linux/tools/testing/selftests/drivers/net/stats.py

290 lines
9.4 KiB
Python
Raw Permalink Normal View History

#!/usr/bin/env python3
# SPDX-License-Identifier: GPL-2.0
"""
Tests related to standard netdevice statistics.
"""
import errno
import subprocess
import time
from lib.py import ksft_run, ksft_exit, ksft_pr
from lib.py import ksft_ge, ksft_eq, ksft_is, ksft_in, ksft_lt, ksft_true, ksft_raises
from lib.py import KsftSkipEx, KsftFailEx
from lib.py import ksft_disruptive
from lib.py import EthtoolFamily, NetdevFamily, RtnlFamily, NlError
from lib.py import NetDrvEnv
from lib.py import cmd, ip, defer
ethnl = EthtoolFamily()
netfam = NetdevFamily()
rtnl = RtnlFamily()
def check_pause(cfg) -> None:
"""
Check that drivers which support Pause config also report standard
pause stats.
"""
try:
ethnl.pause_get({"header": {"dev-index": cfg.ifindex}})
except NlError as e:
if e.error == errno.EOPNOTSUPP:
raise KsftSkipEx("pause not supported by the device") from e
raise
data = ethnl.pause_get({"header": {"dev-index": cfg.ifindex,
"flags": {'stats'}}})
ksft_true(data['stats'], "driver does not report stats")
def check_fec(cfg) -> None:
"""
Check that drivers which support FEC config also report standard
FEC stats.
"""
try:
ethnl.fec_get({"header": {"dev-index": cfg.ifindex}})
except NlError as e:
if e.error == errno.EOPNOTSUPP:
raise KsftSkipEx("FEC not supported by the device") from e
raise
data = ethnl.fec_get({"header": {"dev-index": cfg.ifindex,
"flags": {'stats'}}})
ksft_true(data['stats'], "driver does not report stats")
def pkt_byte_sum(cfg) -> None:
"""
Check that qstat and interface stats match in value.
"""
def get_qstat(test):
stats = netfam.qstats_get({}, dump=True)
if stats:
for qs in stats:
if qs["ifindex"]== test.ifindex:
return qs
return None
qstat = get_qstat(cfg)
if qstat is None:
raise KsftSkipEx("qstats not supported by the device")
for key in ['tx-packets', 'tx-bytes', 'rx-packets', 'rx-bytes']:
ksft_in(key, qstat, "Drivers should always report basic keys")
# Compare stats, rtnl stats and qstats must match,
# but the interface may be up, so do a series of dumps
# each time the more "recent" stats must be higher or same.
def stat_cmp(rstat, qstat):
for key in ['tx-packets', 'tx-bytes', 'rx-packets', 'rx-bytes']:
if rstat[key] != qstat[key]:
return rstat[key] - qstat[key]
return 0
for _ in range(10):
testing: net-drv: use stats64 for testing Testing a network device that has large numbers of bytes/packets may overflow. Using stats64 when comparing fixes this problem. I tripped on this while iterating on a qstats patch for mlx5. See below for confirmation without my added code that this is a bug. Before this patch (with added debugging output): $ NETIF=eth0 tools/testing/selftests/drivers/net/stats.py KTAP version 1 1..4 ok 1 stats.check_pause ok 2 stats.check_fec rstat: 481708634 qstat: 666201639514 key: tx-bytes not ok 3 stats.pkt_byte_sum ok 4 stats.qstat_by_ifindex Note the huge delta above ^^^ in the rtnl vs qstats. After this patch: $ NETIF=eth0 tools/testing/selftests/drivers/net/stats.py KTAP version 1 1..4 ok 1 stats.check_pause ok 2 stats.check_fec ok 3 stats.pkt_byte_sum ok 4 stats.qstat_by_ifindex It looks like rtnl_fill_stats in net/core/rtnetlink.c will attempt to copy the 64bit stats into a 32bit structure which is probably why this behavior is occurring. To show this is happening, you can get the underlying stats that the stats.py test uses like this: $ ./cli.py --spec ../../../Documentation/netlink/specs/rt_link.yaml \ --do getlink --json '{"ifi-index": 7}' And examine the output (heavily snipped to show relevant fields): 'stats': { 'multicast': 3739197, 'rx-bytes': 1201525399, 'rx-packets': 56807158, 'tx-bytes': 492404458, 'tx-packets': 1200285371, 'stats64': { 'multicast': 3739197, 'rx-bytes': 35561263767, 'rx-packets': 56807158, 'tx-bytes': 666212335338, 'tx-packets': 1200285371, The stats.py test prior to this patch was using the 'stats' structure above, which matches the failure output on