qemu-e2k/tests/qemu-iotests/136
Max Reitz 19026817f7 iotests: Make 136 less flaky
136 executes some AIO requests without a final aio_flush; then it
advances the virtual clock and thus expects the last access time of the
device to be less than the current time when queried (i.e. idle_time_ns
to be greater than 0).  However, without the aio_flush, some requests
may be settled after the clock_step invocation.  In that case,
idle_time_ns would be 0 and the test fails.

Fix this by adding an aio_flush if any AIO request other than some other
aio_flush has been executed.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20171109203025.27493-6-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
2017-11-14 18:06:25 +01:00

343 lines
13 KiB
Python

#!/usr/bin/env python
#
# Tests for block device statistics
#
# Copyright (C) 2015 Igalia, S.L.
# Author: Alberto Garcia <berto@igalia.com>
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#
import iotests
import os
interval_length = 10
nsec_per_sec = 1000000000
op_latency = nsec_per_sec / 1000 # See qtest_latency_ns in accounting.c
bad_sector = 8192
bad_offset = bad_sector * 512
blkdebug_file = os.path.join(iotests.test_dir, 'blkdebug.conf')
class BlockDeviceStatsTestCase(iotests.QMPTestCase):
test_img = "null-aio://"
total_rd_bytes = 0
total_rd_ops = 0
total_wr_bytes = 0
total_wr_ops = 0
total_wr_merged = 0
total_flush_ops = 0
failed_rd_ops = 0
failed_wr_ops = 0
invalid_rd_ops = 0
invalid_wr_ops = 0
wr_highest_offset = 0
account_invalid = False
account_failed = False
def blockstats(self, device):
result = self.vm.qmp("query-blockstats")
for r in result['return']:
if r['device'] == device:
return r['stats']
raise Exception("Device not found for blockstats: %s" % device)
def create_blkdebug_file(self):
file = open(blkdebug_file, 'w')
file.write('''
[inject-error]
event = "read_aio"
errno = "5"
sector = "%d"
[inject-error]
event = "write_aio"
errno = "5"
sector = "%d"
''' % (bad_sector, bad_sector))
file.close()
def setUp(self):
drive_args = []
drive_args.append("stats-intervals.0=%d" % interval_length)
drive_args.append("stats-account-invalid=%s" %
(self.account_invalid and "on" or "off"))
drive_args.append("stats-account-failed=%s" %
(self.account_failed and "on" or "off"))
self.create_blkdebug_file()
self.vm = iotests.VM().add_drive('blkdebug:%s:%s' %
(blkdebug_file, self.test_img),
','.join(drive_args))
self.vm.launch()
# Set an initial value for the clock
self.vm.qtest("clock_step %d" % nsec_per_sec)
def tearDown(self):
self.vm.shutdown()
os.remove(blkdebug_file)
def accounted_ops(self, read = False, write = False, flush = False):
ops = 0
if write:
ops += self.total_wr_ops
if self.account_failed:
ops += self.failed_wr_ops
if self.account_invalid:
ops += self.invalid_wr_ops
if read:
ops += self.total_rd_ops
if self.account_failed:
ops += self.failed_rd_ops
if self.account_invalid:
ops += self.invalid_rd_ops
if flush:
ops += self.total_flush_ops
return ops
def accounted_latency(self, read = False, write = False, flush = False):
latency = 0
if write:
latency += self.total_wr_ops * op_latency
if self.account_failed:
latency += self.failed_wr_ops * op_latency
if read:
latency += self.total_rd_ops * op_latency
if self.account_failed:
latency += self.failed_rd_ops * op_latency
if flush:
latency += self.total_flush_ops * op_latency
return latency
def check_values(self):
stats = self.blockstats('drive0')
# Check that the totals match with what we have calculated
self.assertEqual(self.total_rd_bytes, stats['rd_bytes'])
self.assertEqual(self.total_wr_bytes, stats['wr_bytes'])
self.assertEqual(self.total_rd_ops, stats['rd_operations'])
self.assertEqual(self.total_wr_ops, stats['wr_operations'])
