#------------------------------------------------------------------------------- # elftools: elf/elffile.py # # ELFFile - main class for accessing ELF files # # Eli Bendersky (eliben@gmail.com) # This code is in the public domain #------------------------------------------------------------------------------- import io from io import BytesIO import os import struct import zlib try: import resource PAGESIZE = resource.getpagesize() except ImportError: try: # Windows system import mmap PAGESIZE = mmap.PAGESIZE except ImportError: # Jython PAGESIZE = 4096 from ..common.exceptions import ELFError, ELFParseError from ..common.utils import struct_parse, elf_assert from .structs import ELFStructs from .sections import ( Section, StringTableSection, SymbolTableSection, SymbolTableIndexSection, SUNWSyminfoTableSection, NullSection, NoteSection, StabSection, ARMAttributesSection, RISCVAttributesSection) from .dynamic import DynamicSection, DynamicSegment from .relocation import (RelocationSection, RelocationHandler, RelrRelocationSection) from .gnuversions import ( GNUVerNeedSection, GNUVerDefSection, GNUVerSymSection) from .segments import Segment, InterpSegment, NoteSegment from ..dwarf.dwarfinfo import DWARFInfo, DebugSectionDescriptor, DwarfConfig from ..ehabi.ehabiinfo import EHABIInfo from .hash import ELFHashSection, GNUHashSection from .constants import SHN_INDICES class ELFFile(object): """ Creation: the constructor accepts a stream (file-like object) with the contents of an ELF file. Optionally, a stream_loader function can be passed as the second argument. This stream_loader function takes a relative file path to load a supplementary object file, and returns a stream suitable for creating a new ELFFile. Currently, the only such relative file path is obtained from the supplementary object files. Accessible attributes: stream: The stream holding the data of the file - must be a binary stream (bytes, not string). elfclass: 32 or 64 - specifies the word size of the target machine little_endian: boolean - specifies the target machine's endianness elftype: string or int, either known value of E_TYPE enum defining ELF type (e.g. executable, dynamic library or core dump) or integral unparsed value header: the complete ELF file header e_ident_raw: the raw e_ident field of the header """ def __init__(self, stream, stream_loader=None): self.stream = stream self.stream.seek(0, io.SEEK_END) self.stream_len = self.stream.tell() self._identify_file() self.structs = ELFStructs( little_endian=self.little_endian, elfclass=self.elfclass) self.structs.create_basic_structs() self.header = self._parse_elf_header() self.structs.create_advanced_structs( self['e_type'], self['e_machine'], self['e_ident']['EI_OSABI']) self.stream.seek(0) self.e_ident_raw = self.stream.read(16) self._section_header_stringtable = \ self._get_section_header_stringtable() self._section_name_map = None self.stream_loader = stream_loader @classmethod def load_from_path(cls, path): """Takes a path to a file on the local filesystem, and returns an ELFFile from it, setting up a correct stream_loader relative to the original file. """ base_directory = os.path.dirname(path) def loader(elf_path): # FIXME: use actual path instead of str/bytes if not os.path.isabs(elf_path): elf_path = os.path.join(base_directory, elf_path) return open(elf_path, 'rb') stream = open(path, 'rb') return ELFFile(stream, loader) def num_sections(self): """ Number of sections in the file """ if self['e_shoff'] == 0: return 