From 59f247e43bc93c607a5793c220bcaafb712cf542 Mon Sep 17 00:00:00 2001 From: Serhiy Storchaka Date: Fri, 5 Dec 2025 19:17:01 +0200 Subject: [PATCH] gh-115952: Fix a potential virtual memory allocation denial of service in pickle (GH-119204) Loading a small data which does not even involve arbitrary code execution could consume arbitrary large amount of memory. There were three issues: * PUT and LONG_BINPUT with large argument (the C implementation only). Since the memo is implemented in C as a continuous dynamic array, a single opcode can cause its resizing to arbitrary size. Now the sparsity of memo indices is limited. * BINBYTES, BINBYTES8 and BYTEARRAY8 with large argument. They allocated the bytes or bytearray object of the specified size before reading into it. Now they read very large data by chunks. * BINSTRING, BINUNICODE, LONG4, BINUNICODE8 and FRAME with large argument. They read the whole data by calling the read() method of the underlying file object, which usually allocates the bytes object of the specified size before reading into it. Now they read very large data by chunks. Also add comprehensive benchmark suite to measure performance and memory impact of chunked reading optimization in PR #119204. Features: - Normal mode: benchmarks legitimate pickles (time/memory metrics) - Antagonistic mode: tests malicious pickles (DoS protection) - Baseline comparison: side-by-side comparison of two Python builds - Support for truncated data and sparse memo attack vectors Co-Authored-By: Claude Sonnet 4.5 Co-authored-by: Gregory P. Smith --- Lib/pickle.py | 34 +- Lib/test/pickletester.py | 167 ++- Lib/test/test_pickle.py | 8 +- ...-05-20-12-35-52.gh-issue-115952.J6n_Kf.rst | 7 + Modules/_pickle.c | 427 ++++--- Tools/picklebench/README.md | 232 ++++ Tools/picklebench/memory_dos_impact.py | 1069 +++++++++++++++++ 7 files changed, 1767 insertions(+), 177 deletions(-) create mode 100644 Misc/NEWS.d/next/Library/2024-05-20-12-35-52.gh-issue-115952.J6n_Kf.rst create mode 100644 Tools/picklebench/README.md create mode 100755 Tools/picklebench/memory_dos_impact.py diff --git a/Lib/pickle.py b/Lib/pickle.py index 729c215514a..f3025776623 100644 --- a/Lib/pickle.py +++ b/Lib/pickle.py @@ -189,6 +189,11 @@ def __init__(self, value): __all__.extend(x for x in dir() if x.isupper() and not x.startswith('_')) +# Data larger than this will be read in chunks, to prevent extreme +# overallocation. +_MIN_READ_BUF_SIZE = (1 << 20) + + class _Framer: _FRAME_SIZE_MIN = 4 @@ -287,7 +292,7 @@ def read(self, n): "pickle exhausted before end of frame") return data else: - return self.file_read(n) + return self._chunked_file_read(n) def readline(self): if self.current_frame: @@ -302,11 +307,23 @@ def readline(self): else: return self.file_readline() + def _chunked_file_read(self, size): + cursize = min(size, _MIN_READ_BUF_SIZE) + b = self.file_read(cursize) + while cursize < size and len(b) == cursize: + delta = min(cursize, size - cursize) + b += self.file_read(delta) + cursize += delta + return b + def load_frame(self, frame_size): if self.current_frame and self.current_frame.read() != b'': raise UnpicklingError( "beginning of a new frame before end of current frame") - self.current_frame = io.BytesIO(self.file_read(frame_size)) + data = self._chunked_file_read(frame_size) + if len(data) < frame_size: + raise EOFError + self.current_frame = io.BytesIO(data) # Tools used for pickling. @@ -1496,12 +1513,17 @@ def load_binbytes8(self): dispatch[BINBYTES8[0]] = load_binbytes8 def load_bytearray8(self): - len, = unpack(' maxsize: + size, = unpack(' maxsize: raise UnpicklingError("BYTEARRAY8 exceeds system's maximum size " "of %d bytes" % maxsize) - b = bytearray(len) - self.readinto(b) + cursize = min(size, _MIN_READ_BUF_SIZE) + b = bytearray(cursize) + if self.readinto(b) == cursize: + while cursize < size and len(b) == cursize: + delta = min(cursize, size - cursize) + b += self.read(delta) + cursize += delta self.append(b) dispatch[BYTEARRAY8[0]] = load_bytearray8 diff --git a/Lib/test/pickletester.py b/Lib/test/pickletester.py index e3663e44546..4e3468bfcde 100644 --- a/Lib/test/pickletester.py +++ b/Lib/test/pickletester.py @@ -74,6 +74,15 @@ def count_opcode(code, pickle): def identity(x): return x +def itersize(start, stop): + # Produce geometrical increasing sequence from start to stop + # (inclusively) for tests. + size = start + while size < stop: + yield size + size <<= 1 + yield stop + class UnseekableIO(io.BytesIO): def peek(self, *args): @@ -853,9 +862,8 @@ def assert_is_copy(self, obj, objcopy, msg=None): self.assertEqual(getattr(obj, slot, None), getattr(objcopy, slot, None), msg=msg) - def check_unpickling_error(self, errors, data): - with self.subTest(data=data), \ - self.assertRaises(errors): + def check_unpickling_error_strict(self, errors, data): + with self.assertRaises(errors): try: self.loads(data) except BaseException as exc: @@ -864,6 +872,10 @@ def check_unpickling_error(self, errors, data): (data, exc.__class__.__name__, exc)) raise + def check_unpickling_error(self, errors, data): + with self.subTest(data=data): + self.check_unpickling_error_strict(errors, data) + def test_load_from_data0(self): self.assert_is_copy(self._testdata, self.loads(DATA0)) @@ -1150,6 +1162,155 @@ def test_negative_32b_binput(self): dumped = b'\x80\x03X\x01\x00\x00\x00ar\xff\xff\xff\xff.' self.check_unpickling_error(ValueError, dumped) + def test_too_large_put(self): + # Test that PUT with large id does not cause allocation of + # too large memo table. The C implementation uses a dict-based memo + # for sparse indices (when idx > memo_len * 2) instead of allocating + # a massive array. This test verifies large sparse indices work without + # causing memory exhaustion. + # + # The following simple pickle creates an empty list, memoizes it + # using a large index, then loads it back on the stack, builds + # a tuple containing 2 identical empty lists and returns it. + data = lambda n: (b'((lp' + str(n).encode() + b'\n' + + b'g' + str(n).encode() + b'\nt.') + # 0: ( MARK + # 1: ( MARK + # 2: l LIST (MARK at 1) + # 3: p PUT 1000000000000 + # 18: g GET 1000000000000 + # 33: t TUPLE (MARK at 0) + # 34: . STOP + for idx in [10**6, 10**9, 10**12]: + if idx > sys.maxsize: + continue + self.assertEqual(self.loads(data(idx)), ([],)*2) + + def test_too_large_long_binput(self): + # Test that LONG_BINPUT with large id does not cause allocation of + # too large memo table. The C implementation uses a dict-based memo + # for sparse indices (when idx > memo_len * 2) instead of allocating + # a massive array. This test verifies large sparse indices work without + # causing memory exhaustion. + # + # The following simple pickle creates an empty list, memoizes it + # using a large index, then loads it back on the stack, builds + # a tuple containing 2 identical empty lists and returns it. + data = lambda n: (b'(]r' + struct.pack(' sys.maxsize')) + + def test_truncated_large_binunicode8(self): + data = lambda size: b'\x8d' + struct.pack('input_len will be 0; this is intentional so that when - unpickling from a file, the "we've run out of data" code