scripts/update_payload/checker.py

# Copyright (c) 2013 The Chromium OS Authors. All rights reserved.
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.

"""Verifying the integrity of a Chrome OS update payload.

This module is used internally by the main Payload class for verifying the
integrity of an update payload. The interface for invoking the checks is as
follows:

  checker = PayloadChecker(payload)
  checker.Run(...)

"""

import array
import base64
import hashlib
import subprocess

import common
from error import PayloadError
import format_utils
import histogram
import update_metadata_pb2


#
# Constants / helper functions.
#
_TYPE_FULL = 'full'
_TYPE_DELTA = 'delta'

_DEFAULT_BLOCK_SIZE = 4096


#
# Helper functions.
#
def _IsPowerOfTwo(val):
  """Returns True iff val is a power of two."""
  return val > 0 and (val & (val - 1)) == 0


def _AddFormat(format_func, value):
  """Adds a custom formatted representation to ordinary string representation.

  Args:
    format_func: a value formatter
    value: value to be formatted and returned
  Returns:
    A string 'x (y)' where x = str(value) and y = format_func(value).

  """
  return '%s (%s)' % (value, format_func(value))


def _AddHumanReadableSize(size):
  """Adds a human readable representation to a byte size value."""
  return _AddFormat(format_utils.BytesToHumanReadable, size)


#
# Payload report generator.
#
class _PayloadReport(object):
  """A payload report generator.

  A report is essentially a sequence of nodes, which represent data points. It
  is initialized to have a "global", untitled section. A node may be a
  sub-report itself.

  """

  # Report nodes: field, sub-report, section.
  class Node(object):
    """A report node interface."""

    @staticmethod
    def _Indent(indent, line):
      """Indents a line by a given indentation amount.

      Args:
        indent: the indentation amount
        line: the line content (string)
      Returns:
        The properly indented line (string).

      """
      return '%*s%s' % (indent, '', line)

    def GenerateLines(self, base_indent, sub_indent, curr_section):
      """Generates the report lines for this node.

      Args:
        base_indent: base indentation for each line
        sub_indent: additional indentation for sub-nodes
        curr_section: the current report section object
      Returns:
        A pair consisting of a list of properly indented report lines and a new
        current section object.

      """
      raise NotImplementedError()

  class FieldNode(Node):
    """A field report node, representing a (name, value) pair."""

    def __init__(self, name, value, linebreak, indent):
      super(_PayloadReport.FieldNode, self).__init__()
      self.name = name
      self.value = value
      self.linebreak = linebreak
      self.indent = indent

    def GenerateLines(self, base_indent, sub_indent, curr_section):
      """Generates a properly formatted 'name : value' entry."""
      report_output = ''
      if self.name:
        report_output += self.name.ljust(curr_section.max_field_name_len) + ' :'
      value_lines = str(self.value).splitlines()
      if self.linebreak and self.name:
        report_output += '\n' + '\n'.join(
            ['%*s%s' % (self.indent, '', line) for line in value_lines])
      else:
        if self.name:
          report_output += ' '
        report_output += '%*s' % (self.indent, '')
        cont_line_indent = len(report_output)
        indented_value_lines = [value_lines[0]]
        indented_value_lines.extend(['%*s%s' % (cont_line_indent, '', line)
                                     for line in value_lines[1:]])
        report_output += '\n'.join(indented_value_lines)

      report_lines = [self._Indent(base_indent, line + '\n')
                      for line in report_output.split('\n')]
      return report_lines, curr_section

  class SubReportNode(Node):
    """A sub-report node, representing a nested report."""

    def __init__(self, title, report):
      super(_PayloadReport.SubReportNode, self).__init__()
      self.title = title
      self.report = report

    def GenerateLines(self, base_indent, sub_indent, curr_section):
      """Recurse with indentation."""
      report_lines = [self._Indent(base_indent, self.title + ' =>\n')]
      report_lines.extend(self.report.GenerateLines(base_indent + sub_indent,
                                                    sub_indent))
      return report_lines, curr_section

  class SectionNode(Node):
    """A section header node."""

    def __init__(self, title=None):
      super(_PayloadReport.SectionNode, self).__init__()
      self.title = title
      self.max_field_name_len = 0

    def GenerateLines(self, base_indent, sub_indent, curr_section):
      """Dump a title line, return self as the (new) current section."""
      report_lines = []
      if self.title:
        report_lines.append(self._Indent(base_indent,
                                         '=== %s ===\n' % self.title))
      return report_lines, self

  def __init__(self):
    self.report = []
    self.last_section = self.global_section = self.SectionNode()
    self.is_finalized = False

  def GenerateLines(self, base_indent, sub_indent):
    """Generates the lines in the report, properly indented.

