diff --git a/net/rds/ib_cm.c b/net/rds/ib_cm.c index 0c64c504f79d..6b40345ba44d 100644 --- a/net/rds/ib_cm.c +++ b/net/rds/ib_cm.c @@ -1038,6 +1038,19 @@ int rds_ib_conn_path_connect(struct rds_conn_path *cp) return ret; } +static unsigned long rds_ib_conn_path_shutdown_check_wait(struct rds_conn_path *cp) +{ + struct rds_connection *conn = cp->cp_conn; + struct rds_ib_connection *ic = conn->c_transport_data; + + return (!ic->i_cm_id || + (rds_ib_ring_empty(&ic->i_recv_ring) && + (atomic_read(&ic->i_signaled_sends) == 0) && + (atomic_read(&ic->i_fastreg_inuse_count)) == 0 && + (atomic_read(&ic->i_fastreg_wrs) == RDS_IB_DEFAULT_FR_WR))) ? 0 + : msecs_to_jiffies(1000); +} + /* * This is so careful about only cleaning up resources that were built up * so that it can be called at any point during startup. In fact it @@ -1078,11 +1091,13 @@ void rds_ib_conn_path_shutdown(struct rds_conn_path *cp) * sends to complete we're ensured that there will be no * more tx processing. */ - wait_event(rds_ib_ring_empty_wait, - rds_ib_ring_empty(&ic->i_recv_ring) && - (atomic_read(&ic->i_signaled_sends) == 0) && - (atomic_read(&ic->i_fastreg_inuse_count) == 0) && - (atomic_read(&ic->i_fastreg_wrs) == RDS_IB_DEFAULT_FR_WR)); + while (!wait_event_timeout(rds_ib_ring_empty_wait, + rds_ib_conn_path_shutdown_check_wait(cp) == 0, + msecs_to_jiffies(1000))) { + tasklet_schedule(&ic->i_send_tasklet); + tasklet_schedule(&ic->i_recv_tasklet); + } + tasklet_kill(&ic->i_send_tasklet); tasklet_kill(&ic->i_recv_tasklet); diff --git a/tools/testing/selftests/net/rds/README.txt b/tools/testing/selftests/net/rds/README.txt index 295dc82c0770..bac6f15a80d5 100644 --- a/tools/testing/selftests/net/rds/README.txt +++ b/tools/testing/selftests/net/rds/README.txt @@ -1,18 +1,22 @@ RDS self-tests ============== -These scripts provide a coverage test for RDS-TCP by creating two -network namespaces and running rds packets between them. A loopback -network is provisioned with optional probability of packet loss or -corruption. A workload of 50000 hashes, each 64 characters in size, -are passed over an RDS socket on this test network. A passing test means -the RDS-TCP stack was able to recover properly. The provided config.sh -can be used to compile the kernel with the necessary gcov options. The -kernel may optionally be configured to omit the coverage report as well. +These scripts provide a coverage test for RDS-TCP and RDS-RDMA (over +RoCE/RXE) by setting up two endpoints and running RDS packets between +them. The TCP path creates two network namespaces; the RDMA path uses +an RXE (soft RoCE) device backed by a veth pair. A workload of 50000 +hashes, each 64 characters in size, is passed over an RDS socket on +this test network with an optional probability of packet loss or +corruption. A passing test means the RDS stack was able to recover +properly. The provided config.sh can be used to compile the kernel +with the necessary gcov options; pass -r to also enable the kernel +configs required for the RDMA transport. The kernel may optionally be +configured to omit the coverage report as well. USAGE: run.sh [-d logdir] [-l packet_loss] [-c packet_corruption] [-u packet_duplicate] [-t timeout] + [-T tcp|rdma|tcp,rdma] OPTIONS: -d Log directory. If set, logs will be stored in the @@ -27,6 +31,10 @@ OPTIONS: -t Test timeout. Defaults to tools/testing/selftests/net/rds/settings + -T Comma-separated list of transports to test. Accepts + "tcp", "rdma", or "tcp,rdma". Defaults to "tcp". Use + config.sh -r to enable required RDMA configs + ENV