK")

 expect_synrx=$((stat_synrx_last_l))
 expect_ackrx=$((stat_ackrx_last_l))

 cookies=$(ip netns exec ${listener_ns} sysctl net.ipv4.tcp_syncookies)
 cookies=${cookies##*=}

 if [ ${cl_proto} = "MPTCP" ] && [ ${srv_proto} = "MPTCP" ]; then
  expect_synrx=$((stat_synrx_last_l+connect_per_transfer))
  expect_ackrx=$((stat_ackrx_last_l+connect_per_transfer))
 fi

 if [ ${stat_synrx_now_l} -lt ${expect_synrx} ]; then
  mptcp_lib_pr_fail "lower MPC SYN rx (${stat_synrx_now_l})" \
      "than expected (${expect_synrx})"
  retc=1
 fi
 if [ ${stat_ackrx_now_l} -lt ${expect_ackrx} ]; then
  if [ ${stat_ooo_now} -eq 0 ]; then
   mptcp_lib_pr_fail "lower MPC ACK rx (${stat_ackrx_now_l})" \
       "than expected (${expect_ackrx})"
   rets=1
  else
   extra+=" [ Note ] fallback due to TCP OoO"
  fi
 fi

 if $checksum; then
  local csum_err_s
  local csum_err_c
  csum_err_s=$(mptcp_lib_get_counter "${listener_ns}" "MPTcpExtDataCsumErr")
  csum_err_c=$(mptcp_lib_get_counter "${connector_ns}" "MPTcpExtDataCsumErr")

  local csum_err_s_nr=$((csum_err_s - stat_csum_err_s))
  if [ $csum_err_s_nr -gt 0 ]; then
   mptcp_lib_pr_fail "server got ${csum_err_s_nr} data checksum error[s]"
   rets=1
  fi

  local csum_err_c_nr=$((csum_err_c - stat_csum_err_c))
  if [ $csum_err_c_nr -gt 0 ]; then
   mptcp_lib_pr_fail "client got ${csum_err_c_nr} data checksum error[s]"
   retc=1
  fi
 fi

 if [ ${stat_ooo_now} -eq 0 ] && [ ${stat_tcpfb_last_l} -ne ${stat_tcpfb_now_l} ]; then
  mptcp_lib_pr_fail "unexpected fallback to TCP"
  rets=1
 fi

 if [ $cookies -eq 2 ];then
  if [ $stat_cookietx_last -ge $stat_cookietx_now ] ;then
   extra+=" WARN: CookieSent: did not advance"
  fi
  if [ $stat_cookierx_last -ge $stat_cookierx_now ] ;then
   extra+=" WARN: CookieRecv: did not advance"
  fi
 else
  if [ $stat_cookietx_last -ne $stat_cookietx_now ] ;then
   extra+=" WARN: CookieSent: changed"
  fi
  if [ $stat_cookierx_last -ne $stat_cookierx_now ] ;then
   extra+=" WARN: CookieRecv: changed"
  fi
 fi

 if [ ${stat_synrx_now_l} -gt ${expect_synrx} ]; then
  extra+=" WARN: SYNRX: expect ${expect_synrx},"
  extra+=" got ${stat_synrx_now_l} (probably retransmissions)"
 fi
 if [ ${stat_ackrx_now_l} -gt ${expect_ackrx} ]; then
  extra+=" WARN: ACKRX: expect ${expect_ackrx},"
  extra+=" got ${stat_ackrx_now_l} (probably retransmissions)"
 fi

 if [ $retc -eq 0 ] && [ $rets -eq 0 ]; then
  mptcp_lib_pr_ok "${extra:1}"
  mptcp_lib_result_pass "${TEST_GROUP}: ${tap_title}"
 else
  if [ -n "${extra}" ]; then
   mptcp_lib_print_warn "${extra:1}"
  fi
  mptcp_lib_result_fail "${TEST_GROUP}: ${tap_title}"
 fi

 cat "$capout"
 [ $retc -eq 0 ] && [ $rets -eq 0 ]
}

make_file()
{
 local name=$1
 local who=$2
 local SIZE=$filesize
 local ksize
 local rem

 if [ $SIZE -eq 0 ]; then
  local MAXSIZE=$((1024 * 1024 * 8))
  local MINSIZE=$((1024 * 256))