    Args:
      base_indent: the indentation used for root-level report lines
      sub_indent: the indentation offset used for sub-reports
    Returns:
      A list of indented report lines.

    """
    report_lines = []
    curr_section = self.global_section
    for node in self.report:
      node_report_lines, curr_section = node.GenerateLines(
          base_indent, sub_indent, curr_section)
      report_lines.extend(node_report_lines)

    return report_lines

  def Dump(self, out_file, base_indent=0, sub_indent=2):
    """Dumps the report to a file.

    Args:
      out_file: file object to output the content to
      base_indent: base indentation for report lines
      sub_indent: added indentation for sub-reports

    """

    report_lines = self.GenerateLines(base_indent, sub_indent)
    if report_lines and not self.is_finalized:
      report_lines.append('(incomplete report)\n')

    for line in report_lines:
      out_file.write(line)

  def AddField(self, name, value, linebreak=False, indent=0):
    """Adds a field/value pair to the payload report.

    Args:
      name: the field's name
      value: the field's value
      linebreak: whether the value should be printed on a new line
      indent: amount of extra indent for each line of the value

    """
    assert not self.is_finalized
    if name and self.last_section.max_field_name_len < len(name):
      self.last_section.max_field_name_len = len(name)
    self.report.append(self.FieldNode(name, value, linebreak, indent))

  def AddSubReport(self, title):
    """Adds and returns a sub-report with a title."""
    assert not self.is_finalized
    sub_report = self.SubReportNode(title, type(self)())
    self.report.append(sub_report)
    return sub_report.report

  def AddSection(self, title):
    """Adds a new section title."""
    assert not self.is_finalized
    self.last_section = self.SectionNode(title)
    self.report.append(self.last_section)

  def Finalize(self):
    """Seals the report, marking it as complete."""
    self.is_finalized = True


#
# Payload verification.
#
class PayloadChecker(object):
  """Checking the integrity of an update payload.

  This is a short-lived object whose purpose is to isolate the logic used for
  verifying the integrity of an update payload.

  """

  def __init__(self, payload):
    assert payload.is_init, 'uninitialized update payload'
    self.payload = payload

    # Reset state; these will be assigned when the manifest is checked.
    self.block_size = _DEFAULT_BLOCK_SIZE
    self.sigs_offset = 0
    self.sigs_size = 0
    self.old_rootfs_size = 0
    self.old_kernel_size = 0
    self.new_rootfs_size = 0
    self.new_kernel_size = 0
    self.payload_type = None

  @staticmethod
  def _CheckElem(msg, name, report, is_mandatory, is_submsg, convert=str,
                 msg_name=None, linebreak=False, indent=0):
    """Adds an element from a protobuf message to the payload report.

    Checks to see whether a message contains a given element, and if so adds
    the element value to the provided report. A missing mandatory element
    causes an exception to be raised.

    Args:
      msg: the message containing the element
      name: the name of the element
      report: a report object to add the element name/value to
      is_mandatory: whether or not this element must be present
      is_submsg: whether this element is itself a message
      convert: a function for converting the element value for reporting
      msg_name: the name of the message object (for error reporting)
      linebreak: whether the value report should induce a line break
      indent: amount of indent used for reporting the value
    Returns:
      A pair consisting of the element value and the generated sub-report for
      it (if the element is a sub-message, None otherwise). If the element is
      missing, returns (None, None).
    Raises:
      PayloadError if a mandatory element is missing.

    """
    if not msg.HasField(name):
      if is_mandatory:
        raise PayloadError("%smissing mandatory %s '%s'" %
                           (msg_name + ' ' if msg_name else '',
                            'sub-message' if is_submsg else 'field',
                            name))
      return (None, None)

    value = getattr(msg, name)
    if is_submsg:
      return (value, report and report.AddSubReport(name))
    else:
      if report:
        report.AddField(name, convert(value), linebreak=linebreak,
                        indent=indent)
      return (value, None)

  @staticmethod
  def _CheckMandatoryField(msg, field_name, report, msg_name, convert=str,
                           linebreak=False, indent=0):
    """Adds a mandatory field; returning first component from _CheckElem."""
    return PayloadChecker._CheckElem(msg, field_name, report, True, False,
                                     convert=convert, msg_name=msg_name,
                                     linebreak=linebreak, indent=indent)[0]

  @staticmethod
  def _CheckOptionalField(msg, field_name, report, convert=str,
                          linebreak=False, indent=0):
    """Adds an optional field; returning first component from _CheckElem."""
    return PayloadChecker._CheckElem(msg, field_name, report, False, False,
                                     convert=convert, linebreak=linebreak,
                                     indent=indent)[0]

  @staticmethod
  def _CheckMandatorySubMsg(msg, submsg_name, report, msg_name):
    """Adds a mandatory sub-message; wrapper for _CheckElem."""
    return PayloadChecker._CheckElem(msg, submsg_name, report, True, True,
                                     msg_name)

  @staticmethod
  def _CheckOptionalSubMsg(msg, submsg_name, report):
    """Adds an optional sub-message; wrapper for _CheckElem."""
    return PayloadChecker._CheckElem(msg, submsg_name, report, False, True)

  @staticmethod
  def _CheckPresentIff(val1, val2, name1, name2, obj_name):
    """Checks that val1 is None iff val2 is None.

