You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
mem_edit/mem_edit/abstract.py

356 lines
13 KiB
Python

"""
Abstract class for cross-platform memory editing.
"""
from typing import List, Tuple, Optional, Union, Generator
from abc import ABCMeta, abstractmethod
from contextlib import contextmanager
import copy
import ctypes
import logging
from . import utils
from .utils import ctypes_buffer_t
logger = logging.getLogger(__name__)
class Process(metaclass=ABCMeta):
"""
This class is used to interact with processes running on the system
(i.e., by reading from or writing to the memory used by a given process).
The static methods
`Process.list_available_pids()`
`Process.get_pid_by_name(executable_filename)`
can be used to help find the process id (pid) of the target process. They are
provided for convenience only; it is probably better to use the tools built
in to your operating system to discover the pid of the specific process you
would like to edit.
Once you have found the pid, you are ready to construct an instance of Process
and use it to read and write to memory. Once you are done with the process,
use `.close()` to free up the process for access by other debuggers etc.
```
p = Process(1239)
p.close()
```
To read/write to memory, first create a buffer using ctypes:
```
buffer0 = (ctypes.c_byte * 5)(39, 50, 03, 40, 30)
buffer1 = ctypes.c_ulong()
```
and then use
```
p.write_memory(0x2fe, buffer0)
val0 = p.read_memory(0x220, buffer0)[:]
val1a = p.read_memory(0x149, buffer1).value
val2b = buffer1.value
assert(val1a == val2b)
```
Searching for a value can be done in a number of ways:
Search a list of addresses:
`found_addresses = p.search_addresses([0x1020, 0x1030], buffer0)`
Search the entire memory space:
`found_addresses = p.search_all_memory(buffer0, writeable_only=False)`
You can also get a list of which regions in memory are mapped (readable):
`regions = p.list_mapped_regions(writeable_only=False)`
which can be used along with search_buffer(...) to re-create .search_all_memory(...):
```
found = []
for region_start, region_stop in regions:
region_buffer = (ctypes.c_byte * (region_stop - region_start))()
p.read_memory(region_start, region_buffer)
found += utils.search_buffer(ctypes.c_ulong(123456790), region_buffer)
```
Other useful methods include the context manager, implemented as a static method:
```
with Process.open_process(pid) as p:
# use p here, no need to call p.close()
```
.get_path(), which reports the path of the executable file which was used
to start the process:
```
executable_path = p.get_path()
```
and deref_struct_pointer, which takes a pointer to a struct and reads out the struct members:
```
# struct is a list of (offset, buffer) pairs
struct_defintion = [(0x0, ctypes.c_ulong()),
(0x20, ctypes.c_byte())]
values = p.deref_struct_pointer(0x0feab4, struct_defintion)
```
which is shorthand for
```
struct_addr = p.read_memory(0x0feab4, ctypes.c_void_p())
values = [p.read_memory(struct_addr + 0x0, ctypes.c_ulong()),
p.read_memory(struct_addr + 0x20, ctypes.c_byte())]
```
=================
Putting all this together, a simple program which alters a magic number in the only running
instance of 'magic.exe' might look like this:
```
import ctypes
from mem_edit import Process
magic_number = ctypes.ulong(1234567890)
pid = Process.get_pid_by_name('magic.exe')
with Process.open_process(pid) as p:
addrs = p.search_all_memory(magic_number)
assert(len(addrs) == 1)
p.write_memory(addrs[0], ctypes.c_ulong(42))
```
Searching for a value which changes:
```
pid = Process.get_pid_by_name('monitor_me.exe')
with Process.open_process(pid) as p:
addrs = p.search_all_memory(ctypes.c_int(40))
input('Press enter when value has changed to 55')
filtered_addrs = p.search_addresses(addrs, ctypes.c_int(55))
print('Found addresses:')
for addr in filtered_addrs:
print(hex(addr))
```
"""
@abstractmethod
def __init__(self, process_id: int):
"""
Constructing a Process object prepares the process with specified process_id for
memory editing. Finding the `process_id` for the process you want to edit is often
easiest using os-specific tools (or by launching the process yourself, e.g. with
`subprocess.Popen(...)`).
Args:
process_id: Process id (pid) of the target process
"""
pass
@abstractmethod
def close(self):
"""
Detach from the process, removing our ability to edit it and
letting other debuggers attach to it instead.
This function should be called after you are done working with the process
and will no longer need it. See the `Process.open_process(...)` context
manager to avoid having to call this function yourself.
"""
pass
@abstractmethod
def write_memory(self, base_address: int, write_buffer: ctypes_buffer_t):
"""
Write the given buffer to the process's address space, starting at `base_address`.
Args:
base_address: The address to write at, in the process's address space.
write_buffer: A ctypes object, for example, `ctypes.c_ulong(48)`,
`(ctypes.c_byte * 3)(43, 21, 0xff)`, or a subclass of `ctypes.Structure`,
which will be written into memory starting at `base_address`.
"""
pass
@abstractmethod
def read_memory(self, base_address: int, read_buffer: ctypes_buffer_t) -> ctypes_buffer_t:
"""
Read into the given buffer from the process's address space, starting at `base_address`.