0 # From the ELF ABI documentation at # https://refspecs.linuxfoundation.org/elf/gabi4+/ch4.sheader.html: # "e_shnum normally tells how many entries the section header table # contains. [...] If the number of sections is greater than or equal to # SHN_LORESERVE (0xff00), e_shnum has the value SHN_UNDEF (0) and the # actual number of section header table entries is contained in the # sh_size field of the section header at index 0 (otherwise, the sh_size # member of the initial entry contains 0)." if self['e_shnum'] == 0: return self._get_section_header(0)['sh_size'] return self['e_shnum'] def get_section(self, n): """ Get the section at index #n from the file (Section object or a subclass) """ section_header = self._get_section_header(n) return self._make_section(section_header) def get_section_by_name(self, name): """ Get a section from the file, by name. Return None if no such section exists. """ # The first time this method is called, construct a name to number # mapping # if self._section_name_map is None: self._make_section_name_map() secnum = self._section_name_map.get(name, None) return None if secnum is None else self.get_section(secnum) def get_section_index(self, section_name): """ Gets the index of the section by name. Return None if no such section name exists. """ # The first time this method is called, construct a name to number # mapping # if self._section_name_map is None: self._make_section_name_map() return self._section_name_map.get(section_name, None) def iter_sections(self, type=None): """ Yield all the sections in the file. If the optional |type| parameter is passed, this method will only yield sections of the given type. The parameter value must be a string containing the name of the type as defined in the ELF specification, e.g. 'SHT_SYMTAB'. """ for i in range(self.num_sections()): section = self.get_section(i) if type is None or section['sh_type'] == type: yield section def num_segments(self): """ Number of segments in the file """ # From: https://github.com/hjl-tools/x86-psABI/wiki/X86-psABI # Section: 4.1.2 Number of Program Headers # If the number of program headers is greater than or equal to # PN_XNUM (0xffff), this member has the value PN_XNUM # (0xffff). The actual number of program header table entries # is contained in the sh_info field of the section header at # index 0. if self['e_phnum'] < 0xffff: return self['e_phnum'] else: return self.get_section(0)['sh_info'] def get_segment(self, n): """ Get the segment at index #n from the file (Segment object) """ segment_header = self._get_segment_header(n) return self._make_segment(segment_header) def iter_segments(self, type=None): """ Yield all the segments in the file. If the optional |type| parameter is passed, this method will only yield segments of the given type. The parameter value must be a string containing the name of the type as defined in the ELF specification, e.g. 'PT_LOAD'. """ for i in range(self.num_segments()): segment = self.get_segment(i) if type is None or segment['p_type'] == type: yield segment def address_offsets(self, start, size=1): """ Yield a file offset for each ELF segment containing a memory region. A memory region is defined by the range [start...start+size). The offset of the region is yielded. """ end = start + size # consider LOAD only to prevent same address being yielded twice for seg in self.iter_segments(type='PT_LOAD'): if (start >= seg['p_vaddr'] and end <= seg['p_vaddr'] + seg['p_filesz']): yield start - seg['p_vaddr'] + seg['p_offset'] def