    Args:
      val1: first value to be compared
      val2: second value to be compared
      name1: name of object holding the first value
      name2: name of object holding the second value
      obj_name: name of the object containing these values
    Raises:
      PayloadError if assertion does not hold.

    """
    if None in (val1, val2) and val1 is not val2:
      present, missing = (name1, name2) if val2 is None else (name2, name1)
      raise PayloadError("'%s' present without '%s'%s" %
                         (present, missing,
                          ' in ' + obj_name if obj_name else ''))

  @staticmethod
  def _Run(cmd, send_data=None):
    """Runs a subprocess, returns its output.

    Args:
      cmd: list of command-line argument for invoking the subprocess
      send_data: data to feed to the process via its stdin
    Returns:
      A tuple containing the stdout and stderr output of the process.

    """
    run_process = subprocess.Popen(cmd, stdin=subprocess.PIPE,
                                   stdout=subprocess.PIPE)
    return run_process.communicate(input=send_data)

  @staticmethod
  def _CheckSha256Signature(sig_data, pubkey_file_name, actual_hash, sig_name):
    """Verifies an actual hash against a signed one.

    Args:
      sig_data: the raw signature data
      pubkey_file_name: public key used for verifying signature
      actual_hash: the actual hash digest
      sig_name: signature name for error reporting
    Raises:
      PayloadError if signature could not be verified.

    """
    if len(sig_data) != 256:
      raise PayloadError('%s: signature size (%d) not as expected (256)' %
                         (sig_name, len(sig_data)))
    signed_data, _ = PayloadChecker._Run(
        ['openssl', 'rsautl', '-verify', '-pubin', '-inkey', pubkey_file_name],
        send_data=sig_data)

    if len(signed_data) != len(common.SIG_ASN1_HEADER) + 32:
      raise PayloadError('%s: unexpected signed data length (%d)' %
                         (sig_name, len(signed_data)))

    if not signed_data.startswith(common.SIG_ASN1_HEADER):
      raise PayloadError('%s: not containing standard ASN.1 prefix' % sig_name)

    signed_hash = signed_data[len(common.SIG_ASN1_HEADER):]
    if signed_hash != actual_hash:
      raise PayloadError('%s: signed hash (%s) different from actual (%s)' %
                         (sig_name, signed_hash.encode('hex'),
                          actual_hash.encode('hex')))

  @staticmethod
  def _CheckBlocksFitLength(length, num_blocks, block_size, length_name,
                            block_name=None):
    """Checks that a given length fits given block space.

    This ensures that the number of blocks allocated is appropriate for the
    length of the data residing in these blocks.

    Args:
      length: the actual length of the data
      num_blocks: the number of blocks allocated for it
      block_size: the size of each block in bytes
      length_name: name of length (used for error reporting)
      block_name: name of block (used for error reporting)
    Raises:
      PayloadError if the aforementioned invariant is not satisfied.

    """
    # Check: length <= num_blocks * block_size.
    if not length <= num_blocks * block_size:
      raise PayloadError(
          '%s (%d) > num %sblocks (%d) * block_size (%d)' %
          (length_name, length, block_name or '', num_blocks, block_size))

    # Check: length > (num_blocks - 1) * block_size.
    if not length > (num_blocks - 1) * block_size:
      raise PayloadError(
          '%s (%d) <= (num %sblocks - 1 (%d)) * block_size (%d)' %
          (length_name, length, block_name or '', num_blocks - 1, block_size))

  def _CheckManifest(self, report):
    """Checks the payload manifest.

    Args:
      report: a report object to add to
    Returns:
      A tuple consisting of the partition block size used during the update
      (integer), the signatures block offset and size.
    Raises:
      PayloadError if any of the checks fail.

