1
0
mirror of https://github.com/ARM-software/devlib.git synced 2025-04-04 08:50:04 +01:00
devlib/devlib/trace/ftrace.py
Brendan Jackman 6cdae6bbe1 ftrace: Poke all CPUs for cpu_idle events before collecting trace
On some systems CPUs sometimes remain idle for several seconds. If a
trace capture begins during one of these long idle periods, that CPU's
idle state is unknown (in practice it is probably in its deepest
available state from cpuidle's perspective but that can't be known for
sure).

The solution to the equivalent problem for cpufreq is to read the
current frequencies from sysfs and inject artificial cpu_frequency
events using trace_marker (see cpu_freq_trace_all_frequencies in
bin/scripts/shutils.in). However we can't read the current idle state
from sysfs.

Instead, wake up each CPU by executing the "true" command on it via
taskset.
2016-11-08 17:00:23 +00:00

332 lines
14 KiB
Python

# Copyright 2015 ARM Limited
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
from __future__ import division
import os
import json
import time
import re
import subprocess
from devlib.trace import TraceCollector
from devlib.host import PACKAGE_BIN_DIRECTORY
from devlib.exception import TargetError, HostError
from devlib.utils.misc import check_output, which
TRACE_MARKER_START = 'TRACE_MARKER_START'
TRACE_MARKER_STOP = 'TRACE_MARKER_STOP'
OUTPUT_TRACE_FILE = 'trace.dat'
OUTPUT_PROFILE_FILE = 'trace_stat.dat'
DEFAULT_EVENTS = [
'cpu_frequency',
'cpu_idle',
'sched_migrate_task',
'sched_process_exec',
'sched_process_fork',
'sched_stat_iowait',
'sched_switch',
'sched_wakeup',
'sched_wakeup_new',
]
TIMEOUT = 180
# Regexps for parsing of function profiling data
CPU_RE = re.compile(r' Function \(CPU([0-9]+)\)')
STATS_RE = re.compile(r'([^ ]*) +([0-9]+) +([0-9.]+) us +([0-9.]+) us +([0-9.]+) us')
class FtraceCollector(TraceCollector):
def __init__(self, target,
events=None,
functions=None,
buffer_size=None,
buffer_size_step=1000,
tracing_path='/sys/kernel/debug/tracing',
automark=True,
autoreport=True,
autoview=False,
no_install=False,
strict=False,
):
super(FtraceCollector, self).__init__(target)
self.events = events if events is not None else DEFAULT_EVENTS
self.functions = functions
self.buffer_size = buffer_size
self.buffer_size_step = buffer_size_step
self.tracing_path = tracing_path
self.automark = automark
self.autoreport = autoreport
self.autoview = autoview
self.target_output_file = os.path.join(self.target.working_directory, OUTPUT_TRACE_FILE)
self.target_binary = None
self.host_binary = None
self.start_time = None
self.stop_time = None
self.event_string = None
self.function_string = None
self._reset_needed = True
# Setup tracing paths
self.available_events_file = self.target.path.join(self.tracing_path, 'available_events')
self.available_functions_file = self.target.path.join(self.tracing_path, 'available_filter_functions')
self.buffer_size_file = self.target.path.join(self.tracing_path, 'buffer_size_kb')
self.current_tracer_file = self.target.path.join(self.tracing_path, 'current_tracer')
self.function_profile_file = self.target.path.join(self.tracing_path, 'function_profile_enabled')
self.marker_file = self.target.path.join(self.tracing_path, 'trace_marker')
self.ftrace_filter_file = self.target.path.join(self.tracing_path, 'set_ftrace_filter')
self.host_binary = which('trace-cmd')
self.kernelshark = which('kernelshark')
if not self.target.is_rooted:
raise TargetError('trace-cmd instrument cannot be used on an unrooted device.')
if self.autoreport and self.host_binary is None:
raise HostError('trace-cmd binary must be installed on the host if autoreport=True.')
if self.autoview and self.kernelshark is None:
raise HostError('kernelshark binary must be installed on the host if autoview=True.')