has_dwarf_info(self): """ Check whether this file appears to have debugging information. We assume that if it has the .debug_info or .zdebug_info section, it has all the other required sections as well. """ return bool(self.get_section_by_name('.debug_info') or self.get_section_by_name('.zdebug_info') or self.get_section_by_name('.eh_frame')) def get_dwarf_info(self, relocate_dwarf_sections=True, follow_links=True): """ Return a DWARFInfo object representing the debugging information in this file. If relocate_dwarf_sections is True, relocations for DWARF sections are looked up and applied. If follow_links is True, we will try to load the supplementary object file (if any), and use it to resolve references and imports. """ # Expect that has_dwarf_info was called, so at least .debug_info is # present. # Sections that aren't found will be passed as None to DWARFInfo. section_names = ('.debug_info', '.debug_aranges', '.debug_abbrev', '.debug_str', '.debug_line', '.debug_frame', '.debug_loc', '.debug_ranges', '.debug_pubtypes', '.debug_pubnames', '.debug_addr', '.debug_str_offsets', '.debug_line_str', '.debug_loclists', '.debug_rnglists', '.debug_sup', '.gnu_debugaltlink') compressed = bool(self.get_section_by_name('.zdebug_info')) if compressed: section_names = tuple(map(lambda x: '.z' + x[1:], section_names)) # As it is loaded in the process image, .eh_frame cannot be compressed section_names += ('.eh_frame', ) (debug_info_sec_name, debug_aranges_sec_name, debug_abbrev_sec_name, debug_str_sec_name, debug_line_sec_name, debug_frame_sec_name, debug_loc_sec_name, debug_ranges_sec_name, debug_pubtypes_name, debug_pubnames_name, debug_addr_name, debug_str_offsets_name, debug_line_str_name, debug_loclists_sec_name, debug_rnglists_sec_name, debug_sup_name, gnu_debugaltlink_name, eh_frame_sec_name) = section_names debug_sections = {} for secname in section_names: section = self.get_section_by_name(secname) if section is None: debug_sections[secname] = None else: dwarf_section = self._read_dwarf_section( section, relocate_dwarf_sections) if compressed and secname.startswith('.z'): dwarf_section = self._decompress_dwarf_section(dwarf_section) debug_sections[secname] = dwarf_section # Lookup if we have any of the .gnu_debugaltlink (GNU proprietary # implementation) or .debug_sup sections, referencing a supplementary # DWARF file dwarfinfo = DWARFInfo( config=DwarfConfig( little_endian=self.little_endian, default_address_size=self.elfclass // 8, machine_arch=self.get_machine_arch()), debug_info_sec=debug_sections[debug_info_sec_name], debug_aranges_sec=debug_sections[debug_aranges_sec_name], debug_abbrev_sec=debug_sections[debug_abbrev_sec_name], debug_frame_sec=debug_sections[debug_frame_sec_name], eh_frame_sec=debug_sections[eh_frame_sec_name], debug_str_sec=debug_sections[debug_str_sec_name], debug_loc_sec=debug_sections[debug_loc_sec_name], debug_ranges_sec=debug_sections[debug_ranges_sec_name], debug_line_sec=debug_sections[debug_line_sec_name], debug_pubtypes_sec=debug_sections[debug_pubtypes_name], debug_pubnames_sec=debug_sections[debug_pubnames_name], debug_addr_sec=debug_sections[debug_addr_name], debug_str_offsets_sec=debug_sections[debug_str_offsets_name], debug_line_str_sec=debug_sections[debug_line_str_name], debug_loclists_sec=debug_sections[debug_loclists_sec_name], debug_rnglists_sec=debug_sections[debug_rnglists_sec_name], debug_sup_sec=debug_sections[debug_sup_name], gnu_debugaltlink_sec=debug_sections[gnu_debugaltlink_name] ) if