    """
    manifest = self.payload.manifest
    report.AddSection('manifest')

    # Check: block_size must exist and match the expected value.
    actual_block_size = self._CheckMandatoryField(manifest, 'block_size',
                                                  report, 'manifest')
    if actual_block_size != self.block_size:
      raise PayloadError('block_size (%d) not as expected (%d)' %
                         (actual_block_size, self.block_size))

    # Check: signatures_offset <==> signatures_size.
    self.sigs_offset = self._CheckOptionalField(manifest, 'signatures_offset',
                                                report)
    self.sigs_size = self._CheckOptionalField(manifest, 'signatures_size',
                                              report)
    self._CheckPresentIff(self.sigs_offset, self.sigs_size,
                          'signatures_offset', 'signatures_size', 'manifest')

    # Check: old_kernel_info <==> old_rootfs_info.
    oki_msg, oki_report = self._CheckOptionalSubMsg(manifest,
                                                    'old_kernel_info', report)
    ori_msg, ori_report = self._CheckOptionalSubMsg(manifest,
                                                    'old_rootfs_info', report)
    self._CheckPresentIff(oki_msg, ori_msg, 'old_kernel_info',
                          'old_rootfs_info', 'manifest')
    if oki_msg:  # equivalently, ori_msg
      # Assert/mark delta payload.
      if self.payload_type == _TYPE_FULL:
        raise PayloadError(
            'apparent full payload contains old_{kernel,rootfs}_info')
      self.payload_type = _TYPE_DELTA

      # Check: {size, hash} present in old_{kernel,rootfs}_info.
      self.old_kernel_size = self._CheckMandatoryField(
          oki_msg, 'size', oki_report, 'old_kernel_info')
      self._CheckMandatoryField(oki_msg, 'hash', oki_report, 'old_kernel_info',
                                convert=common.FormatSha256)
      self.old_rootfs_size = self._CheckMandatoryField(
          ori_msg, 'size', ori_report, 'old_rootfs_info')
      self._CheckMandatoryField(ori_msg, 'hash', ori_report, 'old_rootfs_info',
                                convert=common.FormatSha256)
    else:
      # Assert/mark full payload.
      if self.payload_type == _TYPE_DELTA:
        raise PayloadError(
            'apparent delta payload missing old_{kernel,rootfs}_info')
      self.payload_type = _TYPE_FULL

    # Check: new_kernel_info present; contains {size, hash}.
    nki_msg, nki_report = self._CheckMandatorySubMsg(
        manifest, 'new_kernel_info', report, 'manifest')
    self.new_kernel_size = self._CheckMandatoryField(
        nki_msg, 'size', nki_report, 'new_kernel_info')
    self._CheckMandatoryField(nki_msg, 'hash', nki_report, 'new_kernel_info',
                              convert=common.FormatSha256)

    # Check: new_rootfs_info present; contains {size, hash}.
    nri_msg, nri_report = self._CheckMandatorySubMsg(
        manifest, 'new_rootfs_info', report, 'manifest')
    self.new_rootfs_size = self._CheckMandatoryField(
        nri_msg, 'size', nri_report, 'new_rootfs_info')
    self._CheckMandatoryField(nri_msg, 'hash', nri_report, 'new_rootfs_info',
                              convert=common.FormatSha256)

    # Check: payload must contain at least one operation.
    if not(len(manifest.install_operations) or
           len(manifest.kernel_install_operations)):
      raise PayloadError('payload contains no operations')

  def _CheckLength(self, length, total_blocks, op_name, length_name):
    """Checks whether a length matches the space designated in extents.

    Args:
      length: the total length of the data
      total_blocks: the total number of blocks in extents
      op_name: operation name (for error reporting)
      length_name: length name (for error reporting)
    Raises:
      PayloadError is there a problem with the length.

    """
    # Check: length is non-zero.
    if length == 0:
      raise PayloadError('%s: %s is zero' % (op_name, length_name))

    # Check that length matches number of blocks.
    self._CheckBlocksFitLength(length, total_blocks, self.block_size,
                               '%s: %s' % (op_name, length_name))

  def _CheckExtents(self, extents, part_size, block_counters, name,
                    allow_pseudo=False, allow_signature=False):
    """Checks a sequence of extents.

    Args:
      extents: the sequence of extents to check
      part_size: the total size of the partition to which the extents apply
      block_counters: an array of counters corresponding to the number of blocks
      name: the name of the extent block
      allow_pseudo: whether or not pseudo block numbers are allowed
      allow_signature: whether or not the extents are used for a signature
    Returns:
      The total number of blocks in the extents.
    Raises:
      PayloadError if any of the entailed checks fails.

