esp-idf/components/espcoredump/corefile/elf.py

382 wiersze
12 KiB
Python

#
# Copyright 2021 Espressif Systems (Shanghai) CO., LTD
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import hashlib
import os
from construct import (AlignedStruct, Bytes, Const, Container, GreedyRange, Int16ul, Int32ul, Padding, Pointer,
Sequence, Struct, this)
try:
from typing import Optional
except ImportError:
pass
# Following structs are based on spec
# https://refspecs.linuxfoundation.org/elf/elf.pdf
# and source code
# IDF_PATH/components/espcoredump/include_core_dump/elf.h
ElfIdentification = Struct(
'EI_MAG' / Const(b'\x7fELF'),
'EI_CLASS' / Const(b'\x01'), # ELFCLASS32
'EI_DATA' / Const(b'\x01'), # ELFDATA2LSB
'EI_VERSION' / Const(b'\x01'), # EV_CURRENT
Padding(9),
)
ElfHeader = Struct(
'e_ident' / ElfIdentification,
'e_type' / Int16ul,
'e_machine' / Int16ul,
'e_version' / Int32ul,
'e_entry' / Int32ul,
'e_phoff' / Int32ul,
'e_shoff' / Int32ul,
'e_flags' / Int32ul,
'e_ehsize' / Int16ul,
'e_phentsize' / Int16ul,
'e_phnum' / Int16ul,
'e_shentsize' / Int16ul,
'e_shnum' / Int16ul,
'e_shstrndx' / Int16ul,
)
SectionHeader = Struct(
'sh_name' / Int32ul,
'sh_type' / Int32ul,
'sh_flags' / Int32ul,
'sh_addr' / Int32ul,
'sh_offset' / Int32ul,
'sh_size' / Int32ul,
'sh_link' / Int32ul,
'sh_info' / Int32ul,
'sh_addralign' / Int32ul,
'sh_entsize' / Int32ul,
)
ProgramHeader = Struct(
'p_type' / Int32ul,
'p_offset' / Int32ul,
'p_vaddr' / Int32ul,
'p_paddr' / Int32ul,
'p_filesz' / Int32ul,
'p_memsz' / Int32ul,
'p_flags' / Int32ul,
'p_align' / Int32ul,
)
ElfHeaderTables = Struct(
'elf_header' / ElfHeader,
'program_headers' / Pointer(this.elf_header.e_phoff, ProgramHeader[this.elf_header.e_phnum]),
'section_headers' / Pointer(this.elf_header.e_shoff, SectionHeader[this.elf_header.e_shnum]),
)
NoteSection = AlignedStruct(
4,
'namesz' / Int32ul,
'descsz' / Int32ul,
'type' / Int32ul,
'name' / Bytes(this.namesz),
'desc' / Bytes(this.descsz),
)
NoteSections = GreedyRange(NoteSection)
class ElfFile(object):
"""
Elf class to a single elf file
"""
SHN_UNDEF = 0x00
SHT_PROGBITS = 0x01
SHT_STRTAB = 0x03
SHT_NOBITS = 0x08
PT_LOAD = 0x01
PT_NOTE = 0x04
ET_CORE = 0x04
EV_CURRENT = 0x01
def __init__(self, elf_path=None, e_type=None, e_machine=None):
# type: (Optional[str], Optional[int], Optional[int]) -> None
self.e_type = e_type
self.e_machine = e_machine
self._struct = None # type: Optional[Struct]
self._model = None # type: Optional[Container]
self._section_names = {} # type: dict[int, str]
self.sections = [] # type: list[ElfSection]
self.load_segments = [] # type: list[ElfSegment]
self.note_segments = [] # type: list[ElfNoteSegment]
if elf_path and os.path.isfile(elf_path):
self.read_elf(elf_path)
def read_elf(self, elf_path): # type: (str) -> None
"""
Read elf file, also write to ``self.model``, ``self.program_headers``,
``self.section_headers``
:param elf_path: elf file path
:return: None
"""
with open(elf_path, 'rb') as fr:
elf_bytes = fr.read()
header_tables = ElfHeaderTables.parse(elf_bytes)
self.e_type = header_tables.elf_header.e_type
self.e_machine = header_tables.elf_header.e_machine
self._struct = self._generate_struct_from_headers(header_tables)