follow_links: dwarfinfo.supplementary_dwarfinfo = self.get_supplementary_dwarfinfo(dwarfinfo) return dwarfinfo def get_supplementary_dwarfinfo(self, dwarfinfo): """ Read supplementary dwarfinfo, from either the standared .debug_sup section or the GNU proprietary .gnu_debugaltlink. """ supfilepath = dwarfinfo.parse_debugsupinfo() if supfilepath is not None and self.stream_loader is not None: stream = self.stream_loader(supfilepath) supelffile = ELFFile(stream) dwarf_info = supelffile.get_dwarf_info() stream.close() return dwarf_info return None def has_ehabi_info(self): """ Check whether this file appears to have arm exception handler index table. """ return any(self.iter_sections(type='SHT_ARM_EXIDX')) def get_ehabi_infos(self): """ Generally, shared library and executable contain 1 .ARM.exidx section. Object file contains many .ARM.exidx sections. So we must traverse every section and filter sections whose type is SHT_ARM_EXIDX. """ _ret = [] if self['e_type'] == 'ET_REL': # TODO: support relocatable file assert False, "Current version of pyelftools doesn't support relocatable file." for section in self.iter_sections(type='SHT_ARM_EXIDX'): _ret.append(EHABIInfo(section, self.little_endian)) return _ret if len(_ret) > 0 else None def get_machine_arch(self): """ Return the machine architecture, as detected from the ELF header. """ architectures = { 'EM_M32' : 'AT&T WE 32100', 'EM_SPARC' : 'SPARC', 'EM_386' : 'x86', 'EM_68K' : 'Motorola 68000', 'EM_88K' : 'Motorola 88000', 'EM_IAMCU' : 'Intel MCU', 'EM_860' : 'Intel 80860', 'EM_MIPS' : 'MIPS', 'EM_S370' : 'IBM System/370', 'EM_MIPS_RS3_LE' : 'MIPS RS3000 Little-endian', 'EM_PARISC' : 'Hewlett-Packard PA-RISC', 'EM_VPP500' : 'Fujitsu VPP500', 'EM_SPARC32PLUS' : 'Enhanced SPARC', 'EM_960' : 'Intel 80960', 'EM_PPC' : 'PowerPC', 'EM_PPC64' : '64-bit PowerPC', 'EM_S390' : 'IBM System/390', 'EM_SPU' : 'IBM SPU/SPC', 'EM_V800' : 'NEC V800', 'EM_FR20' : 'Fujitsu FR20', 'EM_RH32' : 'TRW RH-32', 'EM_RCE' : 'Motorola RCE', 'EM_ARM' : 'ARM', 'EM_ALPHA' : 'Digital Alpha', 'EM_SH' : 'Hitachi SH', 'EM_SPARCV9' : 'SPARC Version 9', 'EM_TRICORE' : 'Siemens TriCore embedded processor', 'EM_ARC' : 'Argonaut RISC Core, Argonaut Technologies Inc.', 'EM_H8_300' : 'Hitachi H8/300', 'EM_H8_300H' : 'Hitachi H8/300H', 'EM_H8S' : 'Hitachi H8S', 'EM_H8_500' : 'Hitachi H8/500', 'EM_IA_64' : 'Intel IA-64', 'EM_MIPS_X' : 'MIPS-X', 'EM_COLDFIRE' : 'Motorola ColdFire', 'EM_68HC12' : 'Motorola M68HC12', 'EM_MMA' : 'Fujitsu MMA', 'EM_PCP' : 'Siemens PCP', 'EM_NCPU' : 'Sony nCPU', 'EM_NDR1' : 'Denso NDR1', 'EM_STARCORE' : 'Motorola Star*Core', 'EM_ME16' : 'Toyota ME16', 'EM_ST100' : 'STMicroelectronics ST100', 'EM_TINYJ' : 'Advanced Logic TinyJ', 'EM_X86_64' : 'x64', 'EM_PDSP' : 'Sony DSP', 'EM_PDP10' : 'Digital Equipment PDP-10', 'EM_PDP11' : 'Digital Equipment PDP-11', 'EM_FX66' : 'Siemens FX66', 'EM_ST9PLUS' : 'STMicroelectronics ST9+ 8/16 bit', 'EM_ST7' : 'STMicroelectronics ST7 8-bit', 'EM_68HC16' : 'Motorola MC68HC16', 'EM_68HC11' : 'Motorola MC68HC11', 'EM_68HC08' : 'Motorola MC68HC08', 'EM_68HC05' : 'Motorola MC68HC05', 'EM_SVX' : 'Silicon Graphics SVx', 'EM_ST19' : 'STMicroelectronics ST19 8-bit', 'EM_VAX' : 'Digital VAX', 'EM_CRIS' : 'Axis Communications 32-bit', 'EM_JAVELIN' : 'Infineon Technologies 32-bit', 'EM_FIREPATH' : 'Element 14 64-bit DSP', 'EM_ZSP' : 'LSI Logic 16-bit DSP', 'EM_MMIX' : 'Donald Knuth\'s educational 64-bit', 'EM_HUANY' : 'Harvard University machine-independent object files', 'EM_PRISM' : 