    """
    total_num_blocks = 0
    for ex, ex_name in common.ExtentIter(extents, name):
      # Check: mandatory fields.
      start_block = PayloadChecker._CheckMandatoryField(ex, 'start_block',
                                                        None, ex_name)
      num_blocks = PayloadChecker._CheckMandatoryField(ex, 'num_blocks', None,
                                                       ex_name)
      end_block = start_block + num_blocks

      # Check: num_blocks > 0.
      if num_blocks == 0:
        raise PayloadError('%s: extent length is zero' % ex_name)

      if start_block != common.PSEUDO_EXTENT_MARKER:
        # Check: make sure we're within the partition limit.
        if part_size and end_block * self.block_size > part_size:
          raise PayloadError(
              '%s: extent (%s) exceeds partition size (%d)' %
              (ex_name, common.FormatExtent(ex, self.block_size), part_size))

        # Record block usage.
        for i in range(start_block, end_block):
          block_counters[i] += 1
      elif not (allow_pseudo or (allow_signature and len(extents) == 1)):
        # Pseudo-extents must be allowed explicitly, or otherwise be part of a
        # signature operation (in which case there has to be exactly one).
        raise PayloadError('%s: unexpected pseudo-extent' % ex_name)

      total_num_blocks += num_blocks

    return total_num_blocks

  def _CheckReplaceOperation(self, op, data_length, total_dst_blocks, op_name):
    """Specific checks for REPLACE/REPLACE_BZ operations.

    Args:
      op: the operation object from the manifest
      data_length: the length of the data blob associated with the operation
      total_dst_blocks: total number of blocks in dst_extents
      op_name: operation name for error reporting
    Raises:
      PayloadError if any check fails.

    """
    # Check: does not contain src extents.
    if op.src_extents:
      raise PayloadError('%s: contains src_extents' % op_name)

    # Check: contains data.
    if data_length is None:
      raise PayloadError('%s: missing data_{offset,length}' % op_name)

    if op.type == common.OpType.REPLACE:
      PayloadChecker._CheckBlocksFitLength(data_length, total_dst_blocks,
                                           self.block_size,
                                           op_name + '.data_length', 'dst')
    else:
      # Check: data_length must be smaller than the alotted dst blocks.
      if data_length >= total_dst_blocks * self.block_size:
        raise PayloadError(
            '%s: data_length (%d) must be less than allotted dst block '
            'space (%d * %d)' %
            (op_name, data_length, total_dst_blocks, self.block_size))

  def _CheckMoveOperation(self, op, data_offset, total_src_blocks,
                          total_dst_blocks, op_name):
    """Specific checks for MOVE operations.

    Args:
      op: the operation object from the manifest
      data_offset: the offset of a data blob for the operation
      total_src_blocks: total number of blocks in src_extents
      total_dst_blocks: total number of blocks in dst_extents
      op_name: operation name for error reporting
    Raises:
      PayloadError if any check fails.

    """
    # Check: no data_{offset,length}.
    if data_offset is not None:
      raise PayloadError('%s: contains data_{offset,length}' % op_name)

    # Check: total src blocks == total dst blocks.
    if total_src_blocks != total_dst_blocks:
      raise PayloadError(
          '%s: total src blocks (%d) != total dst blocks (%d)' %
          (op_name, total_src_blocks, total_dst_blocks))

    # Check: for all i, i-th src block index != i-th dst block index.
    i = 0
    src_extent_iter = iter(op.src_extents)
    dst_extent_iter = iter(op.dst_extents)
    src_extent = dst_extent = None
    src_idx = src_num = dst_idx = dst_num = 0
    while i < total_src_blocks:
      # Get the next source extent, if needed.
      if not src_extent:
        try:
          src_extent = src_extent_iter.next()
        except StopIteration:
          raise PayloadError('%s: ran out of src extents (%d/%d)' %
                             (op_name, i, total_src_blocks))
        src_idx = src_extent.start_block
        src_num = src_extent.num_blocks

      # Get the next dest extent, if needed.
      if not dst_extent:
        try:
          dst_extent = dst_extent_iter.next()
        except StopIteration:
          raise PayloadError('%s: ran out of dst extents (%d/%d)' %
                             (op_name, i, total_dst_blocks))
        dst_idx = dst_extent.start_block
        dst_num = dst_extent.num_blocks

      if src_idx == dst_idx:
        raise PayloadError('%s: src/dst block number %d is the same (%d)' %
                           (op_name, i, src_idx))

      advance = min(src_num, dst_num)
      i += advance

      src_idx += advance
      src_num -= advance
      if src_num == 0:
        src_extent = None

      dst_idx += advance
      dst_num -= advance
      if dst_num == 0:
        dst_extent = None

    # Make sure we've exhausted all src/dst extents.
    if src_extent:
      raise PayloadError('%s: excess src blocks' % op_name)
    if dst_extent:
      raise PayloadError('%s: excess dst blocks' % op_name)

  def _CheckBsdiffOperation(self, data_length, total_dst_blocks, op_name):
    """Specific checks for BSDIFF operations.

    Args:
      data_length: the length of the data blob associated with the operation
      total_dst_blocks: total number of blocks in dst_extents
      op_name: operation name for error reporting
    Raises:
      PayloadError if any check fails.

