From 45356ee7b633d5f589ef61ec8e4b81409d21f873 Mon Sep 17 00:00:00 2001 From: Malfurious Date: Fri, 12 Jan 2024 07:37:53 -0500 Subject: builder: Rename package to payload and expose contents This follows in the package contents export change. Additionally, the builder package is renamed to "payload". "payload" is actually the preferred name of this package. It was previously renamed due to the absurdity of importing "sploit.payload.payload.Payload()", and the fact that additional modules were being bundled together so a more broad name _seemed_ desirable. Signed-off-by: Malfurious --- sploit/__init__.py | 2 +- sploit/builder/__init__.py | 5 - sploit/builder/gadhint.py | 109 ------------- sploit/builder/payload.py | 94 ----------- sploit/builder/rop.py | 383 --------------------------------------------- sploit/payload/__init__.py | 3 + sploit/payload/gadhint.py | 109 +++++++++++++ sploit/payload/payload.py | 94 +++++++++++ sploit/payload/rop.py | 383 +++++++++++++++++++++++++++++++++++++++++++++ 9 files changed, 590 insertions(+), 592 deletions(-) delete mode 100644 sploit/builder/__init__.py delete mode 100644 sploit/builder/gadhint.py delete mode 100644 sploit/builder/payload.py delete mode 100644 sploit/builder/rop.py create mode 100644 sploit/payload/__init__.py create mode 100644 sploit/payload/gadhint.py create mode 100644 sploit/payload/payload.py create mode 100644 sploit/payload/rop.py diff --git a/sploit/__init__.py b/sploit/__init__.py index 1eb570c..f761ead 100644 --- a/sploit/__init__.py +++ b/sploit/__init__.py @@ -1,8 +1,8 @@ from sploit import ( arch, - builder, comm, log, + payload, rev, symtbl, until, diff --git a/sploit/builder/__init__.py b/sploit/builder/__init__.py deleted file mode 100644 index 758d511..0000000 --- a/sploit/builder/__init__.py +++ /dev/null @@ -1,5 +0,0 @@ -from . import ( - gadhint, - payload, - rop, -) diff --git a/sploit/builder/gadhint.py b/sploit/builder/gadhint.py deleted file mode 100644 index 9b077fe..0000000 --- a/sploit/builder/gadhint.py +++ /dev/null @@ -1,109 +0,0 @@ -from dataclasses import dataclass, field -from sploit.rev.gadget import Gadget - -@dataclass -class GadHint: - """ - User-annotated gadget description object - - gadget (Gadget|int): The gadget being annotated. May be a Gadget object or - an offset as an int. - - pops (list[str]): The registers popped by this gadget, in order of - occurrence. - - movs (dict{str:str}): The register-to-register moves made by this gadget. - Keys are destination register names, values are source register names. The - order given is insignificant. - - imms (dict{str:int}): The immediate-to-register loads made by this gadget. - Keys are destination register names, values are immediate values. The order - given is insignificant. - - writes (dict{str:str}): The register-to-memory moves (stores) made by this - gadget. Keys are destination register names (expected to hold memory - locations), values are source register names (expected to hold direct - values). The order given is insignificant. - - requirements (dict{str:int}): The register state that is required before - this gadget should be executed. Keys are register names, values are the - required register values. - - stack (list[int]): A list of words to append to the stack following this - gadget. The first element given is nearest to the top of the stack and the - rest follow in order. - - align (bool): If True, this gadget expects the stack to be aligned prior - to entry. - - syscall (bool): If True, this gadget contains a syscall instruction. - - spm (int): "Stack pointer move" - The amount the stack pointer is adjusted - by this gadget. The effect of executing a terminating "return" instruction - should not be accounted for. A value of zero is taken as "unspecified". - """ - - gadget: int = 0 - pops: list = field(default_factory=list) - movs: dict = field(default_factory=dict) - imms: dict = field(default_factory=dict) - writes: dict = field(default_factory=dict) - requirements: dict = field(default_factory=dict) - stack: list = field(default_factory=list) - align: bool = False - syscall: bool = False - spm: int = 0 - - @property - def offset(self): - """Return gadget offset as an integer.""" - return int(self.gadget) - - def __index__(self): - """Convert object to integer using offset value.""" - return self.offset - - def __add__(self, x): - """Return new object with adjusted offset.""" - return GadHint(self.gadget + x, self.pops, self.movs, self.imms, - self.writes, self.requirements, self.stack, self.align, - self.syscall, self.spm) - - def __sub__(self, x): - """Return new object with adjusted offset.""" - return self + (-x) - - def with_requirements(self, reqs): - """Return new object with additional requirements.""" - for k, v in reqs.items(): - if self.requirements.get(k, v) != v: - raise ValueError( - f"GadHint: Conflicting gadget requirements: " - f"{self.requirements}, {reqs}") - - return GadHint(self.gadget, self.pops, self.movs, self.imms, - self.writes, self.requirements | reqs, self.stack, - self.align, self.syscall, self.spm) - - def __repr__(self): - """Return human-readable GadHint.""" - def fmt(name, prop): - if len(prop) > 0: - return f", {name}={prop}" - return "" - - s = hex(self.gadget) - s = f"Gadget({s})" if type(self.gadget) is Gadget else s - s += fmt("pops", self.pops) - s += fmt("movs", self.movs) - s += fmt("imms", self.imms) - s += fmt("writes", self.writes) - s += fmt("requirements", self.requirements) - s += fmt("stack", self.stack) - if self.align: - s += ", align" - if self.syscall: - s += ", syscall" - if self.spm > 0: - s += f", spm={self.spm}" - return f"GadHint({s})" diff --git a/sploit/builder/payload.py b/sploit/builder/payload.py deleted file mode 100644 index cf105c6..0000000 --- a/sploit/builder/payload.py +++ /dev/null @@ -1,94 +0,0 @@ -from sploit.arch import arch, itob -from sploit.symtbl import Symtbl - -class Payload: - MAGIC = b'\xef' - - def __init__(self, **kwargs): - self.payload = b'' - self.sym = Symtbl(**kwargs) - self.ctrs = {} - - def __len__(self): - return len(self.payload) - - def __call__(self, badbytes=b''): - found = [ hex(x) for x in set(self.payload).intersection(badbytes) ] - if len(found) > 0: - raise Exception(f'Payload: bad bytes in content: {found}') - return self.payload - - def _name(self, kind, sym): - if sym is not None: return sym - try: ctr = self.ctrs[kind] - except: ctr = 0 - self.ctrs[kind] = ctr + 1 - return f'{kind}_{ctr}' - - def _append(self, value, sym): - (self.sym @ 0)[sym] = len(self) - self.payload += value - return self - - def _prepend(self, value, sym): - self.sym >>= len(value) - (self.sym @ 0)[sym] = 0 - self.payload = value + self.payload - return self - - def end(self): - return self.sym.base + len(self) - - def bin(self, *values, sym=None): - return self._append(b''.join(values), sym=self._name('bin', sym)) - - def str(self, *values, sym=None): - values = [ v.encode() + b'\x00' for v in values ] - return self.bin(*values, sym=self._name('str', sym)) - - def int(self, *values, sym=None): - values = [ itob(v) for v in values ] - return self.bin(*values, sym=self._name('int', sym)) - - def int8(self, *values, sym=None): - values = [ itob(v, 1) for v in values ] - return self.bin(*values, sym=self._name('int', sym)) - - def int16(self, *values, sym=None): - values = [ itob(v, 2) for v in values ] - return self.bin(*values, sym=self._name('int', sym)) - - def int32(self, *values, sym=None): - values = [ itob(v, 4) for v in values ] - return self.bin(*values, sym=self._name('int', sym)) - - def int64(self, *values, sym=None): - values = [ itob(v, 8) for v in values ] - return self.bin(*values, sym=self._name('int', sym)) - - def ret(self, *values, sym=None): - return self.int(*values, sym=self._name('ret', sym)) - - def sbp(self, *values, sym=None): - if len(values) == 0: - return self.rep(self.MAGIC, arch.wordsize, sym=self._name('sbp', sym)) - return self.int(*values, sym=self._name('sbp', sym)) - - def rep(self, value, size, sym=None): - return self.bin(self._rep_helper(value, size), sym=self._name('rep', sym)) - - def pad(self, size, value=None, sym=None): - return self.bin(self._pad_helper(size, value), sym=self._name('pad', sym)) - - def pad_front(self, size, value=None, sym=None): - return self._prepend(self._pad_helper(size, value), sym=self._name('pad', sym)) - - def _rep_helper(self, value, size, *, explain=''): - if size < 0: - raise Exception(f'Payload: {explain}rep: available space is negative') - if (size := size / len(value)) != int(size): - raise Exception(f'Payload: {explain}rep: element does not divide the space evenly') - return value * int(size) - - def _pad_helper(self, size, value): - return self._rep_helper(value or arch.nopcode, size - len(self), explain='pad: ') diff --git a/sploit/builder/rop.py b/sploit/builder/rop.py deleted file mode 100644 index 7b58e0e..0000000 --- a/sploit/builder/rop.py +++ /dev/null @@ -1,383 +0,0 @@ -""" -ROP chain generation utilities - -This module contains tools for automating basic return-oriented-programming -workloads, such as loading register values and calling into arbitrary functions -or syscalls. The tools are currently designed to work on x86 (32 or 64 bit) -and ARM (32 bit only). - -The main appeal of the ROP class is the ability to abstract away the manual -construction of ROP chain data, and instead make declarative statements like -"call this function with these arguments." The ROP class will also utilize its -supplied binary objects to automatically find and use trivial gadgets. - -The user is able to provide annotations for more complicated gadgets, which help -instruct the class how to incorporate them into a ROP chain. This is done with -the GadHint dataclass. GadHint objects are provided to a ROP instance by -including them in the Symtbl of one of the binary objects it is constructed with. -If applicable, a user-supplied gadget will take precedence over automatic gadget -searching. - -See the GadHint class to learn more about the descriptive attributes that are -supported. -""" - -from graphlib import TopologicalSorter - -from sploit.arch import arch, btoi, itob -from sploit.builder.gadhint import GadHint -from sploit.builder.payload import Payload - -class ROP(Payload): - """ - ROP-enabled payload builder - - POP_MAGIC (int): Magic value used for pop instructions where no specific - value is required by the user. - - SPM_MAGIC (bytes): Magic value to fill the stack with when the best - available cleaning gadget is larger than is necessary. - - objects (list[ELF]): The binary objects this ROP instance will consider - for gadget searching. - - safe_syscalls (bool): If True, require that automatically found syscall - instructions are immediately followed by a return instruction. - - align_calls (bool): If True, ensure that the stack return address into - function calls is aligned according to the architecture alignment property. - Knowledge of alignment is taken from the instance Symtbl's base value. - - clean_stack (bool): If True, attempt to locate a cleaning gadget to "pop" - stack data that is leftover from a function call. Required if attempting - to make multiple calls that involve stack-based arguments. - """ - - POP_MAGIC = 0xdead - SPM_MAGIC = b'\x69' - - def __init__(self, *objects, safe_syscalls=True, align_calls=True, - clean_stack=True, **symbols): - """Initialize new ROP builder instance.""" - super().