VARIABLES: RDS_LOG_DIR Log directory. If set, logs will be stored in the given dir, or skipped if unset. Log dir @@ -48,6 +56,9 @@ EXAMPLE: # Create a suitable gcov enabled .config tools/testing/selftests/net/rds/config.sh -g + # Optionally add RDMA configs (CONFIG_RDS_RDMA, CONFIG_RDMA_RXE) + tools/testing/selftests/net/rds/config.sh -r + # Alternatly create a gcov disabled .config tools/testing/selftests/net/rds/config.sh @@ -62,5 +73,5 @@ EXAMPLE: "export PYTHONPATH=tools/testing/selftests/net/; \ export SUDO_USER=example_user; \ export RDS_LOG_DIR=tools/testing/selftests/net/rds/rds_logs; \ - tools/testing/selftests/net/rds/run.sh" + tools/testing/selftests/net/rds/run.sh -T tcp,rdma" diff --git a/tools/testing/selftests/net/rds/config.sh b/tools/testing/selftests/net/rds/config.sh index 29a79314dd60..be0668359a07 100755 --- a/tools/testing/selftests/net/rds/config.sh +++ b/tools/testing/selftests/net/rds/config.sh @@ -10,7 +10,8 @@ CONF_FILE="" FLAGS=() GENERATE_GCOV_REPORT=0 -while getopts "gc:" opt; do +ENABLE_RDMA=0 +while getopts "gc:r" opt; do case ${opt} in g) GENERATE_GCOV_REPORT=1 @@ -18,8 +19,11 @@ while getopts "gc:" opt; do c) CONF_FILE=$OPTARG ;; + r) + ENABLE_RDMA=1 + ;; :) - echo "USAGE: config.sh [-g] [-c config]" + echo "USAGE: config.sh [-g] [-c config] [-r]" exit 1 ;; ?) @@ -58,3 +62,10 @@ scripts/config "${FLAGS[@]}" --enable CONFIG_VETH # simulate packet loss scripts/config "${FLAGS[@]}" --enable CONFIG_NET_SCH_NETEM +if [ "$ENABLE_RDMA" -eq 1 ]; then + # enable RDS over InfiniBand / RDMA (rds_rdma test) + scripts/config "${FLAGS[@]}" --enable CONFIG_INFINIBAND + scripts/config "${FLAGS[@]}" --enable CONFIG_INFINIBAND_ADDR_TRANS + scripts/config "${FLAGS[@]}" --enable CONFIG_RDMA_RXE + scripts/config "${FLAGS[@]}" --enable CONFIG_RDS_RDMA +fi diff --git a/tools/testing/selftests/net/rds/run.sh b/tools/testing/selftests/net/rds/run.sh index 424fd57401d8..07af2f927a2a 100755 --- a/tools/testing/selftests/net/rds/run.sh +++ b/tools/testing/selftests/net/rds/run.sh @@ -101,6 +101,16 @@ check_conf_enabled() { exit 4 fi } + +check_rdma_conf_enabled() { + if ! grep -x "$1=y" "$kconfig" > /dev/null 2>&1; then + echo "selftests: [SKIP] rdma transport requires $1 enabled" + echo "To enable, run " \ + "tools/testing/selftests/net/rds/config.sh -r and rebuild" + exit 4 + fi +} + check_conf_disabled() { if grep -x "$1=y" "$kconfig" > /dev/null 2>&1; then echo "selftests: [SKIP] This test requires $1 disabled" @@ -117,6 +127,28 @@ check_conf() { check_conf_disabled CONFIG_MODULES } +# Check kernel config and host environment for RDS-RDMA support. +# Exits with SKIP (4) if the user requested rdma but prerequisites +# are not met. +check_rdma_conf() +{ + case "$TRANSPORT" in + *rdma*) ;; + *) return ;; + esac + + # Kconfig will enforce CONFIG_INFINIBAND_* as dependencies + # of CONFIG_RDMA_RXE + check_rdma_conf_enabled CONFIG_RDMA_RXE + check_rdma_conf_enabled CONFIG_RDS_RDMA + + if ! which rdma > /dev/null 2>&1; then + echo "selftests: [SKIP] rdma transport requires the 'rdma'" \ + " tool (iproute2)" + exit 4 + fi +} + check_env() { if ! test -d "$obj_dir"; then @@ -153,8 +185,10 @@ check_env() LOG_DIR="${RDS_LOG_DIR:-}" TIMEOUT=$timeout GENERATE_GCOV_REPORT=1 +TRANSPORT=tcp FLAGS=() -while getopts "d:l:c:u:t:" opt; do + +while getopts "d:l:c:u:t:T:" opt; do case ${opt} in d) LOG_DIR=${OPTARG} @@ -171,9 +205,12 @@ while getopts "d:l:c:u:t:" opt; do u) FLAGS+=("-u" "${OPTARG}") ;; + T) + TRANSPORT=${OPTARG} + ;; :) echo "USAGE: run.sh [-d logdir] [-l packet_loss] [-c packet_corruption]" \ - "[-u packet_duplicate] [-t timeout]" + "[-u packet_duplicate] [-t timeout] [-T tcp|rdma|tcp,rdma]" exit 1 ;; ?) @@ -183,9 +220,21 @@ while getopts "d:l:c:u:t:" opt; do esac done +# Validate transport tokens +IFS=',' read -ra transports <<< "$TRANSPORT" +for t in "${transports[@]}"; do + if [ "$t" != "tcp" ] && [ "$t" != "rdma" ]; then + echo "run.sh: unknown transport '$t' (expected tcp or rdma)" + exit 1 + fi +done + +FLAGS+=("--transport" "${TRANSPORT}") + check_env check_conf check_gcov_conf +check_rdma_conf TRACE_CMD=() if [[ -n "$LOG_DIR" ]]; then diff --git a/tools/testing/selftests/net/rds/test.py b/tools/testing/selftests/net/rds/test.py index 6db606779231..08f2a846a8ab 100755 --- a/tools/testing/selftests/net/rds/test.py +++ b/tools/testing/selftests/net/rds/test.py @@ -5,21 +5,24 @@ This module provides functional testing for the net/rds component. """ import argparse +import atexit import ctypes import errno import hashlib import os import select +import re import signal import socket import subprocess import sys +import time # Allow utils module to be imported from different directory this_dir = os.path.dirname(os.path.realpath(__file__)) sys.path.append(os.path.join(this_dir, "../")) # pylint: disable-next=wrong-import-position,import-error,no-name-in-module -from lib.py.utils import ip # noqa: E402 +from lib.py.utils import ip, cmd # noqa: E402 # pylint: disable-next=wrong-import-position,import-error,no-name-in-module from lib.py.ksft import ksft_pr # noqa: E402 @@ -32,6 +35,36 @@ NET1 = 'net1' VETH0 = 'veth0' VETH1 = 'veth1' +tcpdump_procs = [] +tcp_addrs = [ + # we technically don't need different port numbers, but this will + # help identify traffic in the network analyzer + ('10.0.0.1', 10000), + ('10.0.0.2', 20000), +] + +# RDMA network configs +RXE_DEV0 = 'rxe0' +RXE_DEV1 = 'rxe1' + +VETH_RDMA0 = 'veth_rdma0' +VETH_RDMA1 = 'veth_rdma1' + +rdma_addrs = [ + ('10.0.0.3', 30000), + ('10.0.0.4', 30000), +] + +# send_packets flag space +OP_FLAG_TCP = 0x1 +OP_FLAG_RDMA = 0x2 + +signal_handler_label = "" + +tap_idx = 0 +nr_pass = 0 +nr_fail = 0 + # Helper function for creating a socket inside a network namespace. # We need this because otherwise RDS will detect that the two TCP # sockets are on the same interface and use the loop transport instead @@ -47,29 +80,195 @@ def netns_socket(netns, *sock_args): child = os.fork() if child == 0: - # change network namespace - with open(f'/var/run/netns/{netns}', encoding='utf-8') as f: - try: + try: + # change network namespace + with open(f'/var/run/netns/{netns}', encoding='utf-8') as f: setns(f.fileno(), 0) - except IOError as e: - print(e.errno) - print(e) + # create socket in target namespace + sock = socket.socket(*sock_args) - # create socket in target namespace - sock = socket.socket(*sock_args) + # send resulting socket to parent + socket.send_fds(u0, [], [sock.fileno()]) - # send resulting socket to parent - socket.send_fds(u0, [], [sock.fileno()]) - - os._exit(0) + os._exit(0) + except BaseException: + os._exit(1) # receive socket from child _, fds, _, _ = socket.recv_fds(u1, 0, 1) - os.waitpid(child, 0) + _, status = os.waitpid(child, 0) u0.close() u1.close() + if not os.WIFEXITED(status) or os.WEXITSTATUS(status) != 0: + raise RuntimeError( + f"netns_socket child failed in netns {netns} (status={status})") return socket.fromfd(fds[0], *sock_args) +def send_burst(socks, ip_addrs, snd_hashes, nr_sent, nr_total): + """Send until blocked or nr_total reached. Return updated nr_sent.""" + + while nr_sent < nr_total: + data = hashlib.sha256( + f'packet {nr_sent}'.encode('utf-8')).hexdigest().encode('utf-8') + # pseudo-random send/receive pattern + snd_idx = nr_sent % 2 + rcv_idx = 1 - (nr_sent % 3) % 2 + + snd = socks[snd_idx] + rcv = socks[rcv_idx] + try: + snd.sendto(data, ip_addrs[rcv_idx]) + except BlockingIOError: + return nr_sent + except OSError as e: + if e.errno in (errno.ENOBUFS, errno.ECONNRESET, errno.EPIPE): + return nr_sent + raise + snd_hashes.setdefault((snd.fileno(), rcv.fileno()), + hashlib.sha256()).update(f'<{data}>'.encode('utf-8')) + nr_sent += 1 + return nr_sent + +def recv_burst(epoll, socks, ip_addrs, rcv_hashes, nr_rcv): + """Drain whatever's readable from epoll. Return updated nr_recv.""" + for filen, evntmask in epoll.poll(): + if not evntmask & select.EPOLLRDNORM: + continue + rcv = next(s for s in socks if s.fileno() == filen) + while True: + try: + data, adr = rcv.recvfrom(1024) + except BlockingIOError: + break + snd_idx = ip_addrs.index(adr) + snd = socks[snd_idx] + rcv_hashes.setdefault((snd.fileno(), rcv.fileno()), + hashlib.sha256()).update(f'<{data}>'.encode('utf-8')) + nr_rcv += 1 + return nr_rcv + +def check_info(socks): + """ + Check all rds info pages for errors + + :param socks: list of sockets to check + """ + + # the Python socket module doesn't know these + rds_info_first = 10000 + rds_info_last = 10017 + + nr_success = 0 + nr_error = 0 + + for sock in socks: + for optname in range(rds_info_first, rds_info_last + 1): + # Sigh, the Python socket module doesn't allow us to pass + # buffer lengths greater than 1024 for some reason. RDS + # wants multiple pages. + try: + sock.getsockopt(socket.SOL_RDS, optname, 1024) + nr_success = nr_success + 1 + except OSError as e: + nr_error = nr_error + 1 + if e.errno == errno.ENOSPC: + # ignore + pass + + ksft_pr(f"getsockopt(): {nr_success}/{nr_error}") + +def verify_hashes(snd_hashes, rcv_hashes): + """Compare send/recv hashes per (sender, receiver) pair.""" + for key, snd_hash in snd_hashes.items(): + rcv_hash = rcv_hashes.get(key) + if rcv_hash is None: + ksft_pr("FAIL: No data received") + return 1 + if snd_hash.hexdigest() != rcv_hash.hexdigest(): + ksft_pr("FAIL: Send/recv mismatch") + ksft_pr("hash expected:", snd_hash.hexdigest()) + ksft_pr("hash received:", rcv_hash.hexdigest()) + return 1 + ksft_pr(f"{key[0]}/{key[1]}: ok") + return 0 + +def snd_rcv_packets(env): + """ + Send packets on the given network interfaces + + :param env: transport-environment dict for setup_tcp() / setup_rdma(). + "addrs": list of (ip, port) tuples matching the sockets + "netns": list of netns names for TCP or None for RDMA + "flags": OP_FLAG_TCP or OP_FLAG_RDMA, selects sockets + """ + + addrs = env["addrs"] + netns_list = env["netns"] + flags = env.get("flags", 0) + + if (flags & OP_FLAG_TCP) and (flags & OP_FLAG_RDMA): + raise RuntimeError(f"Invalid transport flag sets multiple transports: {flags}") + + if flags & OP_FLAG_TCP: + sockets = [ + netns_socket(netns_list[0], socket.AF_RDS, socket.SOCK_SEQPACKET), + netns_socket(netns_list[1], socket.AF_RDS, socket.SOCK_SEQPACKET), + ] + elif flags & OP_FLAG_RDMA: + sockets = [ + socket.socket(socket.AF_RDS, socket.SOCK_SEQPACKET), + socket.socket(socket.AF_RDS, socket.SOCK_SEQPACKET), + ] + else: + raise RuntimeError(f"Invalid transport flag sets no transports: {flags}") + + for s, addr in zip(sockets, addrs): + s.bind(addr) + s.setblocking(0) + + send_hashes = {} + recv_hashes = {} + + ep = select.epoll() + + for s in sockets: + ep.register(s, select.EPOLLRDNORM) + + num_packets = 50000 + nr_send = 0 + nr_recv = 