    """
    # Check: data_{offset,length} present.
    if data_length is None:
      raise PayloadError('%s: missing data_{offset,length}' % op_name)

    # Check: data_length is strictly smaller than the alotted dst blocks.
    if data_length >= total_dst_blocks * self.block_size:
      raise PayloadError(
          '%s: data_length (%d) must be smaller than allotted dst space '
          '(%d * %d = %d)' %
          (op_name, data_length, total_dst_blocks, self.block_size,
           total_dst_blocks * self.block_size))

  def _CheckOperation(self, op, op_name, is_last, old_block_counters,
                      new_block_counters, old_part_size, new_part_size,
                      prev_data_offset, allow_signature, allow_unhashed,
                      blob_hash_counts):
    """Checks a single update operation.

    Args:
      op: the operation object
      op_name: operation name string for error reporting
      is_last: whether this is the last operation in the sequence
      old_block_counters: arrays of block read counters
      new_block_counters: arrays of block write counters
      old_part_size: the source partition size in bytes
      new_part_size: the target partition size in bytes
      prev_data_offset: offset of last used data bytes
      allow_signature: whether this may be a signature operation
      allow_unhashed: allow operations with unhashed data blobs
      blob_hash_counts: counters for hashed/unhashed blobs
    Returns:
      The amount of data blob associated with the operation.
    Raises:
      PayloadError if any check has failed.

    """
    # Check extents.
    total_src_blocks = self._CheckExtents(
        op.src_extents, old_part_size, old_block_counters,
        op_name + '.src_extents', allow_pseudo=True)
    allow_signature_in_extents = (allow_signature and is_last and
                                  op.type == common.OpType.REPLACE)
    total_dst_blocks = self._CheckExtents(
        op.dst_extents, new_part_size, new_block_counters,
        op_name + '.dst_extents', allow_signature=allow_signature_in_extents)

    # Check: data_offset present <==> data_length present.
    data_offset = self._CheckOptionalField(op, 'data_offset', None)
    data_length = self._CheckOptionalField(op, 'data_length', None)
    self._CheckPresentIff(data_offset, data_length, 'data_offset',
                          'data_length', op_name)

    # Check: at least one dst_extent.
    if not op.dst_extents:
      raise PayloadError('%s: dst_extents is empty' % op_name)

    # Check {src,dst}_length, if present.
    if op.HasField('src_length'):
      self._CheckLength(op.src_length, total_src_blocks, op_name, 'src_length')
    if op.HasField('dst_length'):
      self._CheckLength(op.dst_length, total_dst_blocks, op_name, 'dst_length')

    if op.HasField('data_sha256_hash'):
      blob_hash_counts['hashed'] += 1

      # Check: operation carries data.
      if data_offset is None:
        raise PayloadError(
            '%s: data_sha256_hash present but no data_{offset,length}' %
            op_name)

      # Check: hash verifies correctly.
      # pylint: disable=E1101
      actual_hash = hashlib.sha256(self.payload.ReadDataBlob(data_offset,
                                                             data_length))
      if op.data_sha256_hash != actual_hash.digest():
        raise PayloadError(
            '%s: data_sha256_hash (%s) does not match actual hash (%s)' %
            (op_name, op.data_sha256_hash.encode('hex'),
             actual_hash.hexdigest()))
    elif data_offset is not None:
      if allow_signature_in_extents:
        blob_hash_counts['signature'] += 1
      elif allow_unhashed:
        blob_hash_counts['unhashed'] += 1
      else:
        raise PayloadError('%s: unhashed operation not allowed' % op_name)

    if data_offset is not None:
      # Check: contiguous use of data section.
      if data_offset != prev_data_offset:
        raise PayloadError(
            '%s: data offset (%d) not matching amount used so far (%d)' %
            (op_name, data_offset, prev_data_offset))

    # Type-specific checks.
    if op.type in (common.OpType.REPLACE, common.OpType.REPLACE_BZ):
      self._CheckReplaceOperation(op, data_length, total_dst_blocks, op_name)
    elif self.payload_type == _TYPE_FULL:
      raise PayloadError('%s: non-REPLACE operation in a full payload' %
                         op_name)
    elif op.type == common.OpType.MOVE:
      self._CheckMoveOperation(op, data_offset, total_src_blocks,
                               total_dst_blocks, op_name)
    elif op.type == common.OpType.BSDIFF:
      self._CheckBsdiffOperation(data_length, total_dst_blocks, op_name)
    else:
      assert False, 'cannot get here'

    return data_length if data_length is not None else 0

  def _AllocBlockCounters(self, part_size):
    """Returns a freshly initialized array of block counters.

    Args:
      part_size: the size of the partition
    Returns:
      An array of unsigned char elements initialized to zero, one for each of
      the blocks necessary for containing the partition.