my system. Comparing side by side, rx-bytes and tx-bytes, and getting ethtool -S output: rx-bytes stats: 1201525399 rx-bytes stats64: 35561263767 rx-bytes ethtool: 36203402638 tx-bytes stats: 492404458 tx-bytes stats64: 666212335338 tx-bytes ethtool: 666215360113 Note that the above was taken from a system with an mlx5 NIC, which only exposes ndo_get_stats64. Based on the ethtool output and qstat output, it appears that stats.py should be updated to use the 'stats64' structure for accurate comparisons when packet/byte counters get very large. To confirm that this was not related to the qstats code I was iterating on, I booted a kernel without my driver changes and re-ran the test which shows the qstats are skipped (as they don't exist for mlx5): NETIF=eth0 tools/testing/selftests/drivers/net/stats.py KTAP version 1 1..4 ok 1 stats.check_pause ok 2 stats.check_fec ok 3 stats.pkt_byte_sum # SKIP qstats not supported by the device ok 4 stats.qstat_by_ifindex # SKIP No ifindex supports qstats But, fetching the stats using the CLI $ ./cli.py --spec ../../../Documentation/netlink/specs/rt_link.yaml \ --do getlink --json '{"ifi-index": 7}' Shows the same issue (heavily snipped for relevant fields only): 'stats': { 'multicast': 105489, 'rx-bytes': 530879526, 'rx-packets': 751415, 'tx-bytes': 2510191396, 'tx-packets': 27700323, 'stats64': { 'multicast': 105489, 'rx-bytes': 530879526, 'rx-packets': 751415, 'tx-bytes': 15395093284, 'tx-packets': 27700323, Comparing side by side with ethtool -S on the unmodified mlx5 driver: tx-bytes stats: 2510191396 tx-bytes stats64: 15395093284 tx-bytes ethtool: 17718435810 Fixes: f0e6c86e4bab ("testing: net-drv: add a driver test for stats reporting") Signed-off-by: Joe Damato <jdamato@fastly.com> Link: https://lore.kernel.org/r/20240520235850.190041-1-jdamato@fastly.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
2024-05-20 23:58:43 +00:00
rtstat = rtnl.getlink({"ifi-index": cfg.ifindex})['stats64']
if stat_cmp(rtstat, qstat) < 0:
raise KsftFailEx("RTNL stats are lower, fetched later")
qstat = get_qstat(cfg)
if stat_cmp(rtstat, qstat) > 0:
raise KsftFailEx("Qstats are lower, fetched later")
def qstat_by_ifindex(cfg) -> None:
""" Qstats Netlink API tests - querying by ifindex. """
# Construct a map ifindex -> [dump, by-index, dump]
ifindexes = {}
stats = netfam.qstats_get({}, dump=True)
for entry in stats:
ifindexes[entry['ifindex']] = [entry, None, None]
for ifindex in ifindexes:
entry = netfam.qstats_get({"ifindex": ifindex}, dump=True)
ksft_eq(len(entry), 1)
ifindexes[entry[0]['ifindex']][1] = entry[0]
stats = netfam.qstats_get({}, dump=True)
for entry in stats:
ifindexes[entry['ifindex']][2] = entry
if len(ifindexes) == 0:
raise KsftSkipEx("No ifindex supports qstats")
# Now make sure the stats match/make sense
for ifindex, triple in ifindexes.items():
all_keys = triple[0].keys() | triple[1].keys() | triple[2].keys()
for key in all_keys:
ksft_ge(triple[1][key], triple[0][key], comment="bad key: " + key)
ksft_ge(triple[2][key], triple[1][key], comment="bad key: " + key)
# Sanity check the dumps
queues = NetdevFamily(recv_size=4096).qstats_get({"scope": "queue"}, dump=True)
# Reformat the output into {ifindex: {rx: [id, id, ...], tx: [id, id, ...]}}
parsed = {}
for entry in queues:
ifindex = entry["ifindex"]
if ifindex not in parsed:
parsed[ifindex] = {"rx":[], "tx": []}
parsed[ifindex][entry["queue-type"]].append(entry['queue-id'])
# Now, validate
for ifindex, queues in parsed.items():
for qtype in ['rx', 'tx']:
ksft_eq(len(queues[qtype]), len(set(queues[qtype])),
comment="repeated queue keys")
ksft_eq(len(queues[qtype]), max(queues[qtype]) + 1,
comment="missing queue keys")
# Test invalid dumps
# 0 is invalid
with ksft_raises(NlError) as cm:
netfam.qstats_get({"ifindex": 0}, dump=True)
ksft_eq(cm.exception.nl_msg.error, -34)
ksft_eq(cm.exception.nl_msg.extack['bad-attr'], '.ifindex')
# loopback has no stats
with ksft_raises(NlError) as cm:
netfam.qstats_get({"ifindex": 1}, dump=True)