self._model = self._struct.parse(elf_bytes)
if 'string_table' in self._model:
self._section_names = self._parse_string_table(self._model.string_table)
self.load_segments = [ElfSegment(seg.ph.p_vaddr,
seg.data,
seg.ph.p_flags) for seg in self._model.load_segments]
self.note_segments = [ElfNoteSegment(seg.ph.p_vaddr,
seg.data,
seg.ph.p_flags) for seg in self._model.note_segments]
self.sections = [ElfSection(self._section_names[sec.sh.sh_name],
sec.sh.sh_addr,
sec.data,
sec.sh.sh_flags) for sec in self._model.sections]
@staticmethod
def _parse_string_table(byte_str): # type: (bytes) -> dict
name = ''
index = 0
res = {}
for i, c in enumerate(byte_str):
if c in [0x00, '\x00']: # a workaround for python 2 bytes is actually string
res[index] = name
name = ''
index = i + 1
continue
if isinstance(c, int):
name += chr(c)
else:
name += c
return res
def _generate_struct_from_headers(self, header_tables): # type: (Container) -> Struct
"""
Generate ``construct`` Struct for this file
:param header_tables: contains elf_header, program_headers, section_headers
:return: Struct of the whole file
"""
elf_header = header_tables.elf_header
program_headers = header_tables.program_headers
section_headers = header_tables.section_headers
assert program_headers or section_headers
string_table_sh = None
load_segment_subcons = []
note_segment_subcons = []
# Here we point back to make segments know their program headers
for i, ph in enumerate(program_headers):
args = [
'ph' / Pointer(elf_header.e_phoff + i * ProgramHeader.sizeof(), ProgramHeader),
'data' / Pointer(ph.p_offset, Bytes(ph.p_filesz)),
]
if ph.p_vaddr == 0 and ph.p_type == self.PT_NOTE:
args.append('note_secs' / Pointer(ph.p_offset, NoteSections))
note_segment_subcons.append(Struct(*args))
elif ph.p_vaddr != 0:
load_segment_subcons.append(Struct(*args))
section_subcons = []
for i, sh in enumerate(section_headers):
if sh.sh_type == self.SHT_STRTAB and i == elf_header.e_shstrndx:
string_table_sh = sh
elif sh.sh_addr != 0 and sh.sh_type == self.SHT_PROGBITS:
section_subcons.append(Struct(
'sh' / Pointer(elf_header.e_shoff + i * SectionHeader.sizeof(), SectionHeader),
'data' / Pointer(sh.sh_offset, Bytes(sh.sh_size)),
))
args = [
'elf_header' / ElfHeader,
'load_segments' / Sequence(*load_segment_subcons),
'note_segments' / Sequence(*note_segment_subcons),
'sections' / Sequence(*section_subcons),
]
if string_table_sh is not None:
args.append('string_table' / Pointer(string_table_sh.sh_offset, Bytes(string_table_sh.sh_size)))
return Struct(*args)
@property
def sha256(self): # type: () -> bytes
"""
:return: SHA256 hash of the input ELF file
"""
sha256 = hashlib.sha256()
sha256.update(self._struct.build(self._model)) # type: ignore
return sha256.digest()
class ElfSection(object):
SHF_WRITE = 0x01
SHF_ALLOC = 0x02
SHF_EXECINSTR = 0x04
SHF_MASKPROC = 0xf0000000
def __init__(self, name, addr, data, flags): # type: (str, int, bytes, int) -> None
self.name = name
self.addr = addr
self.data = data
self.flags = flags
def attr_str(self): # type: () -> str
if self.flags & self.SHF_MASKPROC:
return 'MS'
res = 'R'
res += 'W' if self.flags & self.SHF_WRITE else ' '
res += 'X' if self.flags & self.SHF_EXECINSTR else ' '