  SIZE=$(((RANDOM * RANDOM + MINSIZE) % MAXSIZE))
 fi

 ksize=$((SIZE / 1024))
 rem=$((SIZE - (ksize * 1024)))

 mptcp_lib_make_file $name 1024 $ksize
 dd if=/dev/urandom conv=notrunc of="$name" oflag=append bs=1 count=$rem 2> /dev/null

 echo "Created $name (size $(stat -c "%s" "$name") B) containing data sent by $who"
}

run_tests_lo()
{
 local listener_ns="$1"
 local connector_ns="$2"
 local connect_addr="$3"
 local loopback="$4"
 local extra_args="$5"
 local lret=0

 # skip if test programs are running inside same netns for subsequent runs.
 if [ $loopback -eq 0 ] && [ ${listener_ns} = ${connector_ns} ]; then
  return 0
 fi

 # skip if we don't want v6
 if ! $ipv6 && mptcp_lib_is_v6 "${connect_addr}"; then
  return 0
 fi

 local local_addr
 if mptcp_lib_is_v6 "${connect_addr}"; then
  local_addr="::"
 else
  local_addr="0.0.0.0"
 fi

 do_transfer ${listener_ns} ${connector_ns} MPTCP MPTCP \
      ${connect_addr} ${local_addr} "${extra_args}"
 lret=$?
 if [ $lret -ne 0 ]; then
  ret=$lret
  return 1
 fi

 if [ $do_tcp -eq 0 ]; then
  # don't bother testing fallback tcp except for loopback case.
  if [ ${listener_ns} != ${connector_ns} ]; then
   return 0
  fi
 fi

 do_transfer ${listener_ns} ${connector_ns} MPTCP TCP \
      ${connect_addr} ${local_addr} "${extra_args}"
 lret=$?
 if [ $lret -ne 0 ]; then
  ret=$lret
  return 1
 fi

 do_transfer ${listener_ns} ${connector_ns} TCP MPTCP \
      ${connect_addr} ${local_addr} "${extra_args}"
 lret=$?
 if [ $lret -ne 0 ]; then
  ret=$lret
  return 1
 fi

 if [ $do_tcp -gt 1 ] ;then
  do_transfer ${listener_ns} ${connector_ns} TCP TCP \
       ${connect_addr} ${local_addr} "${extra_args}"
  lret=$?
  if [ $lret -ne 0 ]; then
   ret=$lret
   return 1
  fi
 fi

 return 0
}

run_tests()
{
 run_tests_lo $1 $2 $3 0
}

run_test_transparent()
{
 local connect_addr="$1"
 local msg="$2"

 local connector_ns="$ns1"
 local listener_ns="$ns2"
 local lret=0
 local r6flag=""

 TEST_GROUP="${msg}"

 # skip if we don't want v6
 if ! $ipv6 && mptcp_lib_is_v6 "${connect_addr}"; then
  return 0
 fi

 # IP(V6)_TRANSPARENT has been added after TOS support which came with
 # the required infrastructure in MPTCP sockopt code. To support TOS, the
 # following function has been exported (T). Not great but better than
 # checking for a specific kernel version.
 if ! mptcp_lib_kallsyms_has "T __ip_sock_set_tos$"; then
  mptcp_lib_pr_skip "${msg} not supported by the kernel"
  mptcp_lib_result_skip "${TEST_GROUP}"
  return
 fi

 if ! ip netns exec "$listener_ns" nft -f /dev/stdin <<"EOF"
flush ruleset
table inet mangle {
 chain divert {
  type filter hook prerouting priority -150;

  meta l4proto tcp socket transparent 1 meta mark set 1 accept
  tcp dport 20000 tproxy to :20000 meta mark set 1 accept
 }
}
EOF
 then
  mptcp_lib_pr_skip "$msg, could not load nft ruleset"
  mptcp_lib_fail_if_expected_feature "nft rules"
  mptcp_lib_result_skip "${TEST_GROUP}"
  return
 fi

 local local_addr
 if mptcp_lib_is_v6 "${connect_addr}"; then
  local_addr="::"
  r6flag="-6"
 else
  local_addr="0.0.0.0"
 fi

 if ! ip -net "$listener_ns" $r6flag rule add fwmark 1 lookup 100; then
  ip netns exec "$listener_ns" nft flush ruleset
  mptcp_lib_pr_skip "$msg, ip $r6flag rule failed"
  mptcp_lib_fail_if_expected_feature "ip rule"
  mptcp_lib_result_skip "${TEST_GROUP}"
  return
 fi

 if ! ip -net "$listener_ns" route add local $local_addr/0 dev lo table 100; then
  ip netns exec "$listener_ns" nft flush ruleset
  ip -net "$listener_ns" $r6flag rule del fwmark 1 lookup 100
  mptcp_lib_pr_skip "$msg, ip route add local $local_addr failed"
  mptcp_lib_fail_if_expected_feature "ip route"
  mptcp_lib_result_skip "${TEST_GROUP}"
  return
 fi

 mptcp_lib_pr_info "test $msg"

 port=$((20000 - 1))
 local extra_args="-o TRANSPARENT"
 do_transfer ${listener_ns} ${connector_ns} MPTCP MPTCP \
      ${connect_addr} ${local_addr} "${extra_args}"
 lret=$?