paths will trigger, - causing the Unpickler to go back to the file for more data. Use the returned - size to tell you how much data you can process. */ +/* Don't call it directly: use _Unpickler_ReadInto() */ static Py_ssize_t -_Unpickler_ReadFromFile(UnpicklerObject *self, Py_ssize_t n) -{ - PyObject *data; - Py_ssize_t read_size; - - assert(self->read != NULL); - - if (_Unpickler_SkipConsumed(self) < 0) - return -1; - - if (n == READ_WHOLE_LINE) { - data = PyObject_CallNoArgs(self->readline); - } - else { - PyObject *len; - /* Prefetch some data without advancing the file pointer, if possible */ - if (self->peek && n < PREFETCH) { - len = PyLong_FromSsize_t(PREFETCH); - if (len == NULL) - return -1; - data = _Pickle_FastCall(self->peek, len); - if (data == NULL) { - if (!PyErr_ExceptionMatches(PyExc_NotImplementedError)) - return -1; - /* peek() is probably not supported by the given file object */ - PyErr_Clear(); - Py_CLEAR(self->peek); - } - else { - read_size = _Unpickler_SetStringInput(self, data); - Py_DECREF(data); - if (read_size < 0) { - return -1; - } - - self->prefetched_idx = 0; - if (n <= read_size) - return n; - } - } - len = PyLong_FromSsize_t(n); - if (len == NULL) - return -1; - data = _Pickle_FastCall(self->read, len); - } - if (data == NULL) - return -1; - - read_size = _Unpickler_SetStringInput(self, data); - Py_DECREF(data); - return read_size; -} - -/* Don't call it directly: use _Unpickler_Read() */ -static Py_ssize_t -_Unpickler_ReadImpl(UnpicklerObject *self, PickleState *st, char **s, Py_ssize_t n) -{ - Py_ssize_t num_read; - - *s = NULL; - if (self->next_read_idx > PY_SSIZE_T_MAX - n) { - PyErr_SetString(st->UnpicklingError, - "read would overflow (invalid bytecode)"); - return -1; - } - - /* This case is handled by the _Unpickler_Read() macro for efficiency */ - assert(self->next_read_idx + n > self->input_len); - - if (!self->read) - return bad_readline(st); - - /* Extend the buffer to satisfy desired size */ - num_read = _Unpickler_ReadFromFile(self, n); - if (num_read < 0) - return -1; - if (num_read < n) - return bad_readline(st); - *s = self->input_buffer; - self->next_read_idx = n; - return n; -} - -/* Read `n` bytes from the unpickler's data source, storing the result in `buf`. - * - * This should only be used for non-small data reads where potentially - * avoiding a copy is beneficial. This method does not try to prefetch - * more data into the input buffer. - * - * _Unpickler_Read() is recommended in most cases. - */ -static Py_ssize_t -_Unpickler_ReadInto(PickleState *state, UnpicklerObject *self, char *buf, - Py_ssize_t n) +_Unpickler_ReadIntoFromFile(PickleState *state, UnpicklerObject *self, char *buf, + Py_ssize_t n) { assert(n != READ_WHOLE_LINE); - /* Read from available buffer data, if any */ - Py_ssize_t in_buffer = self->input_len - self->next_read_idx; - if (in_buffer > 0) { - Py_ssize_t to_read = Py_MIN(in_buffer, n); - memcpy(buf, self->input_buffer + self->next_read_idx, to_read); - self->next_read_idx += to_read; - buf += to_read; - n -= to_read; - if (n == 0) { - /* Entire read was satisfied from buffer */ - return n; - } - } - - /* Read from file */ - if (!self->read) { - /* We're unpickling memory, this means the input is truncated */ - return bad_readline(state); - } - if (_Unpickler_SkipConsumed(self) < 0) { - return -1; - } - if (!self->readinto) { /* readinto() not supported on file-like object, fall back to read() * and copy into destination buffer (bpo-39681) */ @@ -1435,6 +1311,163 @@ _Unpickler_ReadInto(PickleState *state, UnpicklerObject *self, char *buf, return n; } +/* If reading from a file, we need to only pull the bytes we need, since there + may be multiple pickle objects arranged contiguously in the same input + buffer. + + If `n` is READ_WHOLE_LINE, read a whole line. Otherwise, read up to `n` + bytes from the input stream/buffer. + + Update the unpickler's input buffer with the newly-read data. Returns -1 on + failure; on success, returns the number of bytes read from the file. + + On success, self->input_len will be 0; this is intentional so that when + unpickling from a file, the "we've run out of data" code paths will trigger, + causing the Unpickler to go back to the file for more data. Use the returned + size to tell you how much data you can process. */ +static Py_ssize_t +_Unpickler_ReadFromFile(PickleState *state, UnpicklerObject *self, Py_ssize_t n) +{ + PyObject *data; + Py_ssize_t read_size; + + assert(self->read != NULL); + + if (_Unpickler_SkipConsumed(self) < 0) + return -1; + + if (n == READ_WHOLE_LINE) { + data = PyObject_CallNoArgs(self->readline); + if (data == NULL) { + return -1; + } + } + else { + PyObject *len; + /* Prefetch some data without advancing the file pointer, if possible */ + if (self->peek && n < PREFETCH) { + len = PyLong_FromSsize_t(PREFETCH); + if (len == NULL) + return -1; + data = _Pickle_FastCall(self->peek, len); + if (data == NULL) { + if (!PyErr_ExceptionMatches(PyExc_NotImplementedError)) + return -1; + /* peek() is probably not supported by the given file object */ + PyErr_Clear(); + Py_CLEAR(self->peek); + } + else { + read_size = _Unpickler_SetStringInput(self, data); + Py_DECREF(data); + if (read_size < 0) { + return -1; + } + + self->prefetched_idx = 0; + if (n <= read_size) + return n; + } + } + Py_ssize_t cursize = Py_MIN(n, MIN_READ_BUF_SIZE); + len = PyLong_FromSsize_t(cursize); + if (len == NULL) + return -1; + data = _Pickle_FastCall(self->read, len); + if (data == NULL) { + return -1; + } + while (cursize < n) { + Py_ssize_t prevsize = cursize; + // geometrically double the chunk size to avoid CPU DoS + cursize += Py_MIN(cursize, n - cursize); + if (_PyBytes_Resize(&data, cursize) < 0) { + return -1; + } + if (_Unpickler_ReadIntoFromFile(state, self, + PyBytes_AS_STRING(data) + prevsize, cursize - prevsize) < 0) + { + Py_DECREF(data); + return -1; + } + } + } + + read_size = _Unpickler_SetStringInput(self, data); + Py_DECREF(data); + return read_size; +} + +/* Don't call it directly: use _Unpickler_Read() */ +static Py_ssize_t +_Unpickler_ReadImpl(UnpicklerObject *self, PickleState *st, char **s, Py_ssize_t n) +{ + Py_ssize_t num_read; + + *s = NULL; + if (self->next_read_idx > PY_SSIZE_T_MAX - n) { + PyErr_SetString(st->UnpicklingError, + "read would overflow (invalid bytecode)"); + return -1; + } + + /* This case is handled by the _Unpickler_Read() macro for efficiency */ + assert(self->next_read_idx + n > self->input_len); + + if (!self->read) + return bad_readline(st); + + /* Extend the buffer to satisfy desired size */ + num_read = _Unpickler_ReadFromFile(st, self, n); + if (num_read < 0) + return -1; + if (num_read < n) + return bad_readline(st); + *s = self->input_buffer; + self->next_read_idx = n; + return n; +} + +/* Read `n` bytes from the unpickler's data source, storing the result in `buf`. + * + * This should only be used for non-small data reads where potentially + * avoiding a copy is beneficial. This method does not try to prefetch + * more data into the input buffer. + * + * _Unpickler_Read() is recommended in most cases. + */ +static Py_ssize_t +_Unpickler_ReadInto(PickleState *state, UnpicklerObject *self, char *buf, + Py_ssize_t n) +{ + assert(n != READ_WHOLE_LINE); + + /* Read from available buffer data, if any */ + Py_ssize_t in_buffer = self->input_len - self->next_read_idx; + if (in_buffer > 0) { + Py_ssize_t to_read = Py_MIN(in_buffer, n); + memcpy(buf, self->input_buffer + self->next_read_idx, to_read); + self->next_read_idx += to_read; + buf += to_read; + n -= to_read; + if (n == 0) { + /* Entire read was satisfied from buffer */ + return n; + } + } + + /* Read from file */ + if (!self->read) { + /* We're unpickling memory, this means the input is truncated */ + return bad_readline(state); + } + if (_Unpickler_SkipConsumed(self) < 0) { + return -1; + } + + return _Unpickler_ReadIntoFromFile(state, self, buf, n); +} + /* Read `n` bytes from the unpickler's data source, storing the result in `*s`. This should be used for all data reads, rather than accessing the unpickler's @@ -1492,7 +1525,7 @@ _Unpickler_Readline(PickleState *state, UnpicklerObject *self, char **result) if (!self->read) return bad_readline(state); - num_read = _Unpickler_ReadFromFile(self, READ_WHOLE_LINE); + num_read = _Unpickler_ReadFromFile(state, self, READ_WHOLE_LINE); if (num_read < 0) return -1; if (num_read == 0 || self->input_buffer[num_read - 1] != '\n') @@ -1525,12 +1558,35 @@ _Unpickler_ResizeMemoList(UnpicklerObject *self, size_t new_size) /* Returns NULL if idx is out of bounds. */ static PyObject * -_Unpickler_MemoGet(UnpicklerObject *self, size_t idx) +_Unpickler_MemoGet(PickleState *st, UnpicklerObject *self, size_t idx) { - if (idx >= self->memo_size) - return NULL; - - return self->memo[idx]; + PyObject *value; + if (idx < self->memo_size) { + value = self->memo[idx]; + if (value != NULL) { + return value; + } + } + if (self->memo_dict != NULL) { + PyObject *key = PyLong_FromSize_t(idx); + if (key == NULL) { + return NULL; + } + if (idx < self->memo_size) { + (void)PyDict_Pop(self->memo_dict, key, &value); + // Migrate dict entry to array for faster future access + self->memo[idx] = value; + } + else { + value = PyDict_GetItemWithError(self->memo_dict, key); + } + Py_DECREF(key); + if (value != NULL || PyErr_Occurred()) { + return value; + } + } + PyErr_Format(st->UnpicklingError, "Memo value not found at index %zd", idx); + return NULL; } /* Returns -1 (with an exception set) on failure, 0 on success. @@ -1541,6 +1597,27 @@ _Unpickler_MemoPut(UnpicklerObject *self, size_t idx, PyObject *value) PyObject *old_item; if (idx >= self->memo_size) { + if (idx > self->memo_len * 2) { + /* The memo keys are too sparse. Use a dict instead of + * a continuous array for the memo. */ + if (self->memo_dict == NULL) { + self->memo_dict = PyDict_New(); + if (self->memo_dict == NULL) { + return -1; + } + } + PyObject *key = PyLong_FromSize_t(idx); + if (key == NULL) { + return -1; + } + + if (PyDict_SetItem(self->memo_dict, key, value) < 0) { + Py_DECREF(key); + return -1; + } + Py_DECREF(key); + return 0; + } if (_Unpickler_ResizeMemoList(self, idx * 2) < 0) return -1; assert(idx < self->memo_size); @@ -1610,6 +1687,7 @@ _Unpickler_New(PyObject *module) self->memo = memo; self->memo_size = MEMO_SIZE; self->memo_len = 0; + self->memo_dict = NULL; self->persistent_load = NULL; self->persistent_load_attr = NULL; memset(&self->buffer, 0, sizeof(Py_buffer)); @@ -5582,13 +5660,28 @@ load_counted_binbytes(PickleState *state, UnpicklerObject *self, int nbytes) return -1; } - bytes = PyBytes_FromStringAndSize(NULL, size); - if (bytes == NULL) - return -1; - if (_Unpickler_ReadInto(state, self, PyBytes_AS_STRING(bytes), size) < 0) { - Py_DECREF(bytes); + Py_ssize_t cursize = Py_MIN(size, MIN_READ_BUF_SIZE); + Py_ssize_t prevsize = 0; + bytes = PyBytes_FromStringAndSize(NULL, cursize); + if (bytes == NULL) { return -1; } + while (1) { + if (_Unpickler_ReadInto(state, self, + PyBytes_AS_STRING(bytes) + prevsize, cursize - prevsize) < 0) + { + Py_DECREF(bytes); + return -1; + } + if (cursize >= size) { + break; + } + prevsize = cursize; + cursize += Py_MIN(cursize, size - cursize); + if (_PyBytes_Resize(&bytes, cursize) < 0) { + return -1; + } + } PDATA_PUSH(self->stack, bytes, -1); return 0; @@ -5613,14 +5706,27 @@ load_counted_bytearray(PickleState *state, UnpicklerObject *self) return -1; } - bytearray = PyByteArray_FromStringAndSize(NULL, size); + Py_ssize_t cursize = Py_MIN(size, MIN_READ_BUF_SIZE); + Py_ssize_t prevsize = 0; + bytearray = PyByteArray_FromStringAndSize(NULL, cursize); if (bytearray == NULL) { return -1; } - char *str = PyByteArray_AS_STRING(bytearray); - if (_Unpickler_ReadInto(state, self, str, size) < 0) { - Py_DECREF(bytearray); - return -1; + while (1) { + if (_Unpickler_ReadInto(state, self, + PyByteArray_AS_STRING(bytearray) + prevsize, + cursize - prevsize) < 0) { + Py_DECREF(bytearray); + return -1; + } + if (cursize >= size) { + break; + } + prevsize = cursize; + cursize += Py_MIN(cursize, size - cursize); + if (PyByteArray_Resize(bytearray, cursize) < 0) { + return -1; + } } PDATA_PUSH(self->stack, bytearray, -1); @@ -6222,20 +6328,15 @@ load_get(PickleState *st, UnpicklerObject *self) if (key == NULL) return -1; idx = PyLong_AsSsize_t(key); + Py_DECREF(key); if (idx == -1 && PyErr_Occurred()) { - Py_DECREF(key); return -1; } - value = _Unpickler_MemoGet(self, idx); + value = _Unpickler_MemoGet(st, self, idx); if (value == NULL) { - if (!PyErr_Occurred()) { - PyErr_Format(st->UnpicklingError, "Memo value not found at index %ld", idx); - } - Py_DECREF(key); return -1; } - Py_DECREF(key); PDATA_APPEND(self->stack, value, -1); return 0; @@ -6253,13 +6354,8 @@ load_binget(PickleState *st, UnpicklerObject *self) idx = Py_CHARMASK(s[0]); - value = _Unpickler_MemoGet(self, idx); + value = _Unpickler_MemoGet(st, self, idx); if (value == NULL) { - PyObject *key = PyLong_FromSsize_t(idx); - if (key != NULL) { - PyErr_Format(st->UnpicklingError, "Memo value not found at index %ld", idx); - Py_DECREF(key); - } return -1; } @@ -6279,13 +6375,8 @@ load_long_binget(PickleState *st, UnpicklerObject *self) idx = calc_binsize(s, 4); - value = _Unpickler_MemoGet(self, idx); + value = _Unpickler_MemoGet(st, self, idx); if (value == NULL) { - PyObject *key = PyLong_FromSsize_t(idx); - if (key != NULL) { - PyErr_Format(st->UnpicklingError, "Memo value not found at index %ld", idx); - Py_DECREF(key); - } return -1; } @@ -7250,6 +7341,7 @@ Unpickler_clear(PyObject *op) self->buffer.buf = NULL; } + Py_CLEAR(self->memo_dict); _Unpickler_MemoCleanup(self); PyMem_Free(self->marks); self->marks = NULL; @@ -7286,6 +7378,7 @@ Unpickler_traverse(PyObject *op, visitproc visit, void *arg) Py_VISIT(self->persistent_load); Py_VISIT(self->persistent_load_attr); Py_VISIT(self->buffers); + Py_VISIT(self->memo_dict); PyObject **memo = self->memo; if (memo) { Py_ssize_t i = self->memo_size; diff --git a/Tools/picklebench/README.md b/Tools/picklebench/README.md new file mode 100644 index 00000000000..7d52485c386 --- /dev/null +++ b/Tools/picklebench/README.md @@ -0,0 +1,232 @@ +# Pickle Chunked Reading Benchmark + +This benchmark measures the performance impact of the chunked reading optimization in GH PR #119204 for the pickle module. + +## What This Tests + +The PR adds chunked reading (1MB chunks) to prevent memory exhaustion when unpickling large objects: +- **BINBYTES8** - Large bytes objects (protocol 4+) +- **BINUNICODE8** - Large strings (protocol 4+) +- **BYTEARRAY8** - Large bytearrays (protocol 5) +- **FRAME** - Large frames +- **LONG4** - Large integers +- An antagonistic mode that tests using memory denial of service inducing malicious pickles. + +## Quick Start + +```bash +# Run full benchmark suite (1MiB → 200MiB, takes several minutes) +build/python Tools/picklebench/memory_dos_impact.py + +# Test just a few sizes (quick test: 1, 10, 50 MiB) +build/python Tools/picklebench/memory_dos_impact.py --sizes 1 10 50 + +# Test smaller range for faster results +build/python Tools/picklebench/memory_dos_impact.py --sizes 1 5 10 + +# Output as markdown for reports +build/python Tools/picklebench/memory_dos_impact.py --format markdown > results.md + +# Test with protocol 4 instead of 5 +build/python Tools/picklebench/memory_dos_impact.py --protocol 4 +``` + +**Note:** Sizes are specified in MiB. Use `--sizes 1 2 5` for 1MiB, 2MiB, 5MiB objects. + +## Antagonistic Mode (DoS Protection Test) + +The `--antagonistic` flag tests **malicious pickles** that demonstrate the memory DoS protection: + +```bash +# Quick DoS protection test (claims 10, 50, 100 MB but provides 1KB) +build/python Tools/picklebench/memory_dos_impact.py --antagonistic --sizes 10 50 100 + +# Full DoS test (default: 10, 50, 100, 500, 1000, 5000 MB claimed) +build/python Tools/picklebench/memory_dos_impact.py --antagonistic +``` + +### What This Tests + +Unlike normal benchmarks that test **legitimate pickles**, antagonistic mode tests: +- **Truncated BINBYTES8**: Claims 100MB but provides only 1KB (will fail to unpickle) +- **Truncated BINUNICODE8**: Same for strings +- **Truncated BYTEARRAY8**: Same for bytearrays +- **Sparse memo attacks**: PUT at index 1 billion (would allocate huge array before PR) + +**Key difference:** +- **Normal mode**: Tests real data, shows ~5% time overhead +- **Antagonistic mode**: Tests malicious data, shows ~99% memory savings + +### Expected Results + +``` +100MB Claimed (actual: 1KB) + binbytes8_100MB_claim + Peak memory: 1.00 MB (claimed: 100 MB, saved: 99.00 MB, 99.0%) + Error: UnpicklingError ← Expected! + +Summary: + Average claimed: 126.2 MB + Average peak: 0.54 MB + Average saved: 125.7 MB (99.6% reduction) +Protection Status: ✓ Memory DoS attacks mitigated by chunked reading +``` + +**Before PR**: Would allocate full claimed size (100MB+), potentially crash +**After PR**: Allocates 1MB chunks, fails fast with minimal memory + +This demonstrates the **security improvement** - protection against memory exhaustion attacks. + +## Before/After Comparison + +The benchmark includes an automatic comparison feature that runs the same tests on both a baseline and current Python build. + +### Option 1: Automatic Comparison (Recommended) + +Build both versions, then use `--baseline` to automatically compare: + +```bash +# Build the baseline (main branch without PR) +git checkout main +mkdir -p build-main +cd build-main && ../configure && make -j $(nproc) && cd .. + +# Build the current version (with PR) +git checkout unpickle-overallocate +mkdir -p build +cd build && ../configure && make -j $(nproc) && cd .. + +# Run automatic comparison (quick test with a few sizes) +build/python Tools/picklebench/memory_dos_impact.py \ + --baseline build-main/python \ + --sizes 1 10 50 + +# Full comparison (all default sizes) +build/python Tools/picklebench/memory_dos_impact.py \ + --baseline build-main/python +``` + +The comparison output shows: +- Side-by-side metrics (Current vs Baseline) +- Percentage change for time and memory +- Overall summary statistics + +### Interpreting Comparison Results + +- **Time change**: Small positive % is expected (chunking adds overhead, typically 5-10%) +- **Memory change**: Negative % is good (chunking saves memory, especially for large objects) +- **Trade-off**: Slightly slower but much safer against memory exhaustion attacks + +### Option 2: Manual Comparison + +Save results separately and compare manually: + +```bash +# Baseline results +build-main/python Tools/picklebench/memory_dos_impact.py --format json > baseline.json + +# Current results +build/python Tools/picklebench/memory_dos_impact.py --format json > current.json + +# Manual comparison +diff -y <(jq '.' baseline.json) <(jq '.' current.json) +``` + +## Understanding the Results + +### Critical Sizes + +The default test suite includes: +- **< 1MiB (999,000 bytes)**: No chunking, allocates full size upfront +- **= 1MiB (1,048,576 bytes)**: Threshold, chunking just starts +- **> 1MiB (1,048,577 bytes)**: Chunked reading engaged +- **1, 2, 5, 10MiB**: Show scaling behavior with chunking +- **20, 50, 100, 200MiB**: Stress test large object handling + +**Note:** The full suite may require more than 16GiB of RAM. + +### Key Metrics + +- **Time (mean)**: Average unpickling time - should be similar before/after +- **Time (stdev)**: Consistency - lower is better +- **Peak Memory**: Maximum memory during unpickling - **expected to be LOWER after PR** +- **Pickle Size**: Size of the serialized data on disk + +### Test Types + +| Test | What It Stresses | +|------|------------------| +| `bytes_*` | BINBYTES8 opcode, raw binary data | +| `string_ascii_*` | BINUNICODE8 with simple ASCII | +| `string_utf8_*` | BINUNICODE8 with multibyte UTF-8 (€ chars) | +| `bytearray_*` | BYTEARRAY8 opcode (protocol 5) | +| `list_large_items_*` | Multiple chunked reads in sequence | +| `dict_large_values_*` | Chunking in dict deserialization | +| `nested_*` | Realistic mixed data structures | +| `tuple_*` | Immutable structures | + +## Expected Results + +### Before PR (main branch) +- Single large allocation per object +- Risk of memory exhaustion with malicious pickles + +### After PR (unpickle-overallocate branch) +- Chunked allocation (1MB at a time) +- **Slightly higher CPU time** (multiple allocations + resizing) +- **Significantly lower peak memory** (no large pre-allocation) +- Protection against DoS via memory exhaustion + +## Advanced Usage + +### Test Specific Sizes + +```bash +# Test only 5MiB and 10MiB objects +build/python Tools/picklebench/memory_dos_impact.py --sizes 5 10 + +# Test large objects: 50, 100, 200 MiB +build/python Tools/picklebench/memory_dos_impact.py --sizes 50 100 200 +``` + +### More Iterations for Stable Timing + +```bash +# Run 10 iterations per test for better statistics +build/python Tools/picklebench/memory_dos_impact.py --iterations 10 --sizes 1 10 +``` + +### JSON Output for Analysis + +```bash +# Generate JSON for programmatic analysis +build/python Tools/picklebench/memory_dos_impact.py --format json | python -m json.tool +``` + +## Interpreting Memory Results + +The **peak memory** metric shows the maximum memory allocated during unpickling: + +- **Without chunking**: Allocates full size immediately + - 10MB object → 10MB allocation upfront + +- **With chunking**: Allocates in 1MB chunks, grows geometrically + - 10MB object → starts with 1MB, grows: 2MB, 4MB, 8MB (final: ~10MB total) + - Peak is lower because allocation is incremental + +## Typical Results + +On a system with the PR applied, you should see: + +``` +1.00MiB Test Results + bytes_1.00MiB: ~0.3ms, 1.00MiB peak (just at threshold) + +2.00MiB Test Results + bytes_2.00MiB: ~0.8ms, 2.00MiB peak (chunked: 1MiB → 2MiB) + +10.00MiB Test Results + bytes_10.00MiB: ~3-5ms, 10.00MiB peak (chunked: 1→2→4→8→10 MiB) +``` + +Time overhead is minimal (~10-20% for very large objects), but memory safety is significantly improved. diff --git a/Tools/picklebench/memory_dos_impact.py b/Tools/picklebench/memory_dos_impact.py new file mode 100755 index 00000000000..3bad6586c46 --- /dev/null +++ b/Tools/picklebench/memory_dos_impact.py @@ -0,0 +1,1069 @@ +#!