res += 'A' if self.flags & self.SHF_ALLOC else ' '
return res
def __repr__(self): # type: () -> str
return '{:>32} [Addr] 0x{:>08X}, [Size] 0x{:>08X} {:>4}' \
.format(self.name, self.addr, len(self.data), self.attr_str())
class ElfSegment(object):
PF_X = 0x01
PF_W = 0x02
PF_R = 0x04
def __init__(self, addr, data, flags): # type: (int, bytes, int) -> None
self.addr = addr
self.data = data
self.flags = flags
self.type = ElfFile.PT_LOAD
def attr_str(self): # type: () -> str
res = ''
res += 'R' if self.flags & self.PF_R else ' '
res += 'W' if self.flags & self.PF_W else ' '
res += 'E' if self.flags & self.PF_X else ' '
return res
@staticmethod
def _type_str(): # type: () -> str
return 'LOAD'
def __repr__(self): # type: () -> str
return '{:>8} Addr 0x{:>08X}, Size 0x{:>08X} Flags {:4}' \
.format(self._type_str(), self.addr, len(self.data), self.attr_str())
class ElfNoteSegment(ElfSegment):
def __init__(self, addr, data, flags): # type: (int, bytes, int) -> None
super(ElfNoteSegment, self).__init__(addr, data, flags)
self.type = ElfFile.PT_NOTE
self.note_secs = NoteSections.parse(self.data)
@staticmethod
def _type_str(): # type: () -> str
return 'NOTE'
TASK_STATUS_CORRECT = 0x00
TASK_STATUS_TCB_CORRUPTED = 0x01
TASK_STATUS_STACK_CORRUPTED = 0x02
EspTaskStatus = Struct(
'task_index' / Int32ul,
'task_flags' / Int32ul,
'task_tcb_addr' / Int32ul,
'task_stack_start' / Int32ul,
'task_stack_len' / Int32ul,
'task_name' / Bytes(16),
)
class ESPCoreDumpElfFile(ElfFile):
PT_INFO = 8266
PT_TASK_INFO = 678
PT_EXTRA_INFO = 677
CURR_TASK_MARKER = 0xdeadbeef
# ELF file machine type
EM_XTENSA = 0x5E
EM_RISCV = 0xF3
def __init__(self, elf_path=None, e_type=None, e_machine=None):
# type: (Optional[str], Optional[int], Optional[int]) -> None
_e_type = e_type or self.ET_CORE
_e_machine = e_machine or self.EM_XTENSA
super(ESPCoreDumpElfFile, self).__init__(elf_path, _e_type, _e_machine)
def add_segment(self, addr, data, seg_type, flags): # type: (int, bytes, int, int) -> None
if seg_type != self.PT_NOTE:
self.load_segments.append(ElfSegment(addr, data, flags))
else:
self.note_segments.append(ElfNoteSegment(addr, data, flags))
def dump(self, output_path): # type: (str) -> None
"""
Dump self.model into file
:param output_path: output file path
:return: None
"""
res = b''
res += ElfHeader.build({
'e_type': self.e_type,
'e_machine': self.e_machine,
'e_version': self.EV_CURRENT,
'e_entry': 0,
'e_phoff': ElfHeader.sizeof(),
'e_shoff': 0,
'e_flags': 0,
'e_ehsize': ElfHeader.sizeof(),
'e_phentsize': ProgramHeader.sizeof(),
'e_phnum': len(self.load_segments) + len(self.note_segments),
'e_shentsize': 0,
'e_shnum': 0,
'e_shstrndx': self.SHN_UNDEF,
})
offset = ElfHeader.sizeof() + (len(self.load_segments) + len(self.note_segments)) * ProgramHeader.sizeof()
_segments = self.load_segments + self.note_segments # type: ignore
for seg in _segments:
res += ProgramHeader.build({
'p_type': seg.type,
'p_offset': offset,
'p_vaddr': seg.addr,
'p_paddr': seg.addr,
'p_filesz': len(seg.data),
'p_memsz': len(seg.data),
'p_flags': seg.flags,
'p_align': 0,
})
offset += len(seg.data)
for seg in _segments:
res += seg.data
with open(output_path, 'wb') as fw:
fw.write(res)