 ip netns exec "$listener_ns" nft flush ruleset
 ip -net "$listener_ns" $r6flag rule del fwmark 1 lookup 100
 ip -net "$listener_ns" route del local $local_addr/0 dev lo table 100

 if [ $lret -ne 0 ]; then
  mptcp_lib_pr_fail "$msg, mptcp connection error"
  ret=$lret
  return 1
 fi

 mptcp_lib_pr_info "$msg pass"
 return 0
}

run_tests_peekmode()
{
 local peekmode="$1"

 TEST_GROUP="peek mode: ${peekmode}"
 mptcp_lib_pr_info "with peek mode: ${peekmode}"
 run_tests_lo "$ns1" "$ns1" 10.0.1.1 1 "-P ${peekmode}"
 run_tests_lo "$ns1" "$ns1" dead:beef:1::1 1 "-P ${peekmode}"
}

run_tests_mptfo()
{
 TEST_GROUP="MPTFO"

 if ! mptcp_lib_kallsyms_has "mptcp_fastopen_"; then
  mptcp_lib_pr_skip "TFO not supported by the kernel"
  mptcp_lib_result_skip "${TEST_GROUP}"
  return
 fi

 mptcp_lib_pr_info "with MPTFO start"
 ip netns exec "$ns1" sysctl -q net.ipv4.tcp_fastopen=2
 ip netns exec "$ns2" sysctl -q net.ipv4.tcp_fastopen=1

 run_tests_lo "$ns1" "$ns2" 10.0.1.1 0 "-o MPTFO"
 run_tests_lo "$ns1" "$ns2" 10.0.1.1 0 "-o MPTFO"

 run_tests_lo "$ns1" "$ns2" dead:beef:1::1 0 "-o MPTFO"
 run_tests_lo "$ns1" "$ns2" dead:beef:1::1 0 "-o MPTFO"

 ip netns exec "$ns1" sysctl -q net.ipv4.tcp_fastopen=0
 ip netns exec "$ns2" sysctl -q net.ipv4.tcp_fastopen=0
 mptcp_lib_pr_info "with MPTFO end"
}

run_tests_disconnect()
{
 local old_cin=$cin
 local old_sin=$sin

 TEST_GROUP="full disconnect"

 if ! mptcp_lib_kallsyms_has "mptcp_pm_data_reset$"; then
  mptcp_lib_pr_skip "Full disconnect not supported"
  mptcp_lib_result_skip "${TEST_GROUP}"
  return
 fi

 cat $cin $cin $cin > "$cin".disconnect

 # force do_transfer to cope with the multiple transmissions
 sin="$cin.disconnect"
 cin="$cin.disconnect"
 cin_disconnect="$old_cin"
 connect_per_transfer=3

 mptcp_lib_pr_info "disconnect"
 run_tests_lo "$ns1" "$ns1" 10.0.1.1 1 "-I 3 -i $old_cin"
 run_tests_lo "$ns1" "$ns1" dead:beef:1::1 1 "-I 3 -i $old_cin"

 # restore previous status
 sin=$old_sin
 cin=$old_cin
 cin_disconnect="$cin".disconnect
 connect_per_transfer=1
}

display_time()
{
 time_end=$(date +%s)
 time_run=$((time_end-time_start))

 echo "Time: ${time_run} seconds"
}

log_if_error()
{
 local msg="$1"

 if [ ${ret} -ne 0 ]; then
  mptcp_lib_pr_fail "${msg}"

  final_ret=${ret}
  ret=${KSFT_PASS}

  return ${final_ret}
 fi
}

stop_if_error()
{
 if ! log_if_error "${@}"; then
  display_time
  mptcp_lib_result_print_all_tap
  exit ${final_ret}
 fi
}

make_file "$cin" "client"
make_file "$sin" "server"

mptcp_lib_subtests_last_ts_reset

check_mptcp_disabled

stop_if_error "The kernel configuration is not valid for MPTCP"