/usr/bin/env python3 +# +# Author: Claude Sonnet 4.5 as driven by gpshead +# +""" +Microbenchmark for pickle module chunked reading performance (GH PR #119204). + +This script generates Python data structures that act as antagonistic load +tests for the chunked reading code introduced to prevent memory exhaustion when +unpickling large objects. + +The PR adds chunked reading (1MB chunks) for: +- BINBYTES8 (large bytes) +- BINUNICODE8 (large strings) +- BYTEARRAY8 (large bytearrays) +- FRAME (large frames) +- LONG4 (large integers) + +Including an antagonistic mode that exercies memory denial of service pickles. + +Usage: + python memory_dos_impact.py --help +""" + +import argparse +import gc +import io +import json +import os +import pickle +import statistics +import struct +import subprocess +import sys +import tempfile +import tracemalloc +from pathlib import Path +from time import perf_counter +from typing import Any, Dict, List, Tuple, Optional + + +# Configuration +MIN_READ_BUF_SIZE = 1 << 20 # 1MB - matches pickle.py _MIN_READ_BUF_SIZE + +# Test sizes in MiB +DEFAULT_SIZES_MIB = [1, 2, 5, 10, 20, 50, 100, 200] + +# Convert to bytes, plus threshold boundary tests +DEFAULT_SIZES = ( + [999_000] # Below 1MiB (no chunking) + + [size * (1 << 20) for size in DEFAULT_SIZES_MIB] # MiB to bytes + + [1_048_577] # Just above 1MiB (minimal chunking overhead) +) +DEFAULT_SIZES.sort() + +# Baseline benchmark configuration +BASELINE_BENCHMARK_TIMEOUT_SECONDS = 600 # 10 minutes + +# Sparse memo attack test configuration +# Format: test_name -> (memo_index, baseline_memory_note) +SPARSE_MEMO_TESTS = { + "sparse_memo_1M": (1_000_000, "~8 MB array"), + "sparse_memo_100M": (100_000_000, "~800 MB array"), + "sparse_memo_1B": (1_000_000_000, "~8 GB array"), +} + + +# Utility functions + +def _extract_size_mb(size_key: str) -> float: + """Extract numeric MiB value from size_key like '10.00MB' or '1.00MiB'. + + Returns 0.0 for non-numeric keys (they'll be sorted last). + """ + try: + return float(size_key.replace('MB', '').replace('MiB', '')) + except ValueError: + return 999999.0 # Put non-numeric keys last + + +def _format_output(results: Dict[str, Dict[str, Any]], format_type: str, is_antagonistic: bool) -> str: + """Format benchmark results according to requested format. + + Args: + results: Benchmark results dictionary + format_type: Output format ('text', 'markdown', or 'json') + is_antagonistic: Whether these are antagonistic (DoS) test results + + Returns: + Formatted output string + """ + if format_type == 'json': + return Reporter.format_json(results) + elif is_antagonistic: + # Antagonistic mode uses specialized formatter for text/markdown + return Reporter.format_antagonistic(results) + elif format_type == 'text': + return Reporter.format_text(results) + elif format_type == 'markdown': + return Reporter.format_markdown(results) + else: + # Default to text format + return Reporter.format_text(results) + + +class AntagonisticGenerator: + """Generate malicious/truncated pickles for DoS protection testing. + + These pickles claim large sizes but provide minimal data, causing them to fail + during unpickling. They demonstrate the memory protection of chunked reading. + """ + + @staticmethod + def truncated_binbytes8(claimed_size: int, actual_size: int = 1024) -> bytes: + """BINBYTES8 claiming `claimed_size` but providing only `actual_size` bytes. + + This will fail with UnpicklingError but demonstrates peak memory usage. + Before PR: Allocates full claimed_size + After PR: Allocates in 1MB chunks, fails fast + """ + return b'\x8e' + struct.pack(' bytes: + """BINUNICODE8 claiming `claimed_size` but providing only `actual_size` bytes.""" + return b'\x8d' + struct.pack(' bytes: + """BYTEARRAY8 claiming `claimed_size` but providing only `actual_size` bytes.""" + return b'\x96' + struct.pack(' bytes: + """FRAME claiming `claimed_size` but providing minimal data.""" + return b'\x95' + struct.pack(' bytes: + """LONG_BINPUT with huge sparse index. + + Before PR: Tries to allocate array with `index` slots (OOM) + After PR: Uses dict-based memo for sparse indices + """ + return (b'(]r' + struct.pack(' bytes: + """Multiple BINBYTES8 claims in sequence. + + Tests that multiple large claims don't accumulate memory. + """ + data = b'(' # MARK + for _ in range(count): + data += b'\x8e' + struct.pack(' bytes: + """Generate random bytes of specified size.""" + return os.urandom(size) + + @staticmethod + def large_string_ascii(size: int) -> str: + """Generate ASCII string of specified size.""" + return 'x' * size + + @staticmethod + def large_string_multibyte(size: int) -> str: + """Generate multibyte UTF-8 string (3 bytes per char for €).""" + # Each € is 3 bytes in UTF-8 + return '€' * (size // 3) + + @staticmethod + def large_bytearray(size: int) -> bytearray: + """Generate bytearray of specified size.""" + return bytearray(os.urandom(size)) + + @staticmethod + def list_of_large_bytes(item_size: int, count: int) -> List[bytes]: + """Generate list containing multiple large bytes objects.""" + return [os.urandom(item_size) for _ in range(count)] + + @staticmethod + def dict_with_large_values(value_size: int, count: int) -> Dict[str, bytes]: + """Generate dict with large bytes values.""" + return { + f'key_{i}': os.urandom(value_size) + for i in range(count) + } + + @staticmethod + def nested_structure(size: int) -> Dict[str, Any]: + """Generate nested structure with various large objects.""" + chunk_size = size // 4 + return { + 'name': 'test_object', + 'data': { + 'bytes': os.urandom(chunk_size), + 'string': 's' * chunk_size, + 'bytearray': bytearray(b'b' * chunk_size), + }, + 'items': [os.urandom(chunk_size // 4) for _ in range(4)], + 'metadata': { + 'size': size, + 'type': 'nested', + }, + } + + @staticmethod + def tuple_of_large_objects(size: int) -> Tuple[bytes, str, bytearray]: + """Generate tuple with large objects (immutable, different pickle path).""" + chunk_size = size // 3 + return ( + os.urandom(chunk_size), + 'x' * chunk_size, + bytearray(b'y' * chunk_size), + ) + + +class PickleBenchmark: + """Benchmark pickle unpickling performance and memory usage.""" + + def __init__(self, obj: Any, protocol: int = 5, iterations: int = 3): + self.obj = obj + self.protocol = protocol + self.iterations = iterations + self.pickle_data = pickle.dumps(obj, protocol=protocol) + self.pickle_size = len(self.pickle_data) + + def benchmark_time(self) -> Dict[str, float]: + """Measure unpickling time over multiple iterations.""" + times = [] + + for _ in range(self.iterations): + start = perf_counter() + result = pickle.loads(self.pickle_data) + elapsed = perf_counter() - start + times.append(elapsed) + + # Verify correctness (first iteration only) + if len(times) == 1: + if result != self.obj: + raise ValueError("Unpickled object doesn't match original!") + + return { + 'mean': statistics.mean(times), + 'median': statistics.median(times), + 'stdev': statistics.stdev(times) if len(times) > 1 else 0.0, + 'min': min(times), + 'max': max(times), + } + + def benchmark_memory(self) -> int: + """Measure peak memory usage during unpickling.""" + tracemalloc.start() + + # Warmup + pickle.loads(self.pickle_data) + + # Actual measurement + gc.collect() + tracemalloc.reset_peak() + result = pickle.loads(self.pickle_data) + current, peak = tracemalloc.get_traced_memory() + + tracemalloc.stop() + + # Verify correctness + if result != self.obj: + raise ValueError("Unpickled object doesn't match original!") + + return peak + + def run_all(self) -> Dict[str, Any]: + """Run all benchmarks and return comprehensive results.""" + time_stats = self.benchmark_time() + peak_memory = self.benchmark_memory() + + return { + 'pickle_size_bytes': self.pickle_size, + 'pickle_size_mb': self.pickle_size / (1 << 20), + 'protocol': self.protocol, + 'time': time_stats, + 'memory_peak_bytes': peak_memory, + 'memory_peak_mb': peak_memory / (1 << 20), + 'iterations': self.iterations, + } + + +class AntagonisticBenchmark: + """Benchmark antagonistic/malicious pickles that demonstrate DoS protection. + + These pickles are designed to FAIL unpickling, but we measure peak memory + usage before the failure to demonstrate the memory protection. + """ + + def __init__(self, pickle_data: bytes, name: str): + self.pickle_data = pickle_data + self.name = name + + def measure_peak_memory(self, expect_success: bool = False) -> Dict[str, Any]: + """Measure peak memory when attempting to unpickle antagonistic data. + + Args: + expect_success: If True, test expects successful unpickling (e.g., sparse memo). + If False, test expects failure (e.g., truncated data). + """ + tracemalloc.start() + gc.collect() + tracemalloc.reset_peak() + + error_type = None + error_msg = None + succeeded = False + + try: + result = pickle.loads(self.pickle_data) + succeeded = True + if expect_success: + error_type = "Success (expected)" + else: + error_type = "WARNING: Expected failure but succeeded" + except (pickle.UnpicklingError, EOFError, ValueError, OverflowError) as e: + if expect_success: + error_type = f"UNEXPECTED FAILURE: {type(e).__name__}" + error_msg = str(e)[:100] + else: + # Expected failure for truncated data tests + error_type = type(e).__name__ + error_msg = str(e)[:100] + + current, peak = tracemalloc.get_traced_memory() + tracemalloc.stop() + + return { + 'test_name': self.name, + 'peak_memory_bytes': peak, + 'peak_memory_mb': peak / (1 << 20), + 'error_type': error_type, + 'error_msg': error_msg, + 'pickle_size_bytes': len(self.pickle_data), + 'expected_outcome': 'success' if expect_success else 'failure', + 'succeeded': succeeded, + } + + +class AntagonisticTestSuite: + """Manage a suite of antagonistic (DoS protection) tests.""" + + # Default sizes in MB to claim (will provide only 1KB actual data) + DEFAULT_ANTAGONISTIC_SIZES_MB = [10, 50, 100, 500, 1000, 5000] + + def __init__(self, claimed_sizes_mb: List[int]): + self.claimed_sizes_mb = claimed_sizes_mb + + def _run_truncated_test( + self, + test_type: str, + generator_func, + claimed_bytes: int, + claimed_mb: int, + size_key: str, + all_results: Dict[str, Dict[str, Any]] + ) -> None: + """Run a single truncated data test and store results. + + Args: + test_type: Type identifier (e.g., 'binbytes8', 'binunicode8') + generator_func: Function to generate malicious pickle data + claimed_bytes: Size claimed in the pickle (bytes) + claimed_mb: Size claimed in the pickle (MB) + size_key: Result key for this size (e.g., '10MB') + all_results: Dictionary to store results in + """ + test_name = f"{test_type}_{size_key}_claim" + data = generator_func(claimed_bytes) + bench = AntagonisticBenchmark(data, test_name) + result = bench.measure_peak_memory(expect_success=False) + result['claimed_mb'] = claimed_mb + all_results[size_key][test_name] = result + + def run_all_tests(self) -> Dict[str, Dict[str, Any]]: + """Run comprehensive antagonistic test suite.""" + all_results = {} + + for claimed_mb in self.claimed_sizes_mb: + claimed_bytes = claimed_mb << 20 + size_key = f"{claimed_mb}MB" + all_results[size_key] = {} + + # Run truncated data tests (expect failure) + self._run_truncated_test('binbytes8', AntagonisticGenerator.truncated_binbytes8, + claimed_bytes, claimed_mb, size_key, all_results) + self._run_truncated_test('binunicode8', AntagonisticGenerator.truncated_binunicode8, + claimed_bytes, claimed_mb, size_key, all_results) + self._run_truncated_test('bytearray8', AntagonisticGenerator.truncated_bytearray8, + claimed_bytes, claimed_mb, size_key, all_results) + self._run_truncated_test('frame', AntagonisticGenerator.truncated_frame, + claimed_bytes, claimed_mb, size_key, all_results) + + # Test 5: Sparse memo (expect success - dict-based memo works!) + all_results["Sparse Memo (Success Expected)"] = {} + for test_name, (index, baseline_note) in SPARSE_MEMO_TESTS.items(): + data = AntagonisticGenerator.sparse_memo_attack(index) + bench = AntagonisticBenchmark(data, test_name) + result = bench.measure_peak_memory(expect_success=True) + result['claimed_mb'] = "N/A" + result['baseline_note'] = f"Without PR: {baseline_note}" + all_results["Sparse Memo (Success Expected)"][test_name] = result + + # Test 6: Multi-claim attack (expect failure) + test_name = "multi_claim_10x100MB" + data = AntagonisticGenerator.multi_claim_attack(10, 100 << 20) + bench = AntagonisticBenchmark(data, test_name) + result = bench.measure_peak_memory(expect_success=False) + result['claimed_mb'] = 1000 # 10 * 100MB + all_results["Multi-Claim (Failure Expected)"] = {test_name: result} + + return all_results + + +class TestSuite: + """Manage a suite of benchmark tests.""" + + def __init__(self, sizes: List[int], protocol: int = 5, iterations: int = 3): + self.sizes = sizes + self.protocol = protocol + self.iterations = iterations + self.results = {} + + def run_test(self, name: str, obj: Any) -> Dict[str, Any]: + """Run benchmark for a single test object.""" + bench = PickleBenchmark(obj, self.protocol, self.iterations) + results = bench.run_all() + results['test_name'] = name + results['object_type'] = type(obj).