'SiTera Prism', 'EM_AVR' : 'Atmel AVR 8-bit', 'EM_FR30' : 'Fujitsu FR30', 'EM_D10V' : 'Mitsubishi D10V', 'EM_D30V' : 'Mitsubishi D30V', 'EM_V850' : 'NEC v850', 'EM_M32R' : 'Mitsubishi M32R', 'EM_MN10300' : 'Matsushita MN10300', 'EM_MN10200' : 'Matsushita MN10200', 'EM_PJ' : 'picoJava', 'EM_OPENRISC' : 'OpenRISC 32-bit', 'EM_ARC_COMPACT' : 'ARC International ARCompact', 'EM_XTENSA' : 'Tensilica Xtensa', 'EM_VIDEOCORE' : 'Alphamosaic VideoCore', 'EM_TMM_GPP' : 'Thompson Multimedia', 'EM_NS32K' : 'National Semiconductor 32000 series', 'EM_TPC' : 'Tenor Network TPC', 'EM_SNP1K' : 'Trebia SNP 1000', 'EM_ST200' : 'STMicroelectronics ST200', 'EM_IP2K' : 'Ubicom IP2xxx', 'EM_MAX' : 'MAX', 'EM_CR' : 'National Semiconductor CompactRISC', 'EM_F2MC16' : 'Fujitsu F2MC16', 'EM_MSP430' : 'Texas Instruments msp430', 'EM_BLACKFIN' : 'Analog Devices Blackfin', 'EM_SE_C33' : 'Seiko Epson S1C33', 'EM_SEP' : 'Sharp', 'EM_ARCA' : 'Arca RISC', 'EM_UNICORE' : 'PKU-Unity MPRC', 'EM_EXCESS' : 'eXcess', 'EM_DXP' : 'Icera Semiconductor Deep Execution Processor', 'EM_ALTERA_NIOS2' : 'Altera Nios II', 'EM_CRX' : 'National Semiconductor CompactRISC CRX', 'EM_XGATE' : 'Motorola XGATE', 'EM_C166' : 'Infineon C16x/XC16x', 'EM_M16C' : 'Renesas M16C', 'EM_DSPIC30F' : 'Microchip Technology dsPIC30F', 'EM_CE' : 'Freescale Communication Engine RISC core', 'EM_M32C' : 'Renesas M32C', 'EM_TSK3000' : 'Altium TSK3000', 'EM_RS08' : 'Freescale RS08', 'EM_SHARC' : 'Analog Devices SHARC', 'EM_ECOG2' : 'Cyan Technology eCOG2', 'EM_SCORE7' : 'Sunplus S+core7 RISC', 'EM_DSP24' : 'New Japan Radio (NJR) 24-bit DSP', 'EM_VIDEOCORE3' : 'Broadcom VideoCore III', 'EM_LATTICEMICO32' : 'Lattice FPGA RISC', 'EM_SE_C17' : 'Seiko Epson C17', 'EM_TI_C6000' : 'TI TMS320C6000', 'EM_TI_C2000' : 'TI TMS320C2000', 'EM_TI_C5500' : 'TI TMS320C55x', 'EM_TI_ARP32' : 'TI Application Specific RISC, 32bit', 'EM_TI_PRU' : 'TI Programmable Realtime Unit', 'EM_MMDSP_PLUS' : 'STMicroelectronics 64bit VLIW', 'EM_CYPRESS_M8C' : 'Cypress M8C', 'EM_R32C' : 'Renesas R32C', 'EM_TRIMEDIA' : 'NXP Semiconductors TriMedia', 'EM_QDSP6' : 'QUALCOMM DSP6', 'EM_8051' : 'Intel 8051', 'EM_STXP7X' : 'STMicroelectronics STxP7x', 'EM_NDS32' : 'Andes Technology RISC', 'EM_ECOG1' : 'Cyan Technology eCOG1X', 'EM_ECOG1X' : 'Cyan Technology eCOG1X', 'EM_MAXQ30' : 'Dallas Semiconductor MAXQ30', 'EM_XIMO16' : 'New Japan Radio (NJR) 16-bit', 'EM_MANIK' : 'M2000 Reconfigurable RISC', 'EM_CRAYNV2' : 'Cray Inc. NV2', 'EM_RX' : 'Renesas RX', 'EM_METAG' : 'Imagination Technologies META', 'EM_MCST_ELBRUS' : 'MCST Elbrus', 'EM_ECOG16' : 'Cyan Technology eCOG16', 'EM_CR16' : 'National Semiconductor CompactRISC CR16 16-bit', 'EM_ETPU' : 'Freescale', 'EM_SLE9X' : 'Infineon Technologies SLE9X', 'EM_L10M' : 'Intel L10M', 'EM_K10M' : 'Intel K10M', 'EM_AARCH64' : 'AArch64', 'EM_AVR32' : 'Atmel 32-bit', 'EM_STM8' : 'STMicroeletronics STM8 8-bit', 'EM_TILE64' : 'Tilera TILE64', 'EM_TILEPRO' : 'Tilera TILEPro', 'EM_MICROBLAZE' : 'Xilinx MicroBlaze 32-bit RISC', 'EM_CUDA' : 'NVIDIA CUDA', 'EM_TILEGX' : 'Tilera TILE-Gx', 'EM_CLOUDSHIELD' : 'CloudShield', 'EM_COREA_1ST' : 'KIPO-KAIST Core-A 1st generation', 'EM_COREA_2ND' : 'KIPO-KAIST Core-A 2nd generation', 'EM_ARC_COMPACT2' : 'Synopsys ARCompact V2', 'EM_OPEN8' : 'Open8 8-bit RISC', 'EM_RL78' : 'Renesas RL78', 'EM_VIDEOCORE5' : 'Broadcom VideoCore V', 'EM_78KOR' : 'Renesas 78KOR', 'EM_56800EX' : 'Freescale 56800EX', 'EM_BA1' : 'Beyond BA1', 'EM_BA2' : 'Beyond BA2', 'EM_XCORE' : 'XMOS xCORE', 'EM_MCHP_PIC' : 'Microchip 