1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
|
#!/usr/bin/env python3
# Copyright 2021 The Chromium Authors
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.
from collections import defaultdict
import argparse
import itertools
import json
import logging
import os
import re
import sys
import typing
"""This module contains the utilities necessary to read Dtrace result files and
convert them other format for flamegraph generation, such as pprof profiles.
"""
from protos.third_party.pprof import profile_pb2
def pairwise(iterable):
"s -> (s0,s1), (s1,s2), (s2, s3), ..."
a, b = itertools.tee(iterable)
next(b, None)
return zip(a, b)
class ProfileBuilder:
"""Helper to generate a pprof profile."""
def __init__(self):
self._profile = profile_pb2.Profile()
self._locations = {}
self._profile.string_table.append("")
self._strings = {"": 0}
self._signature_id = self.GetStringId('signature')
def GetStringId(self, string: str) -> int:
"""Returns the id for `string` into the string_table, creating new entry
if not already present.
"""
if string in self._strings:
return self._strings[string]
index = len(self._profile.string_table)
self._strings[string] = index
self._profile.string_table.append(string)
return index
def GetSymbolLocation(self, name: str, system_name: str) -> int:
"""Returns the id for a symbol location defined as
`(name, system_name)`, creating new entry if not already present.
"""
if (name, system_name) in self._locations:
return self._locations[(name, system_name)]
function_id = len(self._profile.location) + 1
self._locations[(name, system_name)] = function_id
function = self._profile.function.add()
function.id = function_id
function.name = self.GetStringId(name)
function.system_name = self.GetStringId(system_name)
# These fields are given default values since they aren't stored
# in dtrace.
function.filename = self.GetStringId("")
function.start_line = 0
location = self._profile.location.add()
location.id = function_id
line = location.line.add()
line.function_id = function_id
return function_id
def AddComment(self, comment: str):
self._profile.comment.append(self.GetStringId(comment))
def AddSampleType(self, type: str, unit: str):
"""
Adds a sample type that describe stackframes in this profile.
See protos/third_party/pprof/src/profile.proto for more details.
"""
assert (len(self._profile.sample) == 0)
sample_type = self._profile.sample_type.add()
sample_type.type = self.GetStringId(type)
sample_type.unit = self.GetStringId(unit)
def AddSample(self, locations: typing.List[int], values: typing.List[int],
signature: str):
"""
Adds a sample in the profile, constructed from a list of locations
representing the stack, and a list of values for that stack (as many values
as sample types described by this profile).
"""
assert (len(self._profile.sample_type) == len(values))
sample = self._profile.sample.add()
for value in values:
sample.value.append(value)
for location in locations:
sample.location_id.append(location)
label = sample.label.add()
label.key = self._signature_id
label.str = self.GetStringId(signature)
def SerializeToString(self) -> str:
return self._profile.SerializeToString()
class DTraceParser:
"""Parses and merges chromium Dtrace profiles.
Typical usage example:
parser = DTraceParser()
parser.ParseDir('./samples/')
parser.ExportToPprof(builder)
"""
def __init__(self, sample_type: str = 'cpu_time'):
"""
Args:
output_filename: The path of the file in which results are written.
"""
self._stack_weights = defaultdict(int)
self._signatures = defaultdict(str)
self._stack_frames = {}
self._sample_type = sample_type
self._post_processing_applied = False
def ParseFile(self, stack_file: typing.TextIO):
"""Parses dtrace `stack_file` and adds the data to this profile.
"""
assert (self._post_processing_applied == False)
stack_frames = []
for line, next_line in pairwise(stack_file):
line_content = line.strip()
if not line_content:
continue
# If the next line is non-empty it's not the last in the stack.
if next_line.strip():
# Matches lines like: "0x17e018987e" or "+0x17e018987e"
if line_content.startswith("0x") or line_content.startswith(
"+0x"):
function = line_content
module = "unsymbolized module"
else:
module, function = line_content.split('`', 1)
# Matches lines with offset like: "module`function+0xf6"
if len(function.split('+0x')) == 2:
[function, offset] = function.split('+0x')
stack_frames.append((module, function))
else:
if len(stack_frames) == 0:
continue
weight = int(line_content)
stack_string = ";".join([
f'{module}`{function}' for (module,
function) in stack_frames
])
self._stack_weights[stack_string] += weight
self._stack_frames[stack_string] = stack_frames
stack_frames = []
if not self._stack_frames:
logging.error("No results found, check directory contents")
sys.exit(-1)
def ParseDir(self, stack_dir: str):
"""Parses all dtrace files in `stack_dir` and adds the data to this
profile.
Args:
stack_dir: The directory where Dtrace profile results can be found.
Raises:
SystemExit: When no results are found in stack_dir.