__init__(**symbols) - self.objects = objects - self.safe_syscalls = safe_syscalls - self.align_calls = align_calls - self.clean_stack = clean_stack - - def gadgets(self, *regexes, cont=False): - """Return a list of matching gadgets, considering all objects.""" - results = [] - for obj in self.objects: - results += obj.gadgets(*regexes, cont=cont) - return results - - def gadget(self, *regexes): - """Return the first matching gadget, considering all objects.""" - for obj in self.objects: - try: - return obj.gadget(*regexes) - except: - pass - raise LookupError( - f"ROP: Need to define gadget symbol for {'; '.join(regexes)}") - - def assign(self, *, sym=None, **sets): - """ - Insert a ROP chain to control given registers. - - **sets (str:int): Keyword arguments specify register assignments to - perform with this ROP chain. Argument names correspond to register - names. - - sym (str): If given, sym is the symbol name used to refer to the - inserted data. - """ - gadget = GadHint(0, requirements=sets) - return self._start_chain(gadget, sym=self._name("assign", sym)) - - def call(self, func, *params, sym=None): - """ - Insert a ROP chain to call function. - - func (int): Entry address of function to call. - *params (int): Remaining positional args are passed to func. - - sym (str): If given, sym is the symbol name used to refer to the - inserted data. - """ - register_params = dict(zip(arch.funcargs, params)) - stack_params = params[len(register_params):] - gadget = GadHint(func, requirements=register_params, stack=stack_params, - align=self.align_calls) - return self._start_chain(gadget, sym=self._name("call", sym)) - - def syscall(self, *params, sym=None): - """ - Insert a ROP chain to call kernel. - - *params (int): The first argument is the syscall number. Remaining - positional arguments are passed to the syscall. - - sym (str): If given, sym is the symbol name used to refer to the - inserted data. - """ - if len(params) > len(arch.kernargs): - raise TypeError("ROP: Too many arguments passed to syscall. " - "Target architecture supports up to {len(arch.kernargs)-1}.") - - register_params = dict(zip(arch.kernargs, params)) - gadget = self._get_gadget("syscall", {}).with_requirements(register_params) - return self._start_chain(gadget, sym=self._name("syscall", sym)) - - def memcpy(self, dst, src, *, sym=None): - """ - Insert a ROP chain to write data into memory. - - dst (int): The destination memory address. - src (bytes): The content to write. - - sym (str): If given, sym is the symbol name used to refer to the - inserted data. - """ - gadgets = [] - for idx in range(0, len(src), arch.wordsize): - g = self._get_write(dst + idx, btoi(src[idx:idx+arch.wordsize])) - gadgets.append(g) - return self._start_chain(*gadgets, sym=self._name("memcpy", sym)) - - def _get_hints(self): - """Return all user-supplied gadget hints.""" - return [h for obj in self.objects for _,h in obj.sym if type(h) is GadHint] - - def _discover_requirements(self, seen, graph, current): - """ - Populate gadget dependency graph. - - This function recursively looks up gadgets to ensure all necessary - required gadgets can be found, and stores this information into the - given graph object. Established dependencies encode the order that the - chain builder should attempt to satisfy register requirements. - Dependency loops are detected by the TopologicalSorter. - - seen (set): Set of (register,value) tuples we have already discovered. - graph (TopologicalSorter): Dependency graph model object. - current (GadHint): Current gadget we are processing. - """ - for r, v in current.requirements.items(): - # We key on register name _and_ value because some gadgets may - # only be capable of storing specific values in a target register. - # Requiring a register to store different values may require the - # use of multiple gadgets. - if (r, v) not in seen: - gadget = self._get_gadget(r, current.requirements) - - # Add gadget's requirements to the dependency graph. - # We say that each requirement is a 'successor' to this - # current gadget 'r', so that the chain builder will satisfy - # 'r' first. This prevents the fulfillment of 'r' from - # colbbering targets it requires, as the builder will satisfy - # them afterward. - for x in gadget.requirements: - graph.add(x, r) - - # Treat gadget's load immediates as pseudo-requirements for - # the sake of target ordering, following the same logic - # as above. - for x in gadget.imms: - graph.add(x, r) - - # Mark node as visited - seen.add((r, v)) - self._discover_requirements(seen, graph, gadget) - - def _get_gadget(self, target, sets): - """ - Get context-specific gadget. - - target (str): Either "ret", "syscall", or the name of a register we - would like to modify. - - sets (dict{str:int}): The set of other register requirements we