self.assertEqual(self.total_flush_ops, stats['flush_operations'])
self.assertEqual(self.wr_highest_offset, stats['wr_highest_offset'])
self.assertEqual(self.failed_rd_ops, stats['failed_rd_operations'])
self.assertEqual(self.failed_wr_ops, stats['failed_wr_operations'])
self.assertEqual(self.invalid_rd_ops, stats['invalid_rd_operations'])
self.assertEqual(self.invalid_wr_ops, stats['invalid_wr_operations'])
self.assertEqual(self.account_invalid, stats['account_invalid'])
self.assertEqual(self.account_failed, stats['account_failed'])
self.assertEqual(self.total_wr_merged, stats['wr_merged'])
# Check that there's exactly one interval with the length we defined
self.assertEqual(1, len(stats['timed_stats']))
timed_stats = stats['timed_stats'][0]
self.assertEqual(interval_length, timed_stats['interval_length'])
total_rd_latency = self.accounted_latency(read = True)
if (total_rd_latency != 0):
self.assertEqual(total_rd_latency, stats['rd_total_time_ns'])
self.assertEqual(op_latency, timed_stats['min_rd_latency_ns'])
self.assertEqual(op_latency, timed_stats['max_rd_latency_ns'])
self.assertEqual(op_latency, timed_stats['avg_rd_latency_ns'])
self.assertLess(0, timed_stats['avg_rd_queue_depth'])
else:
self.assertEqual(0, stats['rd_total_time_ns'])
self.assertEqual(0, timed_stats['min_rd_latency_ns'])
self.assertEqual(0, timed_stats['max_rd_latency_ns'])
self.assertEqual(0, timed_stats['avg_rd_latency_ns'])
self.assertEqual(0, timed_stats['avg_rd_queue_depth'])
# min read latency <= avg read latency <= max read latency
self.assertLessEqual(timed_stats['min_rd_latency_ns'],
timed_stats['avg_rd_latency_ns'])
self.assertLessEqual(timed_stats['avg_rd_latency_ns'],
timed_stats['max_rd_latency_ns'])
total_wr_latency = self.accounted_latency(write = True)
if (total_wr_latency != 0):
self.assertEqual(total_wr_latency, stats['wr_total_time_ns'])
self.assertEqual(op_latency, timed_stats['min_wr_latency_ns'])
self.assertEqual(op_latency, timed_stats['max_wr_latency_ns'])
self.assertEqual(op_latency, timed_stats['avg_wr_latency_ns'])
self.assertLess(0, timed_stats['avg_wr_queue_depth'])
else:
self.assertEqual(0, stats['wr_total_time_ns'])
self.assertEqual(0, timed_stats['min_wr_latency_ns'])
self.assertEqual(0, timed_stats['max_wr_latency_ns'])
self.assertEqual(0, timed_stats['avg_wr_latency_ns'])
self.assertEqual(0, timed_stats['avg_wr_queue_depth'])
# min write latency <= avg write latency <= max write latency
self.assertLessEqual(timed_stats['min_wr_latency_ns'],
timed_stats['avg_wr_latency_ns'])
self.assertLessEqual(timed_stats['avg_wr_latency_ns'],
timed_stats['max_wr_latency_ns'])
total_flush_latency = self.accounted_latency(flush = True)
if (total_flush_latency != 0):
self.assertEqual(total_flush_latency, stats['flush_total_time_ns'])
self.assertEqual(op_latency, timed_stats['min_flush_latency_ns'])
self.assertEqual(op_latency, timed_stats['max_flush_latency_ns'])
self.assertEqual(op_latency, timed_stats['avg_flush_latency_ns'])
else:
self.assertEqual(0, stats['flush_total_time_ns'])
self.assertEqual(0, timed_stats['min_flush_latency_ns'])
self.assertEqual(0, timed_stats['max_flush_latency_ns'])
self.assertEqual(0, timed_stats['avg_flush_latency_ns'])
# min flush latency <= avg flush latency <= max flush latency
self.assertLessEqual(timed_stats['min_flush_latency_ns'],
timed_stats['avg_flush_latency_ns'])
self.assertLessEqual(timed_stats['avg_flush_latency_ns'],
timed_stats['max_flush_latency_ns'])
# idle_time_ns must be > 0 if we have performed any operation
if (self.accounted_ops(read = True, write = True, flush = True) != 0):
self.assertLess(0, stats['idle_time_ns'])
else:
self.assertFalse(stats.has_key('idle_time_ns'))
# This test does not alter these, so they must be all 0
self.assertEqual(0, stats['rd_merged'])