8-bit PIC', 'EM_INTEL205' : 'Reserved by Intel', 'EM_INTEL206' : 'Reserved by Intel', 'EM_INTEL207' : 'Reserved by Intel', 'EM_INTEL208' : 'Reserved by Intel', 'EM_INTEL209' : 'Reserved by Intel', 'EM_KM32' : 'KM211 KM32 32-bit', 'EM_KMX32' : 'KM211 KMX32 32-bit', 'EM_KMX16' : 'KM211 KMX16 16-bit', 'EM_KMX8' : 'KM211 KMX8 8-bit', 'EM_KVARC' : 'KM211 KVARC', 'EM_CDP' : 'Paneve CDP', 'EM_COGE' : 'Cognitive', 'EM_COOL' : 'Bluechip Systems CoolEngine', 'EM_NORC' : 'Nanoradio Optimized RISC', 'EM_CSR_KALIMBA' : 'CSR Kalimba', 'EM_Z80' : 'Zilog Z80', 'EM_VISIUM' : 'VISIUMcore', 'EM_FT32' : 'FTDI Chip FT32 32-bit RISC', 'EM_MOXIE' : 'Moxie', 'EM_AMDGPU' : 'AMD GPU', 'EM_RISCV' : 'RISC-V', 'EM_BPF' : 'Linux BPF - in-kernel virtual machine', 'EM_CSKY' : 'C-SKY', 'EM_LOONGARCH' : 'LoongArch', 'EM_FRV' : 'Fujitsu FR-V' } return architectures.get(self['e_machine'], '') def get_shstrndx(self): """ Find the string table section index for the section header table """ # From https://refspecs.linuxfoundation.org/elf/gabi4+/ch4.eheader.html: # If the section name string table section index is greater than or # equal to SHN_LORESERVE (0xff00), this member has the value SHN_XINDEX # (0xffff) and the actual index of the section name string table section # is contained in the sh_link field of the section header at index 0. if self['e_shstrndx'] != SHN_INDICES.SHN_XINDEX: return self['e_shstrndx'] else: return self._get_section_header(0)['sh_link'] #-------------------------------- PRIVATE --------------------------------# def __getitem__(self, name): """ Implement dict-like access to header entries """ return self.header[name] def _identify_file(self): """ Verify the ELF file and identify its class and endianness. """ # Note: this code reads the stream directly, without using ELFStructs, # since we don't yet know its exact format. ELF was designed to be # read like this - its e_ident field is word-size and endian agnostic. self.stream.seek(0) magic = self.stream.read(4) elf_assert(magic == b'\x7fELF', 'Magic number does not match') ei_class = self.stream.read(1) if ei_class == b'\x01': self.elfclass = 32 elif ei_class == b'\x02': self.elfclass = 64 else: raise ELFError('Invalid EI_CLASS %s' % repr(ei_class)) ei_data = self.stream.read(1) if ei_data == b'\x01': self.little_endian = True elif ei_data == b'\x02': self.little_endian = False else: raise ELFError('Invalid EI_DATA %s' % repr(ei_data)) def _section_offset(self, n): """ Compute the offset of section #n in the file """ return self['e_shoff'] + n * self['e_shentsize'] def _segment_offset(self, n): """ Compute the offset of segment #n in the file """ return self['e_phoff'] + n * self['e_phentsize'] def _make_segment(self, segment_header): """ Create a Segment object of the appropriate type """ segtype = segment_header['p_type'] if segtype == 'PT_INTERP': return InterpSegment(segment_header, self.stream) elif segtype == 'PT_DYNAMIC': return DynamicSegment(segment_header, self.stream, self) elif segtype == 'PT_NOTE': return NoteSegment(segment_header, self.stream, self) else: return Segment(segment_header, self.stream) def _get_section_header(self, n): """ Find the header of section #n, parse it and return the struct """ stream_pos = self._section_offset(n) if stream_pos > self.stream_len: return None return struct_parse( self.structs.Elf_Shdr, self.stream, stream_pos=stream_pos) def _get_section_name(self, section_header): """ Given a section header, find this section's name in the file's string table """ if