0 + + while nr_send < num_packets: + + # Send as much as we can without blocking + ksft_pr("sending...", nr_send, nr_recv) + nr_send = send_burst(sockets, addrs, send_hashes, nr_send, num_packets) + + # Receive as much as we can without blocking + ksft_pr("receiving...", nr_send, nr_recv) + while nr_recv < nr_send: + nr_recv = recv_burst(ep, sockets, addrs, recv_hashes, nr_recv) + + # exercise net/rds/tcp.c:rds_tcp_sysctl_reset() + if netns_list: + for net in netns_list: + ip(f"netns exec {net} /usr/sbin/sysctl net.rds.tcp.rds_tcp_rcvbuf=10000") + ip(f"netns exec {net} /usr/sbin/sysctl net.rds.tcp.rds_tcp_sndbuf=10000") + + ksft_pr("done", nr_send, nr_recv) + + check_info(sockets) + + # We're done sending and receiving stuff, now let's check if what + # we received is what we sent. + rc = verify_hashes(send_hashes, recv_hashes) + + ep.close() + for s in sockets: + s.close() + + return rc + def stop_pcaps(): """Stop tcpdump processes. @@ -95,17 +294,174 @@ def signal_handler(_sig, _frame): """ Test timed out signal handler """ - ksft_pr("Test timed out") - stop_pcaps() - print("not ok 1 rds selftest") + ksft_pr(f"Test timed out: {signal_handler_label}") + print(f"not ok {tap_idx} rds selftest {signal_handler_label}") sys.exit(1) +def setup_tcp(): + """ + Configure tcp network + """ + + # clean up any leftovers from a previously interrupted run + teardown_tcp() + + ip(f"netns add {NET0}") + ip(f"netns add {NET1}") + ip("link add type veth") + + # Move TCP interfaces into separate namespaces so they can no longer be + # bound directly; this prevents rds from switching over from the tcp + # transport to the loop transport. + ip(f"link set {VETH0} netns {NET0} up") + ip(f"link set {VETH1} netns {NET1} up") + + # add addresses + ip(f"-n {NET0} addr add {tcp_addrs[0][0]}/32 dev {VETH0}") + ip(f"-n {NET1} addr add {tcp_addrs[1][0]}/32 dev {VETH1}") + + # add routes + ip(f"-n {NET0} route add {tcp_addrs[1][0]}/32 dev {VETH0}") + ip(f"-n {NET1} route add {tcp_addrs[0][0]}/32 dev {VETH1}") + + # sanity check that our two interfaces/addresses are correctly set up + # and communicating by doing a single ping + ip(f"netns exec {NET0} ping -c 1 {tcp_addrs[1][0]}") + + # Start a packet capture on each network + if logdir is not None: + for netn in [NET0, NET1]: + pcap = logdir+'/rds-'+netn+'.pcap' + + tcpdump_cmd = ['ip', 'netns', 'exec', netn, '/usr/sbin/tcpdump'] + sudo_user = os.environ.get('SUDO_USER') + if sudo_user: + tcpdump_cmd.extend(['-Z', sudo_user]) + tcpdump_cmd.extend(['-i', 'any', '-w', pcap]) + + # pylint: disable-next=consider-using-with + p = subprocess.Popen(tcpdump_cmd, + stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) + tcpdump_procs.append(p) + + # simulate packet loss, duplication and corruption + for netn, iface in [(NET0, VETH0), (NET1, VETH1)]: + ip(f"netns exec {netn} /usr/sbin/tc qdisc add dev {iface} root netem \ + corrupt {PACKET_CORRUPTION} loss {PACKET_LOSS} duplicate \ + {PACKET_DUPLICATE}") + +def teardown_tcp(): + """ + Tear down the tcp network configured by setup_tcp(). + + Removing the namespaces also removes the veth pair, addresses, + routes, and netem qdisc that live inside them. fail=False so + this is safe to call in error paths after a partial or complete setup. + """ + cmd(f"ip netns del {NET0}", fail=False) + cmd(f"ip netns del {NET1}", fail=False) + +def get_iface_mac(iface): + """Return the MAC address of a local network interface.""" + out = subprocess.check_output(['ip', 'link', 'show', iface], text=True) + mac = re.search(r'link/ether\s+([0-9a-f:]+)', out) + if