tor-browser

The Tor Browser
git clone https://git.dasho.dev/tor-browser.git
Log | Files | Refs | README | LICENSE

embed_sections.py (4860B)

      1 #!/usr/bin/env vpython3
      2 #
      3 # Copyright 2024 The Chromium Authors
      4 # Use of this source code is governed by a BSD-style license that can be
      5 # found in the LICENSE file.
      6 '''
      7 Embed the address and size of elf sections into the pre-defined symbols.
      8 
      9 The embedded values are used for performance optimization by mlock(2)ing the
     10 sections on ChromeOS. See chromeos/ash/components/memory/memory.cc for details.
     11 '''
     12 
     13 import argparse
     14 import os
     15 import subprocess
     16 import sys
     17 import shutil
     18 
     19 llvm_readelf = os.path.join(
     20    os.path.dirname(sys.argv[0]), '..', '..', 'third_party', 'llvm-build',
     21    'Release+Asserts', 'bin', 'llvm-readelf')
     22 
     23 TARGET_SECTIONS = {
     24    '.rodata': {
     25        'addr': 'kRodataAddr',
     26        'size': 'kRodataSize'
     27    },
     28    '.text.hot': {
     29        'addr': 'kTextHotAddr',
     30        'size': 'kTextHotSize'
     31    },
     32 }
     33 
     34 
     35 def parse_endianess(objdump_result):
     36  for line in objdump_result.splitlines():
     37    line = line.strip()
     38    if line.startswith('Data:'):
     39      if '1' in line:
     40        return 'big'
     41      if '2' in line:
     42        return 'little'
     43  raise ValueError('No endian found')
     44 
     45 
     46 def assert_elf_type(objdump_result):
     47  for line in objdump_result.splitlines():
     48    line = line.strip()
     49    if line.startswith('Class:'):
     50      if 'ELF64' in line or 'ELF32' in line:
     51        return
     52      raise ValueError('Class is not ELF64 nor ELF32: ' + line)
     53  raise ValueError('No class found')
     54 
     55 
     56 def parse_section_info(objdump_result, section_name):
     57  for line in objdump_result.splitlines():
     58    row = line.strip().split()
     59    if len(row) < 2:
     60      continue
     61    if row[1] == section_name:
     62      # 3: Address, 4: Offset, 5: Size
     63      return (int(row[3], base=16), int(row[4], base=16), int(row[5], base=16))
     64  return (0, 0, 0)
     65 
     66 
     67 def create_symbol_map(binary_input, objdump_result):
     68  (rodata_section_addr, rodata_section_offset,
     69   rodata_section_size) = parse_section_info(objdump_result, '.rodata')
     70 
     71  command = [llvm_readelf, '--symbols', binary_input]
     72  with subprocess.Popen(command, stdout=subprocess.PIPE, text=True) as process:
     73 
     74    variable_names = [
     75        var for maps in TARGET_SECTIONS.values() for var in maps.values()
     76    ]
     77    result = {}
     78    while len(result) < len(variable_names):
     79      line = process.stdout.readline()
     80      if not line:
     81        break
     82      for var in variable_names:
     83        if var in line:
     84          row = line.strip().split()
     85          if row[2] == '8':
     86            size = 8
     87          elif row[2] == '4':
     88            size = 4
     89          else:
     90            raise ValueError('variable size is not 8 or 4: ' + line)
     91          addr = int(row[1], base=16)
     92          rodata_section_end_addr = rodata_section_addr + rodata_section_size
     93          if addr < rodata_section_addr or addr >= rodata_section_end_addr:
     94            raise ValueError(var + ' is not in .rodata section')
     95          offset = addr - rodata_section_addr + rodata_section_offset
     96          result[var] = (offset, size)
     97 
     98  return result
     99 
    100 
    101 def overwrite_variable(file, symbol_map, endianess, section_name, var_type,
    102                       value):
    103  (var_offset, var_size) = symbol_map[TARGET_SECTIONS[section_name][var_type]]
    104  file.seek(var_offset)
    105  if file.write(
    106      value.to_bytes(length=var_size, byteorder=endianess,
    107                     signed=False)) != var_size:
    108    raise ValueError('failed to write value to file')
    109 
    110 
    111 def main():
    112  argparser = argparse.ArgumentParser(
    113      description='embed sections informataion into binary.')
    114 
    115  argparser.add_argument('--binary-input', help='exe file path.')
    116  argparser.add_argument('--binary-output', help='embedded file path.')
    117  args = argparser.parse_args()
    118 
    119  objdump_result = subprocess.run([llvm_readelf, '-e', args.binary_input],
    120                                  stdout=subprocess.PIPE,
    121                                  check=True,
    122                                  text=True).stdout
    123 
    124  assert_elf_type(objdump_result)
    125 
    126  symbol_map = create_symbol_map(args.binary_input, objdump_result)
    127  if len(symbol_map) != len(set(symbol_map.values())):
    128    raise ValueError(f'symbol_map overlaps: {symbol_map}')
    129 
    130  endianess = parse_endianess(objdump_result)
    131 
    132  shutil.copyfile(args.binary_input, args.binary_output)
    133 
    134  with open(args.binary_output, 'r+b') as file:
    135    for section_name in TARGET_SECTIONS:
    136      (addr, _, size) = parse_section_info(objdump_result, section_name)
    137      overwrite_variable(file, symbol_map, endianess, section_name, 'addr',
    138                         addr)
    139      overwrite_variable(file, symbol_map, endianess, section_name, 'size',
    140                         size)
    141 
    142  objdump_result_after = subprocess.run(
    143      [llvm_readelf, '-e', args.binary_output],
    144      stdout=subprocess.PIPE,
    145      check=True,
    146      text=True).stdout
    147  if objdump_result_after != objdump_result:
    148    raise ValueError('realelf result has changed')
    149 
    150  return 0
    151 
    152 
    153 if __name__ == '__main__':
    154  sys.exit(main())