are - trying to fulfill in parallel. Values may affect the gadget we decide - to use. - """ - # First, consider user-provided hints before automatically locating - # gadgets. - for hint in self._get_hints(): - # Setup additional requirements based on hint's register moves. - # If a mov target is in sets, require to set the src to the 'sets' - # value. - addl_reqs = { src:sets[dst] for dst, src in hint.movs.items() if dst in sets } - hint = hint.with_requirements(addl_reqs) - - # Pops will be accounted for by the chain builder. - # Immediates will be handled by gadget ordering in chain builder. - # Writes are a non-issue here. - - if hint.syscall: - # Only consider syscalls if the target is syscall. - if target == "syscall": - return hint - elif target in hint.imms: - if hint.imms[target] == sets[target]: - return hint - elif target in hint.pops: - return hint - elif target in hint.movs: - return hint - - # Automatically locate simple gadgets - if target == "ret": - return GadHint(self.gadget(arch.ret)) - - if target == "syscall": - insns = [arch.syscall, arch.ret] if self.safe_syscalls else [arch.syscall] - return GadHint(self.gadget(*insns), syscall=True) - - # target == register - insns = [ i.format(target) for i in arch.popgad ] - return GadHint(self.gadget(*insns), pops=[target], spm=arch.wordsize) - - def _get_clean(self, size): - """ - Get a stack cleaning gadget that moves sp by _at least_ size. - - size (int): Minimum stack pointer move. - """ - # spm values of zero (the default) can't be trusted, as in this case - # the user likely hasn't annotated the GadHint properly. Returning a - # larger move than requested is fine, since the chain builder can insert - # junk to be popped. - for hint in self._get_hints(): - if hint.spm >= size and hint.spm > 0: - return hint - - results = self.gadgets(*arch.cleangad) - table = { int(g.asm[0].group(1), 0): g for g in results } - sizes = sorted([ x for x in table.keys() if x >= size ]) - - if len(sizes) > 0: - return GadHint(table[sizes[0]], spm=sizes[0]) - - raise LookupError( - f"ROP: Need to define a stack move gadget of at least {size}") - - def _get_write(self, dst, src): - """ - Get a memory write gadget, injected with requirements for user data. - - dst (int): The intended memory write location. - src (int): The intended value to write. - """ - # If any exist, take the first write provided by user hints, assuming - # the user's intent to specifically use _this_ write. Follow-on gadgets - # to setup the dst and src registers must be findable. - for hint in self._get_hints(): - if hint.writes: - d, s = list(hint.writes.items())[0] - return hint.with_requirements({d:dst, s:src}) - - # Only take an automatic write gadget if we can prove up front that its - # requirements can be met, otherwise move on. A later search result may - # pass the test. - results = self.gadgets(*arch.writegad) - - for gad in results: - d = gad.asm[0].group("dst") - s = gad.asm[0].group("src") - - try: - # Assert requirements are met. - gadget = GadHint(gad, writes={d: s}, requirements={d:dst, s:src}) - self._discover_requirements(set(), TopologicalSorter(), gadget) - return gadget - except: - pass - - raise LookupError("ROP: Need to define gadgets for memory write / deps") - - def _start_chain(self, *gadgets, sym=None): - """ - Insert a generic ROP chain. - - *gadgets (GadHint): Annotated gadgets to prepare a chain from. - - sym (str): If given, sym is the symbol name used to refer to the - inserted data. - """ - stack = Payload(base=self.end()) - for g in gadgets: - self._build_chain(stack, g, {}) - return self.bin(stack(), sym=self._name("gadget", sym)) - - def _build_chain(self, stack, gadget, sets): - """ - Generate chain data for a given ROP gadget. - - This function recursively builds a ROP chain for the given gadget and - its requirements, storing data in the 'stack' object. - - stack (Payload): Stack data being constructed. - gadget (GadHint): Current gadget we are processing. - - sets (dict{str:int}): The set of other register requirements we are - trying to fulfill in parallel. - """ - # Form a to-do-list of registers from our immediate requirements, - # attempting to order them such that we avoid overwriting/conflicting - # values (this may not be possible). - reqs = gadget.requirements - graph = TopologicalSorter({ r:set() for r in reqs }) - self._discover_requirements(set(), graph, gadget) - to_do_list = [ x for x in graph.static_order() if x in reqs ] - - # Start chain by satisfying to-do-list requirements. - while len(to_do_list) > 0: - g = self._get_gadget(to_do_list[0], reqs) - self._build_chain(stack, g, reqs) - - # This gadget may satisfy multiple items in the to-do-list. - # Specifically, all of its pop and mov targets, and any load - # immediates that match our requirements. Non-matching - # immediates will be handled by a later gadget. - imms = g.imms.keys() & reqs.keys() - imms = [ x for x in imms if g.imms[x] == reqs[x] ] - done = g.pops + list(g.movs) + imms - to_do_list = [ x for x in to_do_list if x not in done ] - - # Append chain data to execute this gadget, but respect offset == 0 - # as a way to disable this gadget (perform a NULL gadget). - if gadget.offset != 0: - # Stack alignment if required. - if gadget.align: - align = -stack.end() % arch.alignment - stack.rep(itob(self._get_gadget("ret", {})), align) - - # "Return address" entry into this