path: root/lib/spack/external/macholib/MachO.py

"""
Utilities for reading and writing Mach-O headers
"""
from __future__ import print_function

import sys
import struct
import os

from .mach_o import MH_FILETYPE_SHORTNAMES, LC_DYSYMTAB, LC_SYMTAB
from .mach_o import load_command, S_ZEROFILL, section_64, section
from .mach_o import LC_REGISTRY, LC_ID_DYLIB, LC_SEGMENT, fat_header
from .mach_o import LC_SEGMENT_64, MH_CIGAM_64, MH_MAGIC_64, FAT_MAGIC
from .mach_o import mach_header, fat_arch64, FAT_MAGIC_64, fat_arch
from .mach_o import LC_REEXPORT_DYLIB, LC_PREBOUND_DYLIB, LC_LOAD_WEAK_DYLIB
from .mach_o import LC_LOAD_UPWARD_DYLIB, LC_LOAD_DYLIB, mach_header_64
from .mach_o import MH_CIGAM, MH_MAGIC
from .ptypes import sizeof

from macholib.util import fileview
try:
    from macholib.compat import bytes
except ImportError:
    pass

try:
    unicode
except NameError:
    unicode = str

if sys.version_info[0] == 2:
    range = xrange  # noqa: F821

__all__ = ['MachO']

_RELOCATABLE = set((
    # relocatable commands that should be used for dependency walking
    LC_LOAD_DYLIB,
    LC_LOAD_UPWARD_DYLIB,
    LC_LOAD_WEAK_DYLIB,
    LC_PREBOUND_DYLIB,
    LC_REEXPORT_DYLIB,
))

_RELOCATABLE_NAMES = {
    LC_LOAD_DYLIB: 'load_dylib',
    LC_LOAD_UPWARD_DYLIB: 'load_upward_dylib',
    LC_LOAD_WEAK_DYLIB: 'load_weak_dylib',
    LC_PREBOUND_DYLIB: 'prebound_dylib',
    LC_REEXPORT_DYLIB: 'reexport_dylib',
}


def _shouldRelocateCommand(cmd):
    """
    Should this command id be investigated for relocation?
    """
    return cmd in _RELOCATABLE


def lc_str_value(offset, cmd_info):
    """
    Fetch the actual value of a field of type "lc_str"
    """
    cmd_load, cmd_cmd, cmd_data = cmd_info

    offset -= sizeof(cmd_load) + sizeof(cmd_cmd)
    return cmd_data[offset:].strip(b'\x00')


class MachO(object):
    """
    Provides reading/writing the Mach-O header of a specific existing file
    """
    #   filename   - the original filename of this mach-o
    #   sizediff   - the current deviation from the initial mach-o size
    #   header     - the mach-o header
    #   commands   - a list of (load_command, somecommand, data)
    #                data is either a str, or a list of segment structures
    #   total_size - the current mach-o header size (including header)
    #   low_offset - essentially, the maximum mach-o header size
    #   id_cmd     - the index of my id command, or None

    def __init__(self, filename):

        # supports the ObjectGraph protocol
        self.graphident = filename
        self.filename = filename
        self.loader_path = os.path.dirname(filename)

        # initialized by load
        self.fat = None
        self.headers = []
        with open(filename, 'rb') as fp:
            self.load(fp)

    def __repr__(self):
        return "<MachO filename=%r>" % (self.filename,)

    def load(self, fh):
        assert fh.tell() == 0
        header = struct.unpack('>I', fh.read(4))[0]
        fh.seek(0)
        if header in (FAT_MAGIC, FAT_MAGIC_64):
            self.load_fat(fh)
        else:
            fh.seek(0, 2)
            size = fh.tell()
            fh.seek(0)
            self.load_header(fh, 0, size)

    def load_fat(self, fh):
        self.fat = fat_header.from_fileobj(fh)
        if self.fat.magic == FAT_MAGIC:
            archs = [fat_arch.from_fileobj(fh)
                     for i in range(self.fat.nfat_arch)]
        elif self.fat.magic == FAT_MAGIC_64:
            archs = [fat_arch64.from_fileobj(fh)
                     for i in range(self.fat.nfat_arch)]
        else:
            raise ValueError("Unknown fat header magic: %r" % (self.fat.magic))

        for arch in archs:
            self.load_header(fh, arch.offset, arch.size)

    def rewriteLoadCommands(self, *args, **kw):
        changed = False
        for header in self.headers:
            if header.rewriteLoadCommands(*args, **kw):
                changed = True
        return changed

    def load_header(self, fh, offset, size):
        fh.seek(offset)
        header = struct.unpack('>I', fh.read(4))[0]
        fh.seek(offset)
        if header == MH_MAGIC:
            magic, hdr, endian = MH_MAGIC, mach_header, '>'
        elif header == MH_CIGAM:
            magic, hdr, endian = MH_CIGAM, mach_header, '<'
        elif header == MH_MAGIC_64:
            magic, hdr, endian = MH_MAGIC_64, mach_header_64, '>'
        elif header == MH_CIGAM_64:
            magic, hdr, endian = MH_CIGAM_64, mach_header_64, '<'
        else:
            raise ValueError("Unknown Mach-O header: 0x%08x in %r" % (
                header, fh))
        hdr = MachOHeader(self, fh, offset, size, magic, hdr, endian)
        self.headers.append(hdr)