not mac: + raise RuntimeError(f"Cannot determine MAC address of {iface}") + return mac.group(1) + +def setup_rdma(): + """ + Configure rdma network + """ + + # remove links left over by previously interrupted run. + teardown_rdma() + + # use call here since modprobe may fail if the rdma_rxe + # module is built-in + subprocess.call(['modprobe', 'rdma_rxe'], + stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) + + ip(f"link add {VETH_RDMA0} type veth peer name {VETH_RDMA1}") + + ip(f"link set {VETH_RDMA0} up") + ip(f"link set {VETH_RDMA1} up") + + # Since both addresses are in the same namespace, the source address + # is always local, so enable accept_local + cmd(f"/usr/sbin/sysctl -q net.ipv4.conf.{VETH_RDMA0}.accept_local=1") + cmd(f"/usr/sbin/sysctl -q net.ipv4.conf.{VETH_RDMA1}.accept_local=1") + + # Reverse path filters must be disabled so that the local routes don't + # cause RPF failures. + cmd(f"/usr/sbin/sysctl -q net.ipv4.conf.{VETH_RDMA0}.rp_filter=0") + cmd(f"/usr/sbin/sysctl -q net.ipv4.conf.{VETH_RDMA1}.rp_filter=0") + + # add addresses + ip(f"addr add {rdma_addrs[0][0]}/32 dev {VETH_RDMA0}") + ip(f"addr add {rdma_addrs[1][0]}/32 dev {VETH_RDMA1}") + + # add routes + ip(f"route add {rdma_addrs[1][0]}/32 dev {VETH_RDMA0}") + ip(f"route add {rdma_addrs[0][0]}/32 dev {VETH_RDMA1}") + + # ARP will not resolve neighbor IPs on /32 routes without a subnet. + # Avoid this by adding neighbors directly so RDMA CM can populate path + # records with correct mac addrs without waiting for the ARP. + mac0 = get_iface_mac(VETH_RDMA0) + mac1 = get_iface_mac(VETH_RDMA1) + ip(f"neigh add {rdma_addrs[1][0]} lladdr {mac1} dev {VETH_RDMA0} nud permanent") + ip(f"neigh add {rdma_addrs[0][0]} lladdr {mac0} dev {VETH_RDMA1} nud permanent") + + cmd(f'rdma link add {RXE_DEV0} type rxe netdev {VETH_RDMA0}') + cmd(f'rdma link add {RXE_DEV1} type rxe netdev {VETH_RDMA1}') + + time.sleep(1) # allow RXE devices to initialise + + # Start a packet capture on each network + if logdir is not None: + for iface in [VETH_RDMA0, VETH_RDMA1]: + pcap = logdir+'/rds-roce-'+iface+'.pcap' + + tcpdump_cmd = ['/usr/sbin/tcpdump'] + sudo_user = os.environ.get('SUDO_USER') + if sudo_user: + tcpdump_cmd.extend(['-Z', sudo_user]) + tcpdump_cmd.extend(['-i', iface, '-w', pcap]) + + # pylint: disable-next=consider-using-with + p = subprocess.Popen(tcpdump_cmd, + stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) + tcpdump_procs.append(p) + + # simulate packet loss, duplication and corruption + for iface in [VETH_RDMA0, VETH_RDMA1]: + cmd(f"/usr/sbin/tc qdisc add dev {iface} root netem \ + corrupt {PACKET_CORRUPTION} loss {PACKET_LOSS} duplicate \ + {PACKET_DUPLICATE}") + +def teardown_rdma(): + """ + Tear down the rdma network configured by setup_rdma(). + """ + + # remove links left over by previously interrupted run. + cmd(f'rdma link del {RXE_DEV0}', fail=False) + cmd(f'rdma link del {RXE_DEV1}', fail=False) + cmd(f'ip link del {VETH_RDMA0}', fail=False) + + #Parse out command line arguments. We take an optional # timeout parameter and an optional log output folder parser = argparse.ArgumentParser(description="init script args", formatter_class=argparse.ArgumentDefaultsHelpFormatter) parser.add_argument("-d", "--logdir", action="store", help="directory to store logs", default=None) +parser.add_argument("-T", "--transport", default="tcp", + help="Comma-separated list of transports to test: " + "tcp, rdma, or tcp,rdma. Each matching test " + "is run once per transport. " + "'rdma' requires CONFIG_RDS_RDMA and