tools/cygprofile/symbol_extractor.py

   1 #!/usr/bin/python
   2 # Copyright 2015 The Chromium Authors. All rights reserved.
   3 # Use of this source code is governed by a BSD-style license that can be
   4 # found in the LICENSE file.
   5
   6 """Utilities to get and manipulate symbols from a binary."""
   7
   8 import collections
   9 import logging
  10 import os
  11 import re
  12 import subprocess
  13 import sys
  14
  15 sys.path.insert(
  16     0, os.path.join(os.path.dirname(__file__), os.pardir, os.pardir,
  17                     'third_party', 'android_platform', 'development',
  18                     'scripts'))
  19 import symbol
  20
  21 _MAX_WARNINGS_TO_PRINT = 200
  22
  23 SymbolInfo = collections.namedtuple('SymbolInfo', ('name', 'offset', 'size',
  24                                                    'section'))
  25
  26 def SetArchitecture(arch):
  27   """Set the architecture for binaries to be symbolized."""
  28   symbol.ARCH = arch
  29
  30
  31 def _FromObjdumpLine(line):
  32   """Create a SymbolInfo by parsing a properly formatted objdump output line.
  33
  34   Args:
  35     line: line from objdump
  36
  37   Returns:
  38     An instance of SymbolInfo if the line represents a symbol, None otherwise.
  39   """
  40   # All of the symbol lines we care about are in the form
  41   # 0000000000  g    F   .text.foo     000000000 [.hidden] foo
  42   # where g (global) might also be l (local) or w (weak).
  43   parts = line.split()
  44   if len(parts) < 6 or parts[2] != 'F':
  45     return None
  46
  47   assert len(parts) == 6 or (len(parts) == 7 and parts[5] == '.hidden')
  48   accepted_scopes = set(['g', 'l', 'w'])
  49   assert parts[1] in accepted_scopes
  50
  51   offset = int(parts[0], 16)
  52   section = parts[3]
  53   size = int(parts[4], 16)
  54   name = parts[-1].rstrip('\n')
  55   assert re.match('^[a-zA-Z0-9_.]+$', name)
  56   return SymbolInfo(name=name, offset=offset, section=section, size=size)
  57
  58
  59 def _SymbolInfosFromStream(objdump_lines):
  60   """Parses the output of objdump, and get all the symbols from a binary.
  61
  62   Args:
  63     objdump_lines: An iterable of lines
  64
  65   Returns:
  66     A list of SymbolInfo.
  67   """
  68   symbol_infos = []
  69   for line in objdump_lines:
  70     symbol_info = _FromObjdumpLine(line)
  71     if symbol_info is not None:
  72       symbol_infos.append(symbol_info)
  73   return symbol_infos
  74
  75
  76 def SymbolInfosFromBinary(binary_filename):
  77   """Runs objdump to get all the symbols from a binary.
  78
  79   Args:
  80     binary_filename: path to the binary.
  81
  82   Returns:
  83     A list of SymbolInfo from the binary.
  84   """
  85   command = (symbol.ToolPath('objdump'), '-t', '-w', binary_filename)
  86   p = subprocess.Popen(command, shell=False, stdout=subprocess.PIPE)
  87   try:
  88     result = _SymbolInfosFromStream(p.stdout)
  89     return result
  90   finally:
  91     p.wait()
  92
  93
  94 def GroupSymbolInfosByOffset(symbol_infos):
  95   """Create a dict {offset: [symbol_info1, ...], ...}.
  96
  97   As several symbols can be at the same offset, this is a 1-to-many
  98   relationship.
  99
 100   Args:
 101     symbol_infos: iterable of SymbolInfo instances
 102
 103   Returns:
 104     a dict {offset: [symbol_info1, ...], ...}
 105   """
 106   offset_to_symbol_infos = collections.defaultdict(list)
 107   for symbol_info in symbol_infos:
 108     offset_to_symbol_infos[symbol_info.offset].append(symbol_info)
 109   return dict(offset_to_symbol_infos)
 110
 111 def GroupSymbolInfosByName(symbol_infos):
 112   """Create a dict {name: [symbol_info1, ...], ...}.
 113
 114   A symbol can have several offsets, this is a 1-to-many relationship.
 115
 116   Args:
 117     symbol_infos: iterable of SymbolInfo instances
 118
 119   Returns:
 120     a dict {name: [symbol_info1, ...], ...}
 121   """
 122   name_to_symbol_infos = collections.defaultdict(list)
 123   for symbol_info in symbol_infos:
 124     name_to_symbol_infos[symbol_info.name].append(symbol_info)
 125   return dict(name_to_symbol_infos)
 126
 127 def CreateNameToSymbolInfo(symbol_infos):
 128   """Create a dict {name: symbol_info, ...}.
 129
 130   Args:
 131     symbol_infos: iterable of SymbolInfo instances
 132
 133   Returns:
 134     a dict {name: symbol_info, ...}
 135     If a symbol name corresponds to more than one symbol_info, the symbol_info
 136     with the lowest offset is chosen.
 137   """
 138   #TODO(azarchs): move the functionality in this method into check_orderfile.
 139   symbol_infos_by_name = {}
 140   collision_count = 0
 141   for infos in GroupSymbolInfosByName(symbol_infos).itervalues():
 142     first_symbol_info = min(infos, key=lambda x:x.offset)
 143     symbol_infos_by_name[first_symbol_info.name] = first_symbol_info
 144     if len(infos) > 1:
 145       collision_count += 1
 146       if collision_count <= _MAX_WARNINGS_TO_PRINT:
 147         logging.warning('Symbol %s appears at %d offsets: %s' %
 148                         (first_symbol_info.name,
 149                          len(infos),
 150                          ','.join([hex(x.offset) for x in infos])))
 151   if collision_count > _MAX_WARNINGS_TO_PRINT:
 152     logging.warning('%d symbols at multiple offsets.  First %d shown.' %
 153                     (collision_count, _MAX_WARNINGS_TO_PRINT))
 154   return symbol_infos_by_name