    def write(self, f):
        for header in self.headers:
            header.write(f)


class MachOHeader(object):
    """
    Provides reading/writing the Mach-O header of a specific existing file
    """
    #   filename   - the original filename of this mach-o
    #   sizediff   - the current deviation from the initial mach-o size
    #   header     - the mach-o header
    #   commands   - a list of (load_command, somecommand, data)
    #                data is either a str, or a list of segment structures
    #   total_size - the current mach-o header size (including header)
    #   low_offset - essentially, the maximum mach-o header size
    #   id_cmd     - the index of my id command, or None

    def __init__(self, parent, fh, offset, size, magic, hdr, endian):
        self.MH_MAGIC = magic
        self.mach_header = hdr

        # These are all initialized by self.load()
        self.parent = parent
        self.offset = offset
        self.size = size

        self.endian = endian
        self.header = None
        self.commands = None
        self.id_cmd = None
        self.sizediff = None
        self.total_size = None
        self.low_offset = None
        self.filetype = None
        self.headers = []

        self.load(fh)

    def __repr__(self):
        return "<%s filename=%r offset=%d size=%d endian=%r>" % (
            type(self).__name__, self.parent.filename, self.offset, self.size,
            self.endian)

    def load(self, fh):
        fh = fileview(fh, self.offset, self.size)
        fh.seek(0)

        self.sizediff = 0
        kw = {'_endian_': self.endian}
        header = self.mach_header.from_fileobj(fh, **kw)
        self.header = header
        # if header.magic != self.MH_MAGIC:
        #    raise ValueError("header has magic %08x, expecting %08x" % (
        #        header.magic, self.MH_MAGIC))

        cmd = self.commands = []

        self.filetype = self.get_filetype_shortname(header.filetype)

        read_bytes = 0
        low_offset = sys.maxsize
        for i in range(header.ncmds):
            # read the load command
            cmd_load = load_command.from_fileobj(fh, **kw)

            # read the specific command
            klass = LC_REGISTRY.get(cmd_load.cmd, None)
            if klass is None:
                raise ValueError("Unknown load command: %d" % (cmd_load.cmd,))
            cmd_cmd = klass.from_fileobj(fh, **kw)

            if cmd_load.cmd == LC_ID_DYLIB:
                # remember where this command was
                if self.id_cmd is not None:
                    raise ValueError("This dylib already has an id")
                self.id_cmd = i

            if cmd_load.cmd in (LC_SEGMENT, LC_SEGMENT_64):
                # for segment commands, read the list of segments
                segs = []
                # assert that the size makes sense
                if cmd_load.cmd == LC_SEGMENT:
                    section_cls = section
                else:  # LC_SEGMENT_64
                    section_cls = section_64

                expected_size = (
                    sizeof(klass) + sizeof(load_command) +
                    (sizeof(section_cls) * cmd_cmd.nsects)
                )
                if cmd_load.cmdsize != expected_size:
                    raise ValueError("Segment size mismatch")
                # this is a zero block or something
                # so the beginning is wherever the fileoff of this command is
                if cmd_cmd.nsects == 0:
                    if cmd_cmd.filesize != 0:
                        low_offset = min(low_offset, cmd_cmd.fileoff)
                else:
                    # this one has multiple segments
                    for j in range(cmd_cmd.nsects):
                        # read the segment
                        seg = section_cls.from_fileobj(fh, **kw)
                        # if the segment has a size and is not zero filled
                        # then its beginning is the offset of this segment
                        not_zerofill = ((seg.flags & S_ZEROFILL) != S_ZEROFILL)
                        if seg.offset > 0 and seg.size > 0 and not_zerofill:
                            low_offset = min(low_offset, seg.offset)
                        if not_zerofill:
                            c = fh.tell()
                            fh.seek(seg.offset)
                            sd = fh.read(seg.size)
                            seg.add_section_data(sd)
                            fh.seek(c)
                        segs.append(seg)
                # data is a list of segments
                cmd_data = segs

            # XXX: Disabled for now because writing back doesn't work
            # elif cmd_load.cmd == LC_CODE_SIGNATURE:
            #    c = fh.tell()
            #    fh.seek(cmd_cmd.dataoff)
            #    cmd_data = fh.read(cmd_cmd.datasize)
            #    fh.seek(c)
            # elif cmd_load.cmd == LC_SYMTAB:
            #    c = fh.tell()
            #    fh.seek(cmd_cmd.stroff)
            #    cmd_data = fh.read(cmd_cmd.strsize)
            #    fh.seek(c)

            else:
                # data is a raw str
                data_size = (
                    cmd_load.cmdsize - sizeof(klass) - sizeof(load_command)
                )
                cmd_data = fh.read(data_size)
            cmd.append((cmd_load, cmd_cmd, cmd_data))
            read_bytes += cmd_load.cmdsize