gadget. - stack.ret(gadget.offset) - - # The gadget's "inner stack data" will be values to be popped - # and additional junk data to be deallocated by the gadget itself. - sp_dest = len(stack) + gadget.spm - stack.int(*[ sets.get(p, self.POP_MAGIC) for p in gadget.pops ]) - if gadget.spm > 0: - stack.pad(sp_dest, self.SPM_MAGIC) - - # The gadget's "outer stack data" will be the additional values - # explicitly specified by the gadget. Append a separate gadget - # to clean up these values. - if len(gadget.stack) > 0: - size = len(gadget.stack) * arch.wordsize - - if self.clean_stack: - clean = self._get_clean(size) - stack.ret(clean) - sp_dest = len(stack) + clean.spm - else: - ret = self._get_gadget("ret", {}) - stack.ret(ret) - sp_dest = len(stack) + size - - stack.int(*gadget.stack) - stack.pad(sp_dest, self.SPM_MAGIC) diff --git a/sploit/payload/__init__.py b/sploit/payload/__init__.py new file mode 100644 index 0000000..78769b4 --- /dev/null +++ b/sploit/payload/__init__.py @@ -0,0 +1,3 @@ +from .gadhint import * +from .payload import * +from .rop import * diff --git a/sploit/payload/gadhint.py b/sploit/payload/gadhint.py new file mode 100644 index 0000000..9b077fe --- /dev/null +++ b/sploit/payload/gadhint.py @@ -0,0 +1,109 @@ +from dataclasses import dataclass, field +from sploit.rev.gadget import Gadget + +@dataclass +class GadHint: + """ + User-annotated gadget description object + + gadget (Gadget|int): The gadget being annotated. May be a Gadget object or + an offset as an int. + + pops (list[str]): The registers popped by this gadget, in order of + occurrence. + + movs (dict{str:str}): The register-to-register moves made by this gadget. + Keys are destination register names, values are source register names. The + order given is insignificant. + + imms (dict{str:int}): The immediate-to-register loads made by this gadget. + Keys are destination register names, values are immediate values. The order + given is insignificant. + + writes (dict{str:str}): The register-to-memory moves (stores) made by this + gadget. Keys are destination register names (expected to hold memory + locations), values are source register names (expected to hold direct + values). The order given is insignificant. + + requirements (dict{str:int}): The register state that is required before + this gadget should be executed. Keys are register names, values are the + required register values. + + stack (list[int]): A list of words to append to the stack following this + gadget. The first element given is nearest to the top of the stack and the + rest follow in order. + + align (bool): If True, this gadget expects the stack to be aligned prior + to entry. + + syscall (bool): If True, this gadget contains a syscall instruction. + + spm (int): "Stack pointer move" - The amount the stack pointer is adjusted + by this gadget. The effect of executing a terminating "return" instruction + should not be accounted for. A value of zero is taken as "unspecified". + """ + + gadget: int = 0 + pops: list = field(default_factory=list) + movs: dict = field(default_factory=dict) + imms: dict = field(default_factory=dict) + writes: dict = field(default_factory=dict) + requirements: dict = field(default_factory=dict) + stack: list = field(default_factory=list) + align: bool = False + syscall: bool = False + spm: int = 0 + + @property + def offset(self): + """Return gadget offset as an integer.""" + return int(self.gadget) + + def __index__(self): + """Convert object to integer using offset value.""" + return self.offset + + def __add__(self, x): + """Return new object with adjusted offset.""" + return GadHint(self.gadget + x, self.pops, self.movs, self.imms, + self.writes, self.requirements, self.stack, self.align, + self.syscall, self.spm) + + def __sub__(self, x): + """Return new object with adjusted offset.""" + return self + (-x) + + def with_requirements(self, reqs): + """Return new object with additional requirements.""" + for k, v in reqs.items(): + if self.requirements.get(k, v) != v: + raise ValueError( + f"GadHint: Conflicting gadget requirements: " + f"{self.requirements}, {reqs}") + + return GadHint(self.gadget, self.pops, self.movs, self.imms, + self.writes, self.requirements | reqs, self.stack, + self.align, self.syscall, self.spm) + + def __repr__(self): + """Return human-readable GadHint.""" + def fmt(name, prop): + if len(prop) > 0: + return f", {name}={prop}" + return "" + + s = hex(self.gadget) + s = f"Gadget({s})" if type(self.gadget) is Gadget else s + s += fmt("pops", self.pops) + s += fmt("movs", self.movs) + s += fmt("imms", self.imms) + s += fmt("writes", self.writes) + s += fmt("requirements", self.requirements) + s += fmt("stack", self.stack) + if self.align: + s += ", align" + if self.syscall: + s += ", syscall" + if self.spm > 0: + s += f", spm={self.spm}" + return f"GadHint({s})" diff --git a/sploit/payload/payload.py b/sploit/payload/payload.py new file mode 100644 index 0000000..cf105c6 --- /dev/null +++ b/sploit/payload/payload.py @@ -0,0 +1,94 @@ +from sploit.arch import arch, itob +from sploit.symtbl import Symtbl + +class Payload: + MAGIC = b'\xef' + + def __init__(self, **kwargs): + self.payload = b'' + self.sym = Symtbl(**kwargs) + self.ctrs = {} + + def __len__(self): + return len(self.payload) + + def __call__(self, badbytes=b''): + found = [ hex(x) for x in set(self.payload).intersection(badbytes) ] + if len(found) > 0: + raise Exception(f'Payload: bad bytes in content: {found}') + return self.payload + + def _name(self, kind, sym): + if sym is not None: return sym + try: ctr = self.ctrs[kind] + except: ctr = 0 + self.ctrs[kind] = ctr + 1 + return f'{kind}_{ctr}' + + def _append(self, value, sym): + (self.sym @ 0)[sym] = len(self) + self.payload += value + return self + + def _prepend(self, value, sym): + self.sym >>= len(value) + (self.sym @ 0)[sym] = 0 + self.payload = value + self.payload + return self + + def end(self): + return self.sym.base + len(self) + + def bin(self, *values, sym=None): + return self._append(b''.join(values), sym=self._name('bin', sym)) + + def str(self, *values, sym=None): + values = [ v.encode() + b'\x00' for v in values ] + return self.bin(*values, sym=self._name('str', sym)) + + def int(self, *values, sym=None): + values = [ itob(v) for v in values ] + return self.bin(*values, sym=self._name('int', sym)) + + def int8(self, *values, sym=None): + values = [ itob(v, 1) for v in values ] + return self.bin(*values, sym=self._name('int', sym)) + + def int16(self, *values, sym=None): + values = [ itob(v, 2) for v in values ] + return self.bin(*values, sym=self._name('int', sym)) + + def int32(self, *values, sym=None): + values = [ itob(v, 4) for v in values ] + return self.bin(*values, sym=self._name('int', sym)) + + def int64(self, *values, sym=None): + values = [ itob(v, 8) for v in values ] + return self.bin(*values, sym=self._name('int', sym)) + + def ret(self, *values, sym=None): + return self.int(*values, sym=self._name('ret', sym)) + + def sbp(self, *values, sym=None): + if len(values) == 0: + return self.rep(self.MAGIC, arch.wordsize, sym=self._name('sbp', sym)) + return self.int(*values, sym=self._name('sbp', sym)) + + def rep(self, value, size, sym=None): + return self.bin(self._rep_helper(value, size), sym=self._name('rep', sym)) + + def pad(self, size, value=None, sym=None): + return self.bin(self._pad_helper(size, value), sym=self._name('pad', sym)) + + def pad_front(self, size, value=None, sym=None): + return self._prepend(self._pad_helper(size, value), sym=self._name('pad', sym)) + + def _rep_helper(self, value, size, *, explain=''): + if size < 0: + raise Exception(f'Payload: {explain}rep: available space is negative') + if (size := size / len(value)) != int(size): + raise Exception(f'Payload: {explain}rep: element does not divide the space evenly') + return value * int(size) + + def _pad_helper(self, size, value): + return self._rep_helper(value or arch.nopcode, size - len(self), explain='pad: ') diff --git a/sploit/payload/rop.py b/sploit/payload/rop.py new file mode 100644 index 0000000..54226b4 --- /dev/null +++ b/sploit/payload/rop.py @@ -0,0 +1,383 @@ +""" +ROP chain generation utilities + +This module contains tools for automating basic return-oriented-programming +workloads, such as loading register values and calling into arbitrary functions +or syscalls. The tools are currently designed to work on x86 (32 or 64 bit) +and ARM (32 bit only). + +The main appeal of the ROP class is the ability to abstract away the manual +construction of ROP chain data, and instead make declarative statements like +"call this function with these arguments." The ROP class will also utilize its +supplied binary objects to automatically find and use trivial gadgets. + +The user is able to provide annotations for more complicated gadgets, which help +instruct the class how to incorporate them into a ROP chain. This is done with +the GadHint dataclass. GadHint objects are provided to a ROP instance by +including them in the Symtbl of one of the binary objects it is constructed with. +If applicable, a user-supplied gadget will take precedence over automatic gadget +searching. + +See the GadHint class to learn more about the descriptive attributes that are +supported. +""" + +from graphlib import TopologicalSorter + +from sploit.arch import arch, btoi, itob +from sploit.payload.gadhint import GadHint +from sploit.payload.payload import Payload + +class ROP(Payload): + """ + ROP-enabled payload builder + + POP_MAGIC (int): Magic value used for pop instructions where no specific + value is required by the user. + + SPM_MAGIC (bytes): Magic value to fill the stack with when the best + available cleaning gadget is larger than is necessary. + + objects (list[ELF]): The binary objects this ROP instance will consider + for gadget searching. + + safe_syscalls (bool): If True, require that automatically found syscall + instructions are immediately followed by a return instruction. + + align_calls (bool): If True, ensure that the stack return address into + function calls is aligned according to the architecture alignment property. + Knowledge of alignment is taken from the instance Symtbl's base value. + + clean_stack (bool): If True, attempt to locate a cleaning gadget to "pop" + stack data that is leftover from a function call. Required if attempting + to make multiple calls that involve stack-based arguments. + """ + + POP_MAGIC = 0xdead + SPM_MAGIC = b'\x69' + + def __init__(self, *objects, safe_syscalls=True, align_calls=True, + clean_stack=True, **symbols): + """Initialize new ROP builder instance.""" + super().