"""
# Define the format for DTrace stacks filenames.
stack_filename_regex = re.compile('[0-9]*_[0-9]*.txt')
# Treat all files that respect the name stack_filename_regex as DTrace
# stacks.
for root, dirs, files in os.walk(stack_dir):
for stack_filename in files:
if stack_filename_regex.match(stack_filename):
logging.info(f"Processing {stack_filename} ...")
with open(
os.path.join(root, stack_filename),
newline='',
encoding="ISO-8859-1") as stack_file:
self.ParseFile(stack_file)
if not self._stack_frames:
logging.error("No results found, check directory contents")
sys.exit(-1)
def ConvertToPprof(self, profile_builder: ProfileBuilder):
"""Converts this profile to pprof by writing to
`profile_builder`.
"""
profile_builder.AddSampleType(self._sample_type, "counts")
for key in self._stack_frames:
frames = self._stack_frames[key]
weight = self._stack_weights[key]
signature = self._signatures[key]
sample_locations = []
for (module, function) in frames:
sample_locations.append(
profile_builder.GetSymbolLocation(function, module))
profile_builder.AddSample(sample_locations, [weight], signature)
def ConvertToCollapse(self, output_filename: str):
"""Converts this profile to the "collapsed stack" format. In contrast
to the Dtrace format full stacks are writtent on a single line. At
first the different are separated by semi-colons and a space separates
the weight associated with the function.
Example:
base::foo;content::bar;biz::baz 17
base::biz;content::boo;biz::bim 23
...
"""
os.makedirs(
f"{os.path.dirname(os.path.abspath(output_filename))}",
exist_ok=True)
with open(output_filename, 'w') as f:
for key in self._stack_frames:
frames = self._stack_frames[key]
frames_string = ';'.join(
[function for (module, function) in reversed(frames)])
weight = self._stack_weights[key]
# Reform the line in stacked format and write it out.
f.write(f"{frames_string} {weight}\n")
def GetSamplesListForTesting(self):
samples = []
for key in self._stack_frames:
frames = self._stack_frames[key]
weight = self._stack_weights[key]
samples.append({"frames": frames, "weight": weight})
return samples
def AddSamplesForTesting(self, samples):
for sample in samples:
stack_frames = sample['frames']
stack_string = ";".join(
[f'{module}`{function}' for (module, function) in stack_frames])
self._stack_frames[stack_string] = stack_frames
self._stack_weights[stack_string] += sample['weight']
def ApplySignatures(self, stack: typing.List[typing.Tuple[str, str]]):
"""Matches and return known signatures to given stackframe.
"""
if len(stack) >= 512:
return '_OVERFLOWED_'
for module, function in stack:
if function.startswith('safe_browsing::(anonymous namespace)::'
'PlaybackOnBackgroundThread'):
return 'safe_browsing:VisualSignatures'
if function.startswith(
'safe_browsing::(anonymous namespace)::OnModelInputCreated'
):
return 'safe_browsing:VisualSignatures'
if function.startswith('ParkableStringImpl::CompressInBackground'):
return 'ParkableString'
return 'unknown'
def PostProcessStackSamples(self):
"""Applies filtering and enhancing to self.samples(). This function can
only be called once.
Raises:
SystemExit: If this function is called twice on the same object.
"""
if self._post_processing_applied:
logging.error("Post processing cannot be applied twice")
sys.exit(-1)
self._post_processing_applied = True
for key in self._stack_frames:
# Signatures are always added since they are non destructive.
self._signatures[key] = self.ApplySignatures(
self._stack_frames[key])
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description='Export DTrace stack files into another format.')
parser.add_argument(
"--data_dir",
help="Top level directory that contains DTrace stacks. "
"The directory will be fully walked to find stacks "
"and metadata.json files",
required=True)
parser.add_argument(
"--output", help="The file to write the collapsed stacks into.")
parser.add_argument(
'--unit',
dest='unit',
choices=["cpu_samples", "wakeups"],
default="cpu_samples",
help="The unit of counts acquired with DTrace")
parser.add_argument(
'--format',
dest='format',
action='store',
choices=["pprof", "collapsed"],
default="pprof",
help="Output format to generate.")
args = parser.parse_args()
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.INFO)
# Traverse |data_dir| and concatenate all metadata.json into one big comment
# for the pprof.
full_comment = ""
metadata_filename_regex = re.compile('metadata.json')
for root, dirs, files in os.walk(args.data_dir):
for file in files:
if metadata_filename_regex.match(file):
with open(os.path.join(root, file)) as meta_json:
for line in meta_json:
full_comment += line
parser = DTraceParser(args.unit)
parser.ParseDir(args.data_dir)
parser.PostProcessStackSamples()
output_filename = args.output
data_dir = os.path.abspath(os.path.join(args.data_dir, os.pardir))
if args.format == "pprof":
profile_builder = ProfileBuilder()
profile_builder.AddComment(full_comment)
profile_builder.AddComment(f"Unit : {args.unit}")
parser.ConvertToPprof(profile_builder)
if output_filename is None:
output_filename = os.path.join(data_dir, f"profile_{args.unit}.pb")
with open(output_filename, "wb") as output_file:
output_file.write(profile_builder.SerializeToString())
else:
if output_filename is None:
output_filename = os.path.join(data_dir,
f"profile_{args.unit}.collapsed")
parser.ConvertToCollapse(output_filename)
logging.info(f'Outputing profile in {os.path.abspath(output_filename)}')
|