rdma_rxe.") parser.add_argument('-t', '--timeout', help="timeout to terminate hung test", type=int, default=0) parser.add_argument('-l', '--loss', help="Simulate tcp packet loss", @@ -120,200 +476,63 @@ PACKET_LOSS=str(args.loss)+'%' PACKET_CORRUPTION=str(args.corruption)+'%' PACKET_DUPLICATE=str(args.duplicate)+'%' -ip(f"netns add {NET0}") -ip(f"netns add {NET1}") -ip("link add type veth") +# check transport is either tcp or rdma +transports = [t.strip() for t in args.transport.split(',')] +for t in transports: + if t not in ('tcp', 'rdma'): + raise SystemExit(f"test.py: unknown transport: {t!r}") -addrs = [ - # we technically don't need different port numbers, but this will - # help identify traffic in the network analyzer - ('10.0.0.1', 10000), - ('10.0.0.2', 20000), -] +# Register stop_pcaps before any network setups so that any partially setup +# tcpdumps are still cleaned up on error +atexit.register(stop_pcaps) -# move interfaces to separate namespaces so they can no longer be -# bound directly; this prevents rds from switching over from the tcp -# transport to the loop transport. -ip(f"link set {VETH0} netns {NET0} up") -ip(f"link set {VETH1} netns {NET1} up") +# Set up all requested transports upfront so network plumbing is +# ready before any test runs. +transport_envs = {} +FLAGS = 0 +if 'tcp' in transports: + # Register cleanups before setups to handle partial setups that error'd out + atexit.register(teardown_tcp) + setup_tcp() + transport_envs['tcp'] = { + 'addrs': tcp_addrs, + 'netns': [NET0, NET1], + 'flags': FLAGS | OP_FLAG_TCP, + } - - -# add addresses -ip(f"-n {NET0} addr add {addrs[0][0]}/32 dev {VETH0}") -ip(f"-n {NET1} addr add {addrs[1][0]}/32 dev {VETH1}") - -# add routes -ip(f"-n {NET0} route add {addrs[1][0]}/32 dev {VETH0}") -ip(f"-n {NET1} route add {addrs[0][0]}/32 dev {VETH1}") - -# sanity check that our two interfaces/addresses are correctly set up -# and communicating by doing a single ping -ip(f"netns exec {NET0} ping -c 1 {addrs[1][0]}") - -tcpdump_procs = [] -# Start a packet capture on each network -if logdir is not None: - for net in [NET0, NET1]: - pcap = logdir+'/rds-'+net+'.pcap' - - tcpdump_cmd = ['ip', 'netns', 'exec', net, '/usr/sbin/tcpdump'] - sudo_user = os.environ.get('SUDO_USER') - if sudo_user: - tcpdump_cmd.extend(['-Z', sudo_user]) - tcpdump_cmd.extend(['-i', 'any', '-w', pcap]) - - # pylint: disable-next=consider-using-with - p = subprocess.Popen(tcpdump_cmd, - stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) - tcpdump_procs.append(p) - -# simulate packet loss, duplication and corruption -for net, iface in [(NET0, VETH0), (NET1, VETH1)]: - ip(f"netns exec {net} /usr/sbin/tc qdisc add dev {iface} root netem \ - corrupt {PACKET_CORRUPTION} loss {PACKET_LOSS} duplicate \ - {PACKET_DUPLICATE}") +if 'rdma' in transports: + atexit.register(teardown_rdma) + setup_rdma() + transport_envs['rdma'] = { + 'addrs': rdma_addrs, + 'netns': None, + 'flags': FLAGS | OP_FLAG_RDMA, + } print("TAP version 13") -print("1..1") +print(f"1..{len(transport_envs)}") -# add a timeout -if args.timeout > 0: - signal.alarm(args.timeout) - signal.signal(signal.SIGALRM, signal_handler) +for transport, tenv in transport_envs.items(): + tap_idx += 1 -sockets = [ - netns_socket(NET0, socket.AF_RDS, socket.SOCK_SEQPACKET), - netns_socket(NET1, socket.AF_RDS, socket.SOCK_SEQPACKET), -] + # add a timeout + if args.timeout > 0: + signal_handler_label = transport + signal.alarm(args.timeout) + signal.signal(signal.SIGALRM, signal_handler) -for s, addr in