self._section_header_stringtable is None: raise ELFParseError("String Table not found") name_offset = section_header['sh_name'] return self._section_header_stringtable.get_string(name_offset) def _make_section(self, section_header): """ Create a section object of the appropriate type """ name = self._get_section_name(section_header) sectype = section_header['sh_type'] if sectype == 'SHT_STRTAB': return StringTableSection(section_header, name, self) elif sectype == 'SHT_NULL': return NullSection(section_header, name, self) elif sectype in ('SHT_SYMTAB', 'SHT_DYNSYM', 'SHT_SUNW_LDYNSYM'): return self._make_symbol_table_section(section_header, name) elif sectype == 'SHT_SYMTAB_SHNDX': return self._make_symbol_table_index_section(section_header, name) elif sectype == 'SHT_SUNW_syminfo': return self._make_sunwsyminfo_table_section(section_header, name) elif sectype == 'SHT_GNU_verneed': return self._make_gnu_verneed_section(section_header, name) elif sectype == 'SHT_GNU_verdef': return self._make_gnu_verdef_section(section_header, name) elif sectype == 'SHT_GNU_versym': return self._make_gnu_versym_section(section_header, name) elif sectype in ('SHT_REL', 'SHT_RELA'): return RelocationSection(section_header, name, self) elif sectype == 'SHT_DYNAMIC': return DynamicSection(section_header, name, self) elif sectype == 'SHT_NOTE': return NoteSection(section_header, name, self) elif sectype == 'SHT_PROGBITS' and name == '.stab': return StabSection(section_header, name, self) elif sectype == 'SHT_ARM_ATTRIBUTES': return ARMAttributesSection(section_header, name, self) elif sectype == 'SHT_RISCV_ATTRIBUTES': return RISCVAttributesSection(section_header, name, self) elif sectype == 'SHT_HASH': return self._make_elf_hash_section(section_header, name) elif sectype == 'SHT_GNU_HASH': return self._make_gnu_hash_section(section_header, name) elif sectype == 'SHT_RELR': return RelrRelocationSection(section_header, name, self) else: return Section(section_header, name, self) def _make_section_name_map(self): self._section_name_map = {} for i, sec in enumerate(self.iter_sections()): self._section_name_map[sec.name] = i def _make_symbol_table_section(self, section_header, name): """ Create a SymbolTableSection """ linked_strtab_index = section_header['sh_link'] strtab_section = self.get_section(linked_strtab_index) return SymbolTableSection( section_header, name, elffile=self, stringtable=strtab_section) def _make_symbol_table_index_section(self, section_header, name): """ Create a SymbolTableIndexSection object """ linked_symtab_index = section_header['sh_link'] return SymbolTableIndexSection( section_header, name, elffile=self, symboltable=linked_symtab_index) def _make_sunwsyminfo_table_section(self, section_header, name): """ Create a SUNWSyminfoTableSection """ linked_strtab_index = section_header['sh_link'] strtab_section = self.get_section(linked_strtab_index) return SUNWSyminfoTableSection( section_header, name, elffile=self, symboltable=strtab_section) def _make_gnu_verneed_section(self, section_header, name): """ Create a GNUVerNeedSection """ linked_strtab_index = section_header['sh_link'] strtab_section = self.get_section(linked_strtab_index) return GNUVerNeedSection( section_header, name, elffile=self, stringtable=strtab_section) def _make_gnu_verdef_section(self, section_header, name): """ Create a GNUVerDefSection """ linked_strtab_index = section_header['sh_link'] strtab_section = self.get_section(linked_strtab_index) return