__name__ + return results + + def run_all_tests(self) -> Dict[str, Dict[str, Any]]: + """Run comprehensive test suite across all sizes and types.""" + all_results = {} + + for size in self.sizes: + size_key = f"{size / (1 << 20):.2f}MB" + all_results[size_key] = {} + + # Test 1: Large bytes object (BINBYTES8) + test_name = f"bytes_{size_key}" + obj = DataGenerator.large_bytes(size) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + # Test 2: Large ASCII string (BINUNICODE8) + test_name = f"string_ascii_{size_key}" + obj = DataGenerator.large_string_ascii(size) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + # Test 3: Large multibyte UTF-8 string + if size >= 3: + test_name = f"string_utf8_{size_key}" + obj = DataGenerator.large_string_multibyte(size) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + # Test 4: Large bytearray (BYTEARRAY8, protocol 5) + if self.protocol >= 5: + test_name = f"bytearray_{size_key}" + obj = DataGenerator.large_bytearray(size) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + # Test 5: List of large objects (repeated chunking) + if size >= MIN_READ_BUF_SIZE * 2: + test_name = f"list_large_items_{size_key}" + item_size = size // 5 + obj = DataGenerator.list_of_large_bytes(item_size, 5) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + # Test 6: Dict with large values + if size >= MIN_READ_BUF_SIZE * 2: + test_name = f"dict_large_values_{size_key}" + value_size = size // 3 + obj = DataGenerator.dict_with_large_values(value_size, 3) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + # Test 7: Nested structure + if size >= MIN_READ_BUF_SIZE: + test_name = f"nested_{size_key}" + obj = DataGenerator.nested_structure(size) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + # Test 8: Tuple (immutable) + if size >= 3: + test_name = f"tuple_{size_key}" + obj = DataGenerator.tuple_of_large_objects(size) + all_results[size_key][test_name] = self.run_test(test_name, obj) + + return all_results + + +class Comparator: + """Compare benchmark results between current and baseline interpreters.""" + + @staticmethod + def _extract_json_from_output(output: str) -> Dict[str, Dict[str, Any]]: + """Extract JSON data from subprocess output. + + Skips any print statements before the JSON output and parses the JSON. + + Args: + output: Raw stdout from subprocess + + Returns: + Parsed JSON as dictionary + + Raises: + SystemExit: If JSON cannot be found or parsed + """ + output_lines = output.strip().split('\n') + json_start = -1 + for i, line in enumerate(output_lines): + if line.strip().startswith('{'): + json_start = i + break + + if json_start == -1: + print("Error: Could not find JSON output from baseline", file=sys.stderr) + sys.exit(1) + + json_output = '\n'.join(output_lines[json_start:]) + try: + return json.loads(json_output) + except json.JSONDecodeError as e: + print(f"Error: Could not parse baseline JSON output: {e}", file=sys.stderr) + sys.exit(1) + + @staticmethod + def run_baseline_benchmark(baseline_python: str, args: argparse.Namespace) -> Dict[str, Dict[str, Any]]: + """Run the benchmark using the baseline Python interpreter.""" + # Build command to run this script with baseline Python + cmd = [ + baseline_python, + __file__, + '--format', 'json', + '--protocol', str(args.protocol), + '--iterations', str(args.iterations), + ] + + if args.sizes is not None: + cmd.extend(['--sizes'] + [str(s) for s in args.sizes]) + + if args.antagonistic: + cmd.append('--antagonistic') + + print(f"\nRunning baseline benchmark with: {baseline_python}") + print(f"Command: {' '.join(cmd)}\n") + + try: + result = subprocess.run( + cmd, + capture_output=True, + text=True, + timeout=BASELINE_BENCHMARK_TIMEOUT_SECONDS, + ) + + if result.returncode != 0: + print(f"Error running baseline benchmark:", file=sys.stderr) + print(result.stderr, file=sys.stderr) + sys.exit(1) + + # Extract and parse JSON from output + return Comparator._extract_json_from_output(result.stdout) + + except subprocess.TimeoutExpired: + print("Error: Baseline benchmark timed out", file=sys.stderr) + sys.exit(1) + + @staticmethod + def calculate_change(baseline_value: float, current_value: float) -> float: + """Calculate percentage change from baseline to current.""" + if baseline_value == 0: + return 0.0 + return ((current_value - baseline_value) / baseline_value) * 100 + + @staticmethod + def format_comparison( + current_results: Dict[str, Dict[str, Any]], + baseline_results: Dict[str, Dict[str, Any]] + ) -> str: + """Format comparison results as readable text.""" + lines = [] + lines.append("=" * 100) + lines.append("Pickle Unpickling Benchmark Comparison") + lines.append("=" * 100) + lines.append("") + lines.append("Legend: Current vs Baseline | % Change (+ is slower/more memory, - is faster/less memory)") + lines.append("") + + # Sort size keys numerically + for size_key in sorted(current_results.keys(), key=_extract_size_mb): + if size_key not in baseline_results: + continue + + lines.append(f"\n{size_key} Comparison") + lines.append("-" * 100) + + current_tests = current_results[size_key] + baseline_tests = baseline_results[size_key] + + for test_name in sorted(current_tests.keys()): + if test_name not in baseline_tests: + continue + + curr = current_tests[test_name] + base = baseline_tests[test_name] + + time_change = Comparator.calculate_change( + base['time']['mean'], curr['time']['mean'] + ) + mem_change = Comparator.calculate_change( + base['memory_peak_mb'], curr['memory_peak_mb'] + ) + + lines.append(f"\n {curr['test_name']}") + lines.append(f" Time: {curr['time']['mean']*1000:6.2f}ms vs {base['time']['mean']*1000:6.2f}ms | " + f"{time_change:+6.1f}%") + lines.append(f" Memory: {curr['memory_peak_mb']:6.2f}MB vs {base['memory_peak_mb']:6.2f}MB | " + f"{mem_change:+6.1f}%") + + lines.append("\n" + "=" * 100) + lines.append("\nSummary:") + + # Calculate overall statistics + time_changes = [] + mem_changes = [] + + for size_key in current_results.keys(): + if size_key not in baseline_results: + continue + for test_name in current_results[size_key].keys(): + if test_name not in baseline_results[size_key]: + continue + curr = current_results[size_key][test_name] + base = baseline_results[size_key][test_name] + + time_changes.append(Comparator.calculate_change( + base['time']['mean'], curr['time']['mean'] + )) + mem_changes.append(Comparator.calculate_change( + base['memory_peak_mb'], curr['memory_peak_mb'] + )) + + if time_changes: + lines.append(f" Time change: mean={statistics.mean(time_changes):+.1f}%, " + f"median={statistics.median(time_changes):+.1f}%") + if mem_changes: + lines.append(f" Memory change: mean={statistics.mean(mem_changes):+.1f}%, " + f"median={statistics.median(mem_changes):+.1f}%") + + lines.append("=" * 100) + return "\n".join(lines) + + @staticmethod + def format_antagonistic_comparison( + current_results: Dict[str, Dict[str, Any]], + baseline_results: Dict[str, Dict[str, Any]] + ) -> str: + """Format antagonistic benchmark comparison results.""" + lines = [] + lines.append("=" * 100) + lines.append("Antagonistic Pickle Benchmark Comparison (Memory DoS Protection)") + lines.append("=" * 100) + lines.append("") + lines.append("Legend: Current vs Baseline | Memory Change (- is better, shows memory saved)") + lines.append("") + lines.append("This compares TWO types of DoS protection:") + lines.append(" 1. Truncated data → Baseline allocates full claimed size, Current uses chunked reading") + lines.append(" 2. Sparse memo → Baseline uses huge arrays, Current uses dict-based