    """
    num_blocks = (part_size + self.block_size - 1) / self.block_size
    return array.array('B', [0] * num_blocks)

  def _CheckOperations(self, operations, report, base_name, old_part_size,
                       new_part_size, prev_data_offset, allow_unhashed,
                       allow_signature):
    """Checks a sequence of update operations.

    Args:
      operations: the sequence of operations to check
      report: the report object to add to
      base_name: the name of the operation block
      old_part_size: the old partition size in bytes
      new_part_size: the new partition size in bytes
      prev_data_offset: offset of last used data bytes
      allow_unhashed: allow operations with unhashed data blobs
      allow_signature: whether this sequence may contain signature operations
    Returns:
      The total data blob size used.
    Raises:
      PayloadError if any of the checks fails.

    """
    # The total size of data blobs used by operations scanned thus far.
    total_data_used = 0
    # Counts of specific operation types.
    op_counts = {
        common.OpType.REPLACE: 0,
        common.OpType.REPLACE_BZ: 0,
        common.OpType.MOVE: 0,
        common.OpType.BSDIFF: 0,
    }
    # Total blob sizes for each operation type.
    op_blob_totals = {
        common.OpType.REPLACE: 0,
        common.OpType.REPLACE_BZ: 0,
        # MOVE operations don't have blobs
        common.OpType.BSDIFF: 0,
    }
    # Counts of hashed vs unhashed operations.
    blob_hash_counts = {
        'hashed': 0,
        'unhashed': 0,
    }
    if allow_signature:
      blob_hash_counts['signature'] = 0

    # Allocate old and new block counters.
    old_block_counters = (self._AllocBlockCounters(old_part_size)
                          if old_part_size else None)
    new_block_counters = self._AllocBlockCounters(new_part_size)

    # Process and verify each operation.
    op_num = 0
    for op, op_name in common.OperationIter(operations, base_name):
      op_num += 1

      # Check: type is valid.
      if op.type not in op_counts.keys():
        raise PayloadError('%s: invalid type (%d)' % (op_name, op.type))
      op_counts[op.type] += 1

      is_last = op_num == len(operations)
      curr_data_used = self._CheckOperation(
          op, op_name, is_last, old_block_counters, new_block_counters,
          old_part_size, new_part_size, prev_data_offset + total_data_used,
          allow_signature, allow_unhashed, blob_hash_counts)
      if curr_data_used:
        op_blob_totals[op.type] += curr_data_used
        total_data_used += curr_data_used

    # Report totals and breakdown statistics.
    report.AddField('total operations', op_num)
    report.AddField(
        None,
        histogram.Histogram.FromCountDict(op_counts,
                                          key_names=common.OpType.NAMES),
        indent=1)
    report.AddField('total blobs', sum(blob_hash_counts.values()))
    report.AddField(None,
                    histogram.Histogram.FromCountDict(blob_hash_counts),
                    indent=1)
    report.AddField('total blob size', _AddHumanReadableSize(total_data_used))
    report.AddField(
        None,
        histogram.Histogram.FromCountDict(op_blob_totals,
                                          formatter=_AddHumanReadableSize,
                                          key_names=common.OpType.NAMES),
        indent=1)

    # Report read/write histograms.
    if old_block_counters:
      report.AddField('block read hist',
                      histogram.Histogram.FromKeyList(old_block_counters),
                      linebreak=True, indent=1)

    new_write_hist = histogram.Histogram.FromKeyList(new_block_counters)
    # Check: full update must write each dst block once.
    if self.payload_type == _TYPE_FULL and new_write_hist.GetKeys() != [1]:
      raise PayloadError(
          '%s: not all blocks written exactly once during full update' %
          base_name)

    report.AddField('block write hist', new_write_hist, linebreak=True,
                    indent=1)

    return total_data_used

  def _CheckSignatures(self, report, pubkey_file_name):
    """Checks a payload's signature block."""
    sigs_raw = self.payload.ReadDataBlob(self.sigs_offset, self.sigs_size)
    sigs = update_metadata_pb2.Signatures()
    sigs.ParseFromString(sigs_raw)
    report.AddSection('signatures')

    # Check: at least one signature present.
    # pylint: disable=E1101
    if not sigs.signatures:
      raise PayloadError('signature block is empty')

    last_ops_section = (self.payload.manifest.kernel_install_operations or
                        self.payload.manifest.install_operations)
    fake_sig_op = last_ops_section[-1]
    # Check: signatures_{offset,size} must match the last (fake) operation.
    if not (fake_sig_op.type == common.OpType.REPLACE and
            self.sigs_offset == fake_sig_op.data_offset and
            self.sigs_size == fake_sig_op.data_length):
      raise PayloadError(
          'signatures_{offset,size} (%d+%d) does not match last operation '
          '(%d+%d)' %
          (self.sigs_offset, self.sigs_size, fake_sig_op.data_offset,
           fake_sig_op.data_length))