print_larger_title "Validating network environment with pings"
for sender in "$ns1" "$ns2" "$ns3" "$ns4";do
 do_ping "$ns1" $sender 10.0.1.1
 do_ping "$ns1" $sender dead:beef:1::1

 do_ping "$ns2" $sender 10.0.1.2
 do_ping "$ns2" $sender dead:beef:1::2
 do_ping "$ns2" $sender 10.0.2.1
 do_ping "$ns2" $sender dead:beef:2::1

 do_ping "$ns3" $sender 10.0.2.2
 do_ping "$ns3" $sender dead:beef:2::2
 do_ping "$ns3" $sender 10.0.3.2
 do_ping "$ns3" $sender dead:beef:3::2

 do_ping "$ns4" $sender 10.0.3.1
 do_ping "$ns4" $sender dead:beef:3::1
done

mptcp_lib_result_code "${ret}" "ping tests"

stop_if_error "Could not even run ping tests"
mptcp_lib_pr_ok

[ -n "$tc_loss" ] && tc -net "$ns2" qdisc add dev ns2eth3 root netem loss random $tc_loss delay ${tc_delay}ms
tc_info="loss of $tc_loss "
test "$tc_delay" -gt 0 && tc_info+="delay $tc_delay ms "

reorder_delay=$((tc_delay / 4))

if [ -z "${tc_reorder}" ]; then
 reorder1=$((RANDOM%10))
 reorder1=$((100 - reorder1))
 reorder2=$((RANDOM%100))

 if [ $reorder_delay -gt 0 ] && [ $reorder1 -lt 100 ] && [ $reorder2 -gt 0 ]; then
  tc_reorder="reorder ${reorder1}% ${reorder2}%"
  tc_info+="$tc_reorder with delay ${reorder_delay}ms "
 fi
elif [ "$tc_reorder" = "0" ];then
 tc_reorder=""
elif [ "$reorder_delay" -gt 0 ];then
 # reordering requires some delay
 tc_reorder="reorder $tc_reorder"
 tc_info+="$tc_reorder with delay ${reorder_delay}ms "
fi

mptcp_lib_pr_info "Using ${tc_info}on ns3eth4"

tc -net "$ns3" qdisc add dev ns3eth4 root netem delay ${reorder_delay}ms $tc_reorder

TEST_GROUP="loopback v4"
run_tests_lo "$ns1" "$ns1" 10.0.1.1 1
stop_if_error "Could not even run loopback test"

TEST_GROUP="loopback v6"
run_tests_lo "$ns1" "$ns1" dead:beef:1::1 1
stop_if_error "Could not even run loopback v6 test"

TEST_GROUP="multihosts"
for sender in $ns1 $ns2 $ns3 $ns4;do
 # ns1<->ns2 is not subject to reordering/tc delays. Use it to test
 # mptcp syncookie support.
 if [ $sender = $ns1 ]; then
  ip netns exec "$ns2" sysctl -q net.ipv4.tcp_syncookies=2
 else
  ip netns exec "$ns2" sysctl -q net.ipv4.tcp_syncookies=1
 fi

 run_tests "$ns1" $sender 10.0.1.1
 run_tests "$ns1" $sender dead:beef:1::1

 run_tests "$ns2" $sender 10.0.1.2
 run_tests "$ns2" $sender dead:beef:1::2
 run_tests "$ns2" $sender 10.0.2.1
 run_tests "$ns2" $sender dead:beef:2::1

 run_tests "$ns3" $sender 10.0.2.2
 run_tests "$ns3" $sender dead:beef:2::2
 run_tests "$ns3" $sender 10.0.3.2
 run_tests "$ns3" $sender dead:beef:3::2

 run_tests "$ns4" $sender 10.0.3.1
 run_tests "$ns4" $sender dead:beef:3::1

 log_if_error "Tests with $sender as a sender have failed"
done

run_tests_peekmode "saveWithPeek"
run_tests_peekmode "saveAfterPeek"
log_if_error "Tests with peek mode have failed"

# MPTFO (MultiPath TCP Fatopen tests)
run_tests_mptfo
log_if_error "Tests with MPTFO have failed"

# connect to ns4 ip address, ns2 should intercept/proxy
run_test_transparent 10.0.3.1 "tproxy ipv4"
run_test_transparent dead:beef:3::1 "tproxy ipv6"
log_if_error "Tests with tproxy have failed"

run_tests_disconnect
log_if_error "Tests of the full disconnection have failed"

display_time
mptcp_lib_result_print_all_tap
exit ${final_ret}