Args:
base_address: The address to read from, in the process's address space.
read_buffer: A `ctypes` object, for example. `ctypes.c_ulong()`,
`(ctypes.c_byte * 3)()`, or a subclass of `ctypes.Structure`, which will be
overwritten with the contents of the process's memory starting at `base_address`.
Returns:
`read_buffer` is returned as well as being overwritten.
"""
pass
@abstractmethod
def list_mapped_regions(self, writeable_only=True) -> List[Tuple[int, int]]:
"""
Return a list of `(start_address, stop_address)` for the regions of the address space
accessible to (readable and possibly writable by) the process.
By default, this function does not return non-writeable regions.
Args:
writeable_only: If `True`, only return regions which are also writeable.
Default `True`.
Returns:
List of `(start_address, stop_address)` for each accessible memory region.
"""
pass
@abstractmethod
def get_path(self) -> str:
"""
Return the path to the executable file which was run to start this process.
Returns:
A string containing the path.
"""
pass
@staticmethod
@abstractmethod
def list_available_pids() -> List[int]:
"""
Return a list of all process ids (pids) accessible on this system.
Returns:
List of running process ids.
"""
pass
@staticmethod
@abstractmethod
def get_pid_by_name(target_name: str) -> Optional[int]:
"""
Attempt to return the process id (pid) of a process which was run with an executable
file with the provided name. If no process is found, return None.
This is a convenience method for quickly finding a process which is already known
to be unique and has a well-defined executable name.
Don't rely on this method if you can possibly avoid it, since it makes no
attempt to confirm that it found a unique process and breaks trivially (e.g. if the
executable file is renamed).
Args:
target_name: Name of the process to find the PID for
Returns:
Process id (pid) of a process with the provided name, or `None`.
"""
pass
def deref_struct_pointer(self,
base_address: int,
targets: List[Tuple[int, ctypes_buffer_t]],
) -> List[ctypes_buffer_t]:
"""
Take a pointer to a struct and read out the struct members:
```
struct_defintion = [(0x0, ctypes.c_ulong()),
(0x20, ctypes.c_byte())]
values = p.deref_struct_pointer(0x0feab4, struct_defintion)
```
which is shorthand for
```
struct_addr = p.read_memory(0x0feab4, ctypes.c_void_p())
values = [p.read_memory(struct_addr + 0x0, ctypes.c_ulong()),
p.read_memory(struct_addr + 0x20, ctypes.c_byte())]
```
Args:
base_address: Address at which the struct pointer is located.
targets: List of `(offset, read_buffer)` pairs which will be read from the struct.
Return:
List of read values corresponding to the provided targets.
"""
base = self.read_memory(base_address, ctypes.c_void_p()).value
values = [self.read_memory(base + offset, buffer) for offset, buffer in targets]
return values
def search_addresses(self,
addresses: List[int],
needle_buffer: ctypes_buffer_t,
verbatim: bool = True,
) -> List[int]:
"""
Search for the provided value at each of the provided addresses, and return the addresses
where it is found.
Args:
addresses: List of addresses which should be probed.
needle_buffer: The value to search for. This should be a `ctypes` object of the same
sorts as used by `.read_memory(...)`, which will be compared to the contents of
memory at each of the given addresses.
verbatim: If `True`, perform bitwise comparison when searching for `needle_buffer`.
If `False`, perform `utils.ctypes_equal`-based comparison. Default `True`.
Returns:
List of addresses where the `needle_buffer` was found.
"""
found = []
read_buffer = copy.copy(needle_buffer)
if verbatim:
def compare(a, b):
return bytes(read_buffer) == bytes(needle_buffer)
else:
compare = utils.ctypes_equal
for address in addresses:
self.read_memory(address, read_buffer)
if compare(needle_buffer, read_buffer):
found.append(address)
return found
def search_all_memory(self,
needle_buffer: ctypes_buffer_t,
writeable_only: bool = True,
verbatim: bool = True,
) -> List[int]:
"""
Search the entire memory space accessible to the process for the provided value.
Args:
needle_buffer: The value to search for. This should be a ctypes object of the same
sorts as used by `.read_memory(...)`, which will be compared to the contents of
memory at each accessible address.
writeable_only: If `True`, only search regions where the process has write access.
Default `True`.
verbatim: If `True`, perform bitwise comparison when searching for `needle_buffer`.
If `False`, perform `utils.ctypes_equal-based` comparison. Default `True`.
Returns:
List of addresses where the `needle_buffer` was found.
"""
found = []
if verbatim:
search = utils.search_buffer_verbatim
else:
search = utils.search_buffer
for start, stop in self.list_mapped_regions(writeable_only):
try:
region_buffer = (ctypes.c_byte * (stop - start))()
self.read_memory(start, region_buffer)
found += [offset + start for offset in search(needle_buffer, region_buffer)]
except OSError:
logger.error('Failed to read in range 0x{} - 0x{}'.format(start, stop))
return found
@classmethod
@contextmanager
def open_process(cls, process_id: int) -> Generator['Process', None, None]:
"""
Context manager which automatically closes the constructed Process:
```
with Process.open_process(2394) as p:
# use p here
# no need to run p.close()
```
Args:
process_id: Process id (pid), passed to the Process constructor.
Returns:
Constructed Process object.
"""
process = cls(process_id)
yield process
process.close()