    # Compute the checksum of all data up to signature blob.
    # TODO(garnold) we're re-reading the whole data section into a string
    # just to compute the checksum; instead, we could do it incrementally as
    # we read the blobs one-by-one, under the assumption that we're reading
    # them in order (which currently holds). This should be reconsidered.
    payload_hasher = self.payload.manifest_hasher.copy()
    common.Read(self.payload.payload_file, self.sigs_offset,
                offset=self.payload.data_offset, hasher=payload_hasher)

    for sig, sig_name in common.SignatureIter(sigs.signatures, 'signatures'):
      sig_report = report.AddSubReport(sig_name)

      # Check: signature contains mandatory fields.
      self._CheckMandatoryField(sig, 'version', sig_report, sig_name)
      self._CheckMandatoryField(sig, 'data', None, sig_name)
      sig_report.AddField('data len', len(sig.data))

      # Check: signatures pertains to actual payload hash.
      if sig.version == 1:
        self._CheckSha256Signature(sig.data, pubkey_file_name,
                                   payload_hasher.digest(), sig_name)
      else:
        raise PayloadError('unknown signature version (%d)' % sig.version)

  def Run(self, pubkey_file_name=None, metadata_sig_file=None,
          report_out_file=None, assert_type=None, block_size=0,
          allow_unhashed=False):
    """Checker entry point, invoking all checks.

    Args:
      pubkey_file_name: public key used for signature verification
      metadata_sig_file: metadata signature, if verification is desired
      report_out_file: file object to dump the report to
      assert_type: assert that payload is either 'full' or 'delta' (optional)
      block_size: expected filesystem / payload block size
      allow_unhashed: allow operations with unhashed data blobs
    Raises:
      PayloadError if payload verification failed.

    """
    report = _PayloadReport()

    if assert_type not in (None, _TYPE_FULL, _TYPE_DELTA):
      raise PayloadError("invalid assert_type value (`%s')" % assert_type)
    self.payload_type = assert_type

    if block_size:
      self.block_size = block_size
    if not _IsPowerOfTwo(self.block_size):
      raise PayloadError('expected block (%d) size is not a power of two' %
                         self.block_size)

    # Get payload file size.
    self.payload.payload_file.seek(0, 2)
    payload_file_size = self.payload.payload_file.tell()
    self.payload.ResetFile()

    try:
      # Check metadata signature (if provided).
      if metadata_sig_file:
        if not pubkey_file_name:
          raise PayloadError(
              'no public key provided, cannot verify metadata signature')
        metadata_sig = base64.b64decode(metadata_sig_file.read())
        self._CheckSha256Signature(metadata_sig, pubkey_file_name,
                                   self.payload.manifest_hasher.digest(),
                                   'metadata signature')

      # Part 1: check the file header.
      report.AddSection('header')
      # Check: payload version is valid.
      if self.payload.header.version != 1:
        raise PayloadError('unknown payload version (%d)' %
                           self.payload.header.version)
      report.AddField('version', self.payload.header.version)
      report.AddField('manifest len', self.payload.header.manifest_len)

      # Part 2: check the manifest.
      self._CheckManifest(report)
      assert self.payload_type, 'payload type should be known by now'

      # Part 3: examine rootfs operations.
      report.AddSection('rootfs operations')
      total_blob_size = self._CheckOperations(
          self.payload.manifest.install_operations, report,
          'install_operations', self.old_rootfs_size,
          self.new_rootfs_size, 0, allow_unhashed, False)

      # Part 4: examine kernel operations.
      report.AddSection('kernel operations')
      total_blob_size += self._CheckOperations(
          self.payload.manifest.kernel_install_operations, report,
          'kernel_install_operations', self.old_kernel_size,
          self.new_kernel_size, total_blob_size, allow_unhashed, True)

      # Check: operations data reach the end of the payload file.
      used_payload_size = self.payload.data_offset + total_blob_size
      if used_payload_size != payload_file_size:
        raise PayloadError(
            'used payload size (%d) different from actual file size (%d)' %
            (used_payload_size, payload_file_size))

      # Part 5: handle payload signatures message.
      if self.sigs_size:
        if not pubkey_file_name:
          raise PayloadError(
              'no public key provided, cannot verify payload signature')
        self._CheckSignatures(report, pubkey_file_name)

      # Part 6: summary.
      report.AddSection('summary')
      report.AddField('update type', self.payload_type)

      report.Finalize()
    finally:
      if report_out_file:
        report.Dump(report_out_file)