if not no_install:
host_file = os.path.join(PACKAGE_BIN_DIRECTORY, self.target.abi, 'trace-cmd')
self.target_binary = self.target.install(host_file)
else:
if not self.target.is_installed('trace-cmd'):
raise TargetError('No trace-cmd found on device and no_install=True is specified.')
self.target_binary = 'trace-cmd'
# Validate required events to be traced
available_events = self.target.execute(
'cat {}'.format(self.available_events_file),
as_root=True).splitlines()
selected_events = []
for event in self.events:
# Convert globs supported by FTrace into valid regexp globs
_event = event
if event[0] != '*':
_event = '*' + event
event_re = re.compile(_event.replace('*', '.*'))
# Select events matching the required ones
if len(filter(event_re.match, available_events)) == 0:
message = 'Event [{}] not available for tracing'.format(event)
if strict:
raise TargetError(message)
self.target.logger.warning(message)
else:
selected_events.append(event)
# If function profiling is enabled we always need at least one event.
# Thus, if not other events have been specified, try to add at least
# a tracepoint which is always available and possibly triggered few
# times.
if self.functions and len(selected_events) == 0:
selected_events = ['sched_wakeup_new']
self.event_string = _build_trace_events(selected_events)
# Check for function tracing support
if self.functions:
if not self.target.file_exists(self.function_profile_file):
raise TargetError('Function profiling not supported. '\
'A kernel build with CONFIG_FUNCTION_PROFILER enable is required')
# Validate required functions to be traced
available_functions = self.target.execute(
'cat {}'.format(self.available_functions_file),
as_root=True).splitlines()
selected_functions = []
for function in self.functions:
if function not in available_functions:
message = 'Function [{}] not available for profiling'.format(function)
if strict:
raise TargetError(message)
self.target.logger.warning(message)
else:
selected_functions.append(function)
self.function_string = _build_trace_functions(selected_functions)
def reset(self):
if self.buffer_size:
self._set_buffer_size()
self.target.execute('{} reset'.format(self.target_binary),
as_root=True, timeout=TIMEOUT)
self._reset_needed = False
def start(self):
self.start_time = time.time()
if self._reset_needed:
self.reset()
self.target.execute('{} start {}'.format(self.target_binary, self.event_string),
as_root=True)
if self.automark:
self.mark_start()
if 'cpufreq' in self.target.modules:
self.logger.debug('Trace CPUFreq frequencies')
self.target.cpufreq.trace_frequencies()
if 'cpuidle' in self.target.modules:
self.logger.debug('Trace CPUIdle states')
self.target.cpuidle.perturb_cpus()
# Enable kernel function profiling
if self.functions:
self.target.execute('echo nop > {}'.format(self.current_tracer_file),
as_root=True)
self.target.execute('echo 0 > {}'.format(self.function_profile_file),
as_root=True)
self.target.execute('echo {} > {}'.format(self.function_string, self.ftrace_filter_file),
as_root=True)
self.target.execute('echo 1 > {}'.format(self.function_profile_file),
as_root=True)
def stop(self):
# Disable kernel function profiling
if self.functions:
self.target.execute('echo 1 > {}'.format(self.function_profile_file),
as_root=True)
if 'cpufreq' in self.target.modules:
self.logger.debug('Trace CPUFreq frequencies')
self.target.cpufreq.trace_frequencies()
self.stop_time = time.time()
if self.automark:
self.mark_stop()
self.target.execute('{} stop'.format(self.target_binary),
timeout=TIMEOUT, as_root=True)
self._reset_needed = True
def get_trace(self, outfile):
if os.path.isdir(outfile):
outfile = os.path.join(outfile, os.path.dirname(self.target_output_file))
self.target.execute('{} extract -o {}'.format(self.target_binary, self.target_output_file),
timeout=TIMEOUT, as_root=True)
# The size of trace.dat will depend on how long trace-cmd was running.
# Therefore timout for the pull command must also be adjusted
# accordingly.
pull_timeout = 5 * (self.stop_time - self.start_time)
self.target.pull(self.target_output_file, outfile, timeout=pull_timeout)
if not os.path.isfile(outfile):
self.logger.warning('Binary trace not pulled from device.')