zip(sockets, addrs): - s.bind(addr) - s.setblocking(0) + ret = snd_rcv_packets(tenv) -fileno_to_socket = { - s.fileno(): s for s in sockets -} + # cancel timeout + signal.alarm(0) -addr_to_socket = dict(zip(addrs, sockets)) + if ret == 0: + ksft_pr("Success") + print(f"ok {tap_idx} rds selftest {transport}") + nr_pass += 1 + else: + print(f"not ok {tap_idx} rds selftest {transport}") + nr_fail += 1 -socket_to_addr = { - s: addr for addr, s in zip(addrs, sockets) -} - -send_hashes = {} -recv_hashes = {} - -ep = select.epoll() - -for s in sockets: - ep.register(s, select.EPOLLRDNORM) - -NUM_PACKETS = 50000 -nr_send = 0 -nr_recv = 0 - -while nr_send < NUM_PACKETS: - # Send as much as we can without blocking - ksft_pr("sending...", nr_send, nr_recv) - while nr_send < NUM_PACKETS: - send_data = hashlib.sha256( - f'packet {nr_send}'.encode('utf-8')).hexdigest().encode('utf-8') - - # pseudo-random send/receive pattern - sender = sockets[nr_send % 2] - receiver = sockets[1 - (nr_send % 3) % 2] - - try: - sender.sendto(send_data, socket_to_addr[receiver]) - send_hashes.setdefault((sender.fileno(), receiver.fileno()), - hashlib.sha256()).update(f'<{send_data}>'.encode('utf-8')) - nr_send = nr_send + 1 - except BlockingIOError: - break - except OSError as e: - if e.errno in [errno.ENOBUFS, errno.ECONNRESET, errno.EPIPE]: - break - raise - - # Receive as much as we can without blocking - ksft_pr("receiving...", nr_send, nr_recv) - while nr_recv < nr_send: - for fileno, eventmask in ep.poll(): - receiver = fileno_to_socket[fileno] - - if eventmask & select.EPOLLRDNORM: - while True: - try: - recv_data, address = receiver.recvfrom(1024) - sender = addr_to_socket[address] - recv_hashes.setdefault((sender.fileno(), - receiver.fileno()), hashlib.sha256()).update( - f'<{recv_data}>'.encode('utf-8')) - nr_recv = nr_recv + 1 - except BlockingIOError: - break - - # exercise net/rds/tcp.c:rds_tcp_sysctl_reset() - for net in [NET0, NET1]: - ip(f"netns exec {net} /usr/sbin/sysctl net.rds.tcp.rds_tcp_rcvbuf=10000") - ip(f"netns exec {net} /usr/sbin/sysctl net.rds.tcp.rds_tcp_sndbuf=10000") - -ksft_pr("done", nr_send, nr_recv) - -# the Python socket module doesn't know these -RDS_INFO_FIRST = 10000 -RDS_INFO_LAST = 10017 - -nr_success = 0 -nr_error = 0 - -for s in sockets: - for optname in range(RDS_INFO_FIRST, RDS_INFO_LAST + 1): - # Sigh, the Python socket module doesn't allow us to pass - # buffer lengths greater than 1024 for some reason. RDS - # wants multiple pages. - try: - s.getsockopt(socket.SOL_RDS, optname, 1024) - nr_success = nr_success + 1 - except OSError as e: - nr_error = nr_error + 1 - if e.errno == errno.ENOSPC: - # ignore - pass - -ksft_pr(f"getsockopt(): {nr_success}/{nr_error}") - -# cancel timeout -signal.alarm(0) - -stop_pcaps() - -# We're done sending and receiving stuff, now let's check if what -# we received is what we sent. -ret = 0 -for (sender, receiver), send_hash in send_hashes.items(): - recv_hash = recv_hashes.get((sender, receiver)) - - if recv_hash is None: - ksft_pr("FAIL: No data received") - ret = 1 - break - - if send_hash.hexdigest() != recv_hash.hexdigest(): - ksft_pr("FAIL: Send/recv mismatch") - ksft_pr("hash expected:", send_hash.hexdigest()) - ksft_pr("hash received:", recv_hash.hexdigest()) - ret = 1 - break - - ksft_pr(f"{sender}/{receiver}: ok") - -if ret == 0: - ksft_pr("Success") - print("ok 1 rds selftest") -else: - print("not ok 1 rds selftest") - -ksft_pr(f"Totals: pass:{1-ret} fail:{ret} skip:0") -sys.exit(ret) +ksft_pr(f"Totals: pass:{nr_pass} fail:{nr_fail} skip:0") +sys.exit(1 if nr_fail else 0)