        # make sure the header made sense
        if read_bytes != header.sizeofcmds:
            raise ValueError("Read %d bytes, header reports %d bytes" % (
                read_bytes, header.sizeofcmds))
        self.total_size = sizeof(self.mach_header) + read_bytes
        self.low_offset = low_offset

    def walkRelocatables(self, shouldRelocateCommand=_shouldRelocateCommand):
        """
        for all relocatable commands
        yield (command_index, command_name, filename)
        """
        for (idx, (lc, cmd, data)) in enumerate(self.commands):
            if shouldRelocateCommand(lc.cmd):
                name = _RELOCATABLE_NAMES[lc.cmd]
                ofs = cmd.name - sizeof(lc.__class__) - sizeof(cmd.__class__)
                yield idx, name, data[ofs:data.find(b'\x00', ofs)].decode(
                        sys.getfilesystemencoding())

    def rewriteInstallNameCommand(self, loadcmd):
        """Rewrite the load command of this dylib"""
        if self.id_cmd is not None:
            self.rewriteDataForCommand(self.id_cmd, loadcmd)
            return True
        return False

    def changedHeaderSizeBy(self, bytes):
        self.sizediff += bytes
        if (self.total_size + self.sizediff) > self.low_offset:
            print(
                "WARNING: Mach-O header in %r may be too large to relocate" % (
                    self.parent.filename,))

    def rewriteLoadCommands(self, changefunc):
        """
        Rewrite the load commands based upon a change dictionary
        """
        data = changefunc(self.parent.filename)
        changed = False
        if data is not None:
            if self.rewriteInstallNameCommand(
                    data.encode(sys.getfilesystemencoding())):
                changed = True
        for idx, name, filename in self.walkRelocatables():
            data = changefunc(filename)
            if data is not None:
                if self.rewriteDataForCommand(idx, data.encode(
                        sys.getfilesystemencoding())):
                    changed = True
        return changed

    def rewriteDataForCommand(self, idx, data):
        lc, cmd, old_data = self.commands[idx]
        hdrsize = sizeof(lc.__class__) + sizeof(cmd.__class__)
        align = struct.calcsize('Q')
        data = data + (b'\x00' * (align - (len(data) % align)))
        newsize = hdrsize + len(data)
        self.commands[idx] = (lc, cmd, data)
        self.changedHeaderSizeBy(newsize - lc.cmdsize)
        lc.cmdsize, cmd.name = newsize, hdrsize
        return True

    def synchronize_size(self):
        if (self.total_size + self.sizediff) > self.low_offset:
            raise ValueError(
                ("New Mach-O header is too large to relocate in %r "
                 "(new size=%r, max size=%r, delta=%r)") % (
                    self.parent.filename, self.total_size + self.sizediff,
                    self.low_offset, self.sizediff))
        self.header.sizeofcmds += self.sizediff
        self.total_size = sizeof(self.mach_header) + self.header.sizeofcmds
        self.sizediff = 0

    def write(self, fileobj):
        fileobj = fileview(fileobj, self.offset, self.size)
        fileobj.seek(0)

        # serialize all the mach-o commands
        self.synchronize_size()

        self.header.to_fileobj(fileobj)
        for lc, cmd, data in self.commands:
            lc.to_fileobj(fileobj)
            cmd.to_fileobj(fileobj)

            if sys.version_info[0] == 2:
                if isinstance(data, unicode):
                    fileobj.write(data.encode(sys.getfilesystemencoding()))

                elif isinstance(data, (bytes, str)):
                    fileobj.write(data)
                else:
                    # segments..
                    for obj in data:
                        obj.to_fileobj(fileobj)
            else:
                if isinstance(data, str):
                    fileobj.write(data.encode(sys.getfilesystemencoding()))

                elif isinstance(data, bytes):
                    fileobj.write(data)

                else:
                    # segments..
                    for obj in data:
                        obj.to_fileobj(fileobj)

        # zero out the unused space, doubt this is strictly necessary
        # and is generally probably already the case
        fileobj.write(b'\x00' * (self.low_offset - fileobj.tell()))

    def getSymbolTableCommand(self):
        for lc, cmd, data in self.commands:
            if lc.cmd == LC_SYMTAB:
                return cmd
        return None

    def getDynamicSymbolTableCommand(self):
        for lc, cmd, data in self.commands:
            if lc.cmd == LC_DYSYMTAB:
                return cmd
        return None

    def get_filetype_shortname(self, filetype):
        if filetype in MH_FILETYPE_SHORTNAMES:
            return MH_FILETYPE_SHORTNAMES[filetype]
        else:
            return 'unknown'


def main(fn):
    m = MachO(fn)
    seen = set()
    for header in m.headers:
        for idx, name, other in header.walkRelocatables():
            if other not in seen:
                seen.add(other)
                print('\t' + name + ": " + other)


if __name__ == '__main__':
    import sys
    files = sys.argv[1:] or ['/bin/ls']
    for fn in files:
        print(fn)
        main(fn)