__init__(**symbols) + self.objects = objects + self.safe_syscalls = safe_syscalls + self.align_calls = align_calls + self.clean_stack = clean_stack + + def gadgets(self, *regexes, cont=False): + """Return a list of matching gadgets, considering all objects.""" + results = [] + for obj in self.objects: + results += obj.gadgets(*regexes, cont=cont) + return results + + def gadget(self, *regexes): + """Return the first matching gadget, considering all objects.""" + for obj in self.objects: + try: + return obj.gadget(*regexes) + except: + pass + raise LookupError( + f"ROP: Need to define gadget symbol for {'; '.join(regexes)}") + + def assign(self, *, sym=None, **sets): + """ + Insert a ROP chain to control given registers. + + **sets (str:int): Keyword arguments specify register assignments to + perform with this ROP chain. Argument names correspond to register + names. + + sym (str): If given, sym is the symbol name used to refer to the + inserted data. + """ + gadget = GadHint(0, requirements=sets) + return self._start_chain(gadget, sym=self._name("assign", sym)) + + def call(self, func, *params, sym=None): + """ + Insert a ROP chain to call function. + + func (int): Entry address of function to call. + *params (int): Remaining positional args are passed to func. + + sym (str): If given, sym is the symbol name used to refer to the + inserted data. + """ + register_params = dict(zip(arch.funcargs, params)) + stack_params = params[len(register_params):] + gadget = GadHint(func, requirements=register_params, stack=stack_params, + align=self.align_calls) + return self._start_chain(gadget, sym=self._name("call", sym)) + + def syscall(self, *params, sym=None): + """ + Insert a ROP chain to call kernel. + + *params (int): The first argument is the syscall number. Remaining + positional arguments are passed to the syscall. + + sym (str): If given, sym is the symbol name used to refer to the + inserted data. + """ + if len(params) > len(arch.kernargs): + raise TypeError("ROP: Too many arguments passed to syscall. " + "Target architecture supports up to {len(arch.kernargs)-1}.") + + register_params = dict(zip(arch.kernargs, params)) + gadget = self._get_gadget("syscall", {}).with_requirements(register_params) + return self._start_chain(gadget, sym=self._name("syscall", sym)) + + def memcpy(self, dst, src, *, sym=None): + """ + Insert a ROP chain to write data into memory. + + dst (int): The destination memory address. + src (bytes): The content to write. + + sym (str): If given, sym is the symbol name used to refer to the + inserted data. + """ + gadgets = [] + for idx in range(0, len(src), arch.wordsize): + g = self._get_write(dst + idx, btoi(src[idx:idx+arch.wordsize])) + gadgets.append(g) + return self._start_chain(*gadgets, sym=self._name("memcpy", sym)) + + def _get_hints(self): + """Return all user-supplied gadget hints.""" + return [h for obj in self.objects for _,h in obj.sym if type(h) is GadHint] + + def _discover_requirements(self, seen, graph, current): + """ + Populate gadget dependency graph. + + This function recursively looks up gadgets to ensure all necessary + required gadgets can be found, and stores this information into the + given graph object. Established dependencies encode the order that the + chain builder should attempt to satisfy register requirements. + Dependency loops are detected by the TopologicalSorter. + + seen (set): Set of (register,value) tuples we have already discovered. + graph (TopologicalSorter): Dependency graph model object. + current (GadHint): Current gadget we are processing. + """ + for r, v in current.requirements.items(): + # We key on register name _and_ value because some gadgets may + # only be capable of storing specific values in a target register. + # Requiring a register to store different values may require the + # use of multiple gadgets. + if (r, v) not in seen: + gadget = self._get_gadget(r, current.requirements) + + # Add gadget's requirements to the dependency graph. + # We say that each requirement is a 'successor' to this + # current gadget 'r', so that the chain builder will satisfy + # 'r' first. This prevents the fulfillment of 'r' from + # colbbering targets it requires, as the builder will satisfy + # them afterward. + for x in gadget.requirements: + graph.add(x, r) + + # Treat gadget's load immediates as pseudo-requirements for + # the sake of target ordering, following the same logic + # as above. + for x in gadget.imms: + graph.add(x, r) + + # Mark node as visited + seen.add((r, v)) + self._discover_requirements(seen, graph, gadget) + + def _get_gadget(self, target, sets): + """ + Get context-specific gadget. + + target (str): Either "ret", "syscall", or the name of a register we + would like to modify. + + sets (dict{str:int}): The set of other register requirements we are + trying to fulfill in parallel. Values may affect the gadget we decide + to use. + """ + # First, consider user-provided hints before automatically locating + # gadgets. + for hint in self._get_hints(): + # Setup additional requirements based on hint's register moves. + # If a mov target is in sets, require to set the src to the 'sets' + # value. + addl_reqs = { src:sets[dst] for dst, src in hint.movs.items() if dst in sets } + hint = hint.with_requirements(addl_reqs) + + # Pops will be accounted for by the chain builder. + # Immediates will be handled by gadget ordering in chain builder. + # Writes are a non-issue here. + + if hint.syscall: + # Only consider syscalls if the target is syscall. + if target == "syscall": + return hint + elif target in hint.imms: + if hint.imms[target] == sets[target]: + return