self.assertEqual(0, stats['failed_flush_operations'])
self.assertEqual(0, stats['invalid_flush_operations'])
def do_test_stats(self, rd_size = 0, rd_ops = 0, wr_size = 0, wr_ops = 0,
flush_ops = 0, invalid_rd_ops = 0, invalid_wr_ops = 0,
failed_rd_ops = 0, failed_wr_ops = 0, wr_merged = 0):
# The 'ops' list will contain all the requested I/O operations
ops = []
for i in range(rd_ops):
ops.append("aio_read %d %d" % (i * rd_size, rd_size))
for i in range(wr_ops):
ops.append("aio_write %d %d" % (i * wr_size, wr_size))
for i in range(flush_ops):
ops.append("aio_flush")
highest_offset = wr_ops * wr_size
for i in range(invalid_rd_ops):
ops.append("aio_read -i 0 512")
for i in range(invalid_wr_ops):
ops.append("aio_write -i 0 512")
for i in range(failed_rd_ops):
ops.append("aio_read %d 512" % bad_offset)
for i in range(failed_wr_ops):
ops.append("aio_write %d 512" % bad_offset)
# We need an extra aio_flush to settle all outstanding AIO
# operations before we can advance the virtual clock, so that
# the last access happens before clock_step and idle_time_ns
# will be greater than 0
extra_flush = 0
if rd_ops + wr_ops + invalid_rd_ops + invalid_wr_ops + \
failed_rd_ops + failed_wr_ops > 0:
extra_flush = 1
if extra_flush > 0:
ops.append("aio_flush")
if failed_wr_ops > 0:
highest_offset = max(highest_offset, bad_offset + 512)
# Now perform all operations
for op in ops:
self.vm.hmp_qemu_io("drive0", op)
# Update the expected totals
self.total_rd_bytes += rd_ops * rd_size
self.total_rd_ops += rd_ops
self.total_wr_bytes += wr_ops * wr_size
self.total_wr_ops += wr_ops
self.total_wr_merged += wr_merged
self.total_flush_ops += flush_ops + extra_flush
self.invalid_rd_ops += invalid_rd_ops
self.invalid_wr_ops += invalid_wr_ops
self.failed_rd_ops += failed_rd_ops
self.failed_wr_ops += failed_wr_ops
self.wr_highest_offset = max(self.wr_highest_offset, highest_offset)
# Advance the clock so idle_time_ns has a meaningful value
self.vm.qtest("clock_step %d" % nsec_per_sec)
# And check that the actual statistics match the expected ones
self.check_values()
def test_read_only(self):
test_values = [[512, 1],
[65536, 1],
[512, 12],
[65536, 12]]
for i in test_values:
self.do_test_stats(rd_size = i[0], rd_ops = i[1])
def test_write_only(self):
test_values = [[512, 1],
[65536, 1],
[512, 12],
[65536, 12]]
for i in test_values:
self.do_test_stats(wr_size = i[0], wr_ops = i[1])
def test_invalid(self):
self.do_test_stats(invalid_rd_ops = 7)
self.do_test_stats(invalid_wr_ops = 3)
self.do_test_stats(invalid_rd_ops = 4, invalid_wr_ops = 5)
def test_failed(self):
self.do_test_stats(failed_rd_ops = 8)
self.do_test_stats(failed_wr_ops = 6)
self.do_test_stats(failed_rd_ops = 5, failed_wr_ops = 12)
def test_flush(self):
self.do_test_stats(flush_ops = 8)
def test_all(self):
# rd_size, rd_ops, wr_size, wr_ops, flush_ops
# invalid_rd_ops, invalid_wr_ops,
# failed_rd_ops, failed_wr_ops
# wr_merged
test_values = [[512, 1, 512, 1, 1, 4, 7, 5, 2, 0],
[65536, 1, 2048, 12, 7, 7, 5, 2, 5, 0],
[32768, 9, 8192, 1, 4, 3, 2, 4, 6, 0],
[16384, 11, 3584, 16, 9, 8, 6, 7, 3, 0]]
for i in test_values:
self.do_test_stats(*i)
def test_no_op(self):
# All values must be sane before doing any I/O
self.check_values()
class BlockDeviceStatsTestAccountInvalid(BlockDeviceStatsTestCase):
account_invalid = True
account_failed = False
class BlockDeviceStatsTestAccountFailed(BlockDeviceStatsTestCase):
account_invalid = False
account_failed = True
class BlockDeviceStatsTestAccountBoth(BlockDeviceStatsTestCase):
account_invalid = True
account_failed = True
class BlockDeviceStatsTestCoroutine(BlockDeviceStatsTestCase):
test_img = "null-co://"
if __name__ == '__main__':
iotests.main(supported_fmts=["raw"])