ksft_eq(cm.exception.nl_msg.error, -errno.EOPNOTSUPP)
ksft_eq(cm.exception.nl_msg.extack['bad-attr'], '.ifindex')
# Try to get stats for lowest unused ifindex but not 0
devs = rtnl.getlink({}, dump=True)
all_ifindexes = set(dev["ifi-index"] for dev in devs)
lowest = 2
while lowest in all_ifindexes:
lowest += 1
with ksft_raises(NlError) as cm:
netfam.qstats_get({"ifindex": lowest}, dump=True)
ksft_eq(cm.exception.nl_msg.error, -19)
ksft_eq(cm.exception.nl_msg.extack['bad-attr'], '.ifindex')
@ksft_disruptive
def check_down(cfg) -> None:
""" Test statistics (interface and qstat) are not impacted by ifdown """
try:
qstat = netfam.qstats_get({"ifindex": cfg.ifindex}, dump=True)[0]
except NlError as e:
if e.error == errno.EOPNOTSUPP:
raise KsftSkipEx("qstats not supported by the device") from e
raise
ip(f"link set dev {cfg.dev['ifname']} down")
defer(ip, f"link set dev {cfg.dev['ifname']} up")
qstat2 = netfam.qstats_get({"ifindex": cfg.ifindex}, dump=True)[0]
for k in qstat:
ksft_ge(qstat2[k], qstat[k], comment=f"{k} went backwards on device down")
# exercise per-queue API to make sure that "device down" state
# is handled correctly and doesn't crash
netfam.qstats_get({"ifindex": cfg.ifindex, "scope": "queue"}, dump=True)
def __run_inf_loop(body):
body = body.strip()
if body[-1] != ';':
body += ';'
return subprocess.Popen(f"while true; do {body} done", shell=True,
stdout=subprocess.PIPE, stderr=subprocess.PIPE)
def __stats_increase_sanely(old, new) -> None:
for k in old.keys():
ksft_ge(new[k], old[k])
ksft_lt(new[k] - old[k], 1 << 31, comment="likely wrapping error")
def procfs_hammer(cfg) -> None:
"""
Reading stats via procfs only holds the RCU lock, which is not an exclusive
lock, make sure drivers can handle parallel reads of stats.
"""
one = __run_inf_loop("cat /proc/net/dev")
defer(one.kill)
two = __run_inf_loop("cat /proc/net/dev")
defer(two.kill)
time.sleep(1)
# Make sure the processes are running
ksft_is(one.poll(), None)
ksft_is(two.poll(), None)
rtstat1 = rtnl.getlink({"ifi-index": cfg.ifindex})['stats64']
time.sleep(2)
rtstat2 = rtnl.getlink({"ifi-index": cfg.ifindex})['stats64']
__stats_increase_sanely(rtstat1, rtstat2)
# defers will kill the loops
@ksft_disruptive
def procfs_downup_hammer(cfg) -> None:
"""
Reading stats via procfs only holds the RCU lock, drivers often try
to sleep when reading the stats, or don't protect against races.
"""
# Max out the queues, we'll flip between max and 1
channels = ethnl.channels_get({'header': {'dev-index': cfg.ifindex}})
if channels['combined-count'] == 0:
rx_type = 'rx'
else:
rx_type = 'combined'
cur_queue_cnt = channels[f'{rx_type}-count']
max_queue_cnt = channels[f'{rx_type}-max']
cmd(f"ethtool -L {cfg.ifname} {rx_type} {max_queue_cnt}")
defer(cmd, f"ethtool -L {cfg.ifname} {rx_type} {cur_queue_cnt}")
# Real test stats
stats = __run_inf_loop("cat /proc/net/dev")
defer(stats.kill)
ipset = f"ip link set dev {cfg.ifname}"
defer(ip, f"link set dev {cfg.ifname} up")
# The "echo -n 1" lets us count iterations below
updown = f"{ipset} down; sleep 0.05; {ipset} up; sleep 0.05; " + \
f"ethtool -L {cfg.ifname} {rx_type} 1; " + \
f"ethtool -L {cfg.ifname} {rx_type} {max_queue_cnt}; " + \
"echo -n 1"
updown = __run_inf_loop(updown)
kill_updown = defer(updown.kill)
time.sleep(1)
# Make sure the processes are running
ksft_is(stats.poll(), None)
ksft_is(updown.poll(), None)
rtstat1 = rtnl.getlink({"ifi-index": cfg.ifindex})['stats64']
# We're looking for crashes, give it extra time
time.sleep(9)
rtstat2 = rtnl.getlink({"ifi-index": cfg.ifindex})['stats64']
__stats_increase_sanely(rtstat1, rtstat2)
kill_updown.exec()
stdout, _ = updown.communicate(timeout=5)
ksft_pr("completed up/down cycles:", len(stdout.decode('utf-8')))
def main() -> None:
""" Ksft boiler plate main """
with NetDrvEnv(__file__, queue_count=100) as cfg:
ksft_run([check_pause, check_fec, pkt_byte_sum, qstat_by_ifindex,
check_down, procfs_hammer, procfs_downup_hammer],
args=(cfg, ))
ksft_exit()
if __name__ == "__main__":
main()