else:
if self.autoreport:
textfile = os.path.splitext(outfile)[0] + '.txt'
self.report(outfile, textfile)
if self.autoview:
self.view(outfile)
def get_stats(self, outfile):
if not self.functions:
return
if os.path.isdir(outfile):
outfile = os.path.join(outfile, OUTPUT_PROFILE_FILE)
output = self.target._execute_util('ftrace_get_function_stats',
as_root=True)
function_stats = {}
for line in output.splitlines():
# Match a new CPU dataset
match = CPU_RE.search(line)
if match:
cpu_id = int(match.group(1))
function_stats[cpu_id] = {}
self.logger.debug("Processing stats for CPU%d...", cpu_id)
continue
# Match a new function dataset
match = STATS_RE.search(line)
if match:
fname = match.group(1)
function_stats[cpu_id][fname] = {
'hits' : int(match.group(2)),
'time' : float(match.group(3)),
'avg' : float(match.group(4)),
's_2' : float(match.group(5)),
}
self.logger.debug(" %s: %s",
fname, function_stats[cpu_id][fname])
self.logger.debug("FTrace stats output [%s]...", outfile)
with open(outfile, 'w') as fh:
json.dump(function_stats, fh, indent=4)
self.logger.debug("FTrace function stats save in [%s]", outfile)
return function_stats
def report(self, binfile, destfile):
# To get the output of trace.dat, trace-cmd must be installed
# This is done host-side because the generated file is very large
try:
command = '{} report {} > {}'.format(self.host_binary, binfile, destfile)
self.logger.debug(command)
process = subprocess.Popen(command, stderr=subprocess.PIPE, shell=True)
_, error = process.communicate()
if process.returncode:
raise TargetError('trace-cmd returned non-zero exit code {}'.format(process.returncode))
if error:
# logged at debug level, as trace-cmd always outputs some
# errors that seem benign.
self.logger.debug(error)
if os.path.isfile(destfile):
self.logger.debug('Verifying traces.')
with open(destfile) as fh:
for line in fh:
if 'EVENTS DROPPED' in line:
self.logger.warning('Dropped events detected.')
break
else:
self.logger.debug('Trace verified.')
else:
self.logger.warning('Could not generate trace.txt.')
except OSError:
raise HostError('Could not find trace-cmd. Please make sure it is installed and is in PATH.')
def view(self, binfile):
check_output('{} {}'.format(self.kernelshark, binfile), shell=True)
def teardown(self):
self.target.remove(self.target.path.join(self.target.working_directory, OUTPUT_TRACE_FILE))
def mark_start(self):
self.target.write_value(self.marker_file, TRACE_MARKER_START, verify=False)
def mark_stop(self):
self.target.write_value(self.marker_file, TRACE_MARKER_STOP, verify=False)
def _set_buffer_size(self):
target_buffer_size = self.buffer_size
attempt_buffer_size = target_buffer_size
buffer_size = 0
floor = 1000 if target_buffer_size > 1000 else target_buffer_size
while attempt_buffer_size >= floor:
self.target.write_value(self.buffer_size_file, attempt_buffer_size, verify=False)
buffer_size = self.target.read_int(self.buffer_size_file)
if buffer_size == attempt_buffer_size:
break
else:
attempt_buffer_size -= self.buffer_size_step
if buffer_size == target_buffer_size:
return
while attempt_buffer_size < target_buffer_size:
attempt_buffer_size += self.buffer_size_step
self.target.write_value(self.buffer_size_file, attempt_buffer_size, verify=False)
buffer_size = self.target.read_int(self.buffer_size_file)
if attempt_buffer_size != buffer_size:
message = 'Failed to set trace buffer size to {}, value set was {}'
self.logger.warning(message.format(target_buffer_size, buffer_size))
break
def _build_trace_events(events):
event_string = ' '.join(['-e {}'.format(e) for e in events])
return event_string
def _build_trace_functions(functions):
function_string = " ".join(functions)
return function_string