hint + elif target in hint.pops: + return hint + elif target in hint.movs: + return hint + + # Automatically locate simple gadgets + if target == "ret": + return GadHint(self.gadget(arch.ret)) + + if target == "syscall": + insns = [arch.syscall, arch.ret] if self.safe_syscalls else [arch.syscall] + return GadHint(self.gadget(*insns), syscall=True) + + # target == register + insns = [ i.format(target) for i in arch.popgad ] + return GadHint(self.gadget(*insns), pops=[target], spm=arch.wordsize) + + def _get_clean(self, size): + """ + Get a stack cleaning gadget that moves sp by _at least_ size. + + size (int): Minimum stack pointer move. + """ + # spm values of zero (the default) can't be trusted, as in this case + # the user likely hasn't annotated the GadHint properly. Returning a + # larger move than requested is fine, since the chain builder can insert + # junk to be popped. + for hint in self._get_hints(): + if hint.spm >= size and hint.spm > 0: + return hint + + results = self.gadgets(*arch.cleangad) + table = { int(g.asm[0].group(1), 0): g for g in results } + sizes = sorted([ x for x in table.keys() if x >= size ]) + + if len(sizes) > 0: + return GadHint(table[sizes[0]], spm=sizes[0]) + + raise LookupError( + f"ROP: Need to define a stack move gadget of at least {size}") + + def _get_write(self, dst, src): + """ + Get a memory write gadget, injected with requirements for user data. + + dst (int): The intended memory write location. + src (int): The intended value to write. + """ + # If any exist, take the first write provided by user hints, assuming + # the user's intent to specifically use _this_ write. Follow-on gadgets + # to setup the dst and src registers must be findable. + for hint in self._get_hints(): + if hint.writes: + d, s = list(hint.writes.items())[0] + return hint.with_requirements({d:dst, s:src}) + + # Only take an automatic write gadget if we can prove up front that its + # requirements can be met, otherwise move on. A later search result may + # pass the test. + results = self.gadgets(*arch.writegad) + + for gad in results: + d = gad.asm[0].group("dst") + s = gad.asm[0].group("src") + + try: + # Assert requirements are met. + gadget = GadHint(gad, writes={d: s}, requirements={d:dst, s:src}) + self._discover_requirements(set(), TopologicalSorter(), gadget) + return gadget + except: + pass + + raise LookupError("ROP: Need to define gadgets for memory write / deps") + + def _start_chain(self, *gadgets, sym=None): + """ + Insert a generic ROP chain. + + *gadgets (GadHint): Annotated gadgets to prepare a chain from. + + sym (str): If given, sym is the symbol name used to refer to the + inserted data. + """ + stack = Payload(base=self.end()) + for g in gadgets: + self._build_chain(stack, g, {}) + return self.bin(stack(), sym=self._name("gadget", sym)) + + def _build_chain(self, stack, gadget, sets): + """ + Generate chain data for a given ROP gadget. + + This function recursively builds a ROP chain for the given gadget and + its requirements, storing data in the 'stack' object. + + stack (Payload): Stack data being constructed. + gadget (GadHint): Current gadget we are processing. + + sets (dict{str:int}): The set of other register requirements we are + trying to fulfill in parallel. + """ + # Form a to-do-list of registers from our immediate requirements, + # attempting to order them such that we avoid overwriting/conflicting + # values (this may not be possible). + reqs = gadget.requirements + graph = TopologicalSorter({ r:set() for r in reqs }) + self._discover_requirements(set(), graph, gadget) + to_do_list = [ x for x in graph.static_order() if x in reqs ] + + # Start chain by satisfying to-do-list requirements. + while len(to_do_list) > 0: + g = self._get_gadget(to_do_list[0], reqs) + self._build_chain(stack, g, reqs) + + # This gadget may satisfy multiple items in the to-do-list. + # Specifically, all of its pop and mov targets, and any load + # immediates that match our requirements. Non-matching + # immediates will be handled by a later gadget. + imms = g.imms.keys() & reqs.keys() + imms = [ x for x in imms if g.imms[x] == reqs[x] ] + done = g.pops + list(g.movs) + imms + to_do_list = [ x for x in to_do_list if x not in done ] + + # Append chain data to execute this gadget, but respect offset == 0 + # as a way to disable this gadget (perform a NULL gadget). + if gadget.offset != 0: + # Stack alignment if required. + if gadget.align: + align = -stack.end() % arch.alignment + stack.rep(itob(self._get_gadget("ret", {})), align) + + # "Return address" entry into this gadget. + stack.ret(gadget.offset) + + # The gadget's "inner stack data" will be values to be popped + # and additional junk data to be deallocated by the gadget itself. + sp_dest = len(stack) + gadget.spm + stack.int(*[ sets.get(p, self.POP_MAGIC) for p in gadget.pops ]) + if gadget.spm > 0: + stack.pad(sp_dest, self.SPM_MAGIC) + + # The gadget's "outer stack data" will be the additional values + # explicitly specified by the gadget. Append a separate gadget + # to clean up these values. + if len(gadget.stack) > 0: + size = len(gadget.stack) * arch.wordsize + + if self.clean_stack: + clean = self._get_clean(size) + stack.ret(clean) + sp_dest = len(stack) + clean.spm + else: + ret = self._get_gadget("ret", {}) + stack.ret(ret) + sp_dest = len(stack) + size + + stack.int(*gadget.stack) + stack.pad(sp_dest, self.SPM_MAGIC) -- cgit v1.2.3