GNUVerDefSection( section_header, name, elffile=self, stringtable=strtab_section) def _make_gnu_versym_section(self, section_header, name): """ Create a GNUVerSymSection """ linked_strtab_index = section_header['sh_link'] strtab_section = self.get_section(linked_strtab_index) return GNUVerSymSection( section_header, name, elffile=self, symboltable=strtab_section) def _make_elf_hash_section(self, section_header, name): linked_symtab_index = section_header['sh_link'] symtab_section = self.get_section(linked_symtab_index) return ELFHashSection( section_header, name, self, symtab_section ) def _make_gnu_hash_section(self, section_header, name): linked_symtab_index = section_header['sh_link'] symtab_section = self.get_section(linked_symtab_index) return GNUHashSection( section_header, name, self, symtab_section ) def _get_segment_header(self, n): """ Find the header of segment #n, parse it and return the struct """ return struct_parse( self.structs.Elf_Phdr, self.stream, stream_pos=self._segment_offset(n)) def _get_section_header_stringtable(self): """ Get the string table section corresponding to the section header table. """ stringtable_section_num = self.get_shstrndx() stringtable_section_header = self._get_section_header(stringtable_section_num) if stringtable_section_header is None: return None return StringTableSection( header=stringtable_section_header, name='', elffile=self) def _parse_elf_header(self): """ Parses the ELF file header and assigns the result to attributes of this object. """ return struct_parse(self.structs.Elf_Ehdr, self.stream, stream_pos=0) def _read_dwarf_section(self, section, relocate_dwarf_sections): """ Read the contents of a DWARF section from the stream and return a DebugSectionDescriptor. Apply relocations if asked to. """ # The section data is read into a new stream, for processing section_stream = BytesIO() section_stream.write(section.data()) if relocate_dwarf_sections: reloc_handler = RelocationHandler(self) reloc_section = reloc_handler.find_relocations_for_section(section) if reloc_section is not None: reloc_handler.apply_section_relocations( section_stream, reloc_section) return DebugSectionDescriptor( stream=section_stream, name=section.name, global_offset=section['sh_offset'], size=section.data_size, address=section['sh_addr']) @staticmethod def _decompress_dwarf_section(section): """ Returns the uncompressed contents of the provided DWARF section. """ # TODO: support other compression formats from readelf.c assert section.size > 12, 'Unsupported compression format.' section.stream.seek(0) # According to readelf.c the content should contain "ZLIB" # followed by the uncompressed section size - 8 bytes in # big-endian order compression_type = section.stream.read(4) assert compression_type == b'ZLIB', \ 'Invalid compression type: %r' % (compression_type) uncompressed_size = struct.unpack('>Q', section.stream.read(8))[0] decompressor = zlib.decompressobj() uncompressed_stream = BytesIO() while True: chunk = section.stream.read(PAGESIZE) if not chunk: break uncompressed_stream.write(decompressor.decompress(chunk)) uncompressed_stream.write(decompressor.flush()) uncompressed_stream.seek(0, io.SEEK_END) size = uncompressed_stream.tell() assert uncompressed_size == size, \ 'Wrong uncompressed size: expected %r, but got %r' % ( uncompressed_size, size, ) return section._replace(stream=uncompressed_stream, size=size) def close(self): self.stream.close() def __enter__(self): return self def __exit__(self, type, value, traceback): self.close()