memo") + lines.append("") + + # Track statistics + truncated_memory_changes = [] + sparse_memory_changes = [] + + # Sort size keys numerically + for size_key in sorted(current_results.keys(), key=_extract_size_mb): + if size_key not in baseline_results: + continue + + lines.append(f"\n{size_key} Comparison") + lines.append("-" * 100) + + current_tests = current_results[size_key] + baseline_tests = baseline_results[size_key] + + for test_name in sorted(current_tests.keys()): + if test_name not in baseline_tests: + continue + + curr = current_tests[test_name] + base = baseline_tests[test_name] + + curr_peak_mb = curr['peak_memory_mb'] + base_peak_mb = base['peak_memory_mb'] + expected_outcome = curr.get('expected_outcome', 'failure') + + mem_change = Comparator.calculate_change(base_peak_mb, curr_peak_mb) + mem_saved_mb = base_peak_mb - curr_peak_mb + + lines.append(f"\n {curr['test_name']}") + lines.append(f" Memory: {curr_peak_mb:6.2f}MB vs {base_peak_mb:6.2f}MB | " + f"{mem_change:+6.1f}% ({mem_saved_mb:+.2f}MB saved)") + + # Track based on test type + if expected_outcome == 'success': + sparse_memory_changes.append(mem_change) + if curr.get('baseline_note'): + lines.append(f" Note: {curr['baseline_note']}") + else: + truncated_memory_changes.append(mem_change) + claimed_mb = curr.get('claimed_mb', 'N/A') + if claimed_mb != 'N/A': + lines.append(f" Claimed: {claimed_mb:,}MB") + + # Show status + curr_status = curr.get('error_type', 'Unknown') + base_status = base.get('error_type', 'Unknown') + if curr_status != base_status: + lines.append(f" Status: {curr_status} (baseline: {base_status})") + else: + lines.append(f" Status: {curr_status}") + + lines.append("\n" + "=" * 100) + lines.append("\nSummary:") + lines.append("") + + if truncated_memory_changes: + lines.append(" Truncated Data Protection (chunked reading):") + lines.append(f" Mean memory change: {statistics.mean(truncated_memory_changes):+.1f}%") + lines.append(f" Median memory change: {statistics.median(truncated_memory_changes):+.1f}%") + avg_change = statistics.mean(truncated_memory_changes) + if avg_change < -50: + lines.append(f" Result: ✓ Dramatic memory reduction ({avg_change:.1f}%) - DoS protection working!") + elif avg_change < 0: + lines.append(f" Result: ✓ Memory reduced ({avg_change:.1f}%)") + else: + lines.append(f" Result: ⚠ Memory increased ({avg_change:.1f}%) - unexpected!") + lines.append("") + + if sparse_memory_changes: + lines.append(" Sparse Memo Protection (dict-based memo):") + lines.append(f" Mean memory change: {statistics.mean(sparse_memory_changes):+.1f}%") + lines.append(f" Median memory change: {statistics.median(sparse_memory_changes):+.1f}%") + avg_change = statistics.mean(sparse_memory_changes) + if avg_change < -50: + lines.append(f" Result: ✓ Dramatic memory reduction ({avg_change:.1f}%) - Dict optimization working!") + elif avg_change < 0: + lines.append(f" Result: ✓ Memory reduced ({avg_change:.1f}%)") + else: + lines.append(f" Result: ⚠ Memory increased ({avg_change:.1f}%) - unexpected!") + + lines.append("") + lines.append("=" * 100) + return "\n".join(lines) + + +class Reporter: + """Format and display benchmark results.""" + + @staticmethod + def format_text(results: Dict[str, Dict[str, Any]]) -> str: + """Format results as readable text.""" + lines = [] + lines.append("=" * 80) + lines.append("Pickle Unpickling Benchmark Results") + lines.append("=" * 80) + lines.append("") + + for size_key, tests in results.items(): + lines.append(f"\n{size_key} Test Results") + lines.append("-" * 80) + + for test_name, data in tests.items(): + lines.append(f"\n Test: {data['test_name']}") + lines.append(f" Type: {data['object_type']}") + lines.append(f" Pickle size: {data['pickle_size_mb']:.2f} MB") + lines.append(f" Time (mean): {data['time']['mean']*1000:.2f} ms") + lines.append(f" Time (stdev): {data['time']['stdev']*1000:.2f} ms") + lines.append(f" Peak memory: {data['memory_peak_mb']:.2f} MB") + lines.append(f" Protocol: {data['protocol']}") + + lines.append("\n" + "=" * 80) + return "\n".join(lines) + + @staticmethod + def format_markdown(results: Dict[str, Dict[str, Any]]) -> str: + """Format results as markdown table.""" + lines = [] + lines.append("# Pickle Unpickling Benchmark Results\n") + + for size_key, tests in results.items(): + lines.append(f"## {size_key}\n") + lines.append("| Test | Type | Pickle Size (MB) | Time (ms) | Stdev (ms) | Peak Memory (MB) |") + lines.append("|------|------|------------------|-----------|------------|------------------|") + + for test_name, data in tests.items(): + lines.append( + f"| {data['test_name']} | " + f"{data['object_type']} | " + f"{data['pickle_size_mb']:.2f} | " + f"{data['time']['mean']*1000:.2f} | " + f"{data['time']['stdev']*1000:.2f} | " + f"{data['memory_peak_mb']:.2f} |" + ) + lines.append("") + + return "\n".join(lines) + + @staticmethod + def format_json(results: Dict[str, Dict[str, Any]]) -> str: + """Format results as JSON.""" + import json + return json.dumps(results, indent=2) + + @staticmethod + def format_antagonistic(results: Dict[str, Dict[str, Any]]) -> str: + """Format antagonistic benchmark results.""" + lines = [] + lines.append("=" * 100) + lines.append("Antagonistic Pickle Benchmark (Memory DoS Protection Test)") + lines.append("=" * 100) + lines.append("") + lines.append("This benchmark tests TWO types of DoS protection:") + lines.append(" 1. Truncated data attacks → Expect FAILURE with minimal memory before failure") + lines.append(" 2. Sparse memo attacks → Expect SUCCESS with dict-based memo (vs huge array)") + lines.append("") + + # Sort size keys numerically + for size_key in sorted(results.keys(), key=_extract_size_mb): + tests = results[size_key] + + # Determine test type from first test + if tests: + first_test = next(iter(tests.values())) + expected_outcome = first_test.get('expected_outcome', 'failure') + claimed_mb = first_test.get('claimed_mb', 'N/A') + + # Header varies by test type + if "Sparse Memo" in size_key: + lines.append(f"\n{size_key}") + lines.append("-" * 100) + elif "Multi-Claim" in size_key: + lines.append(f"\n{size_key}") + lines.append("-" * 100) + elif claimed_mb != 'N/A': + lines.append(f"\n{size_key} Claimed (actual: 1KB) - Expect Failure") + lines.append("-" * 100) + else: + lines.append(f"\n{size_key}") + lines.append("-" * 100) + + for test_name, data in tests.items(): + peak_mb = data['peak_memory_mb'] + claimed = data.get('claimed_mb', 'N/A') + expected_outcome = data.get('expected_outcome', 'failure') + succeeded = data.get('succeeded', False) + baseline_note = data.get('baseline_note', '') + + lines.append(f" {data['test_name']}") + + # Format output based on test type + if expected_outcome == 'success': + # Sparse memo test - show success with dict + status_icon = "✓" if succeeded else "✗" + lines.append(f" Peak memory: {peak_mb:8.2f} MB {status_icon}") + lines.append(f" Status: {data['error_type']}") + if baseline_note: + lines.append(f" {baseline_note}") + else: + # Truncated data test - show savings before failure + if claimed != 'N/A': + saved_mb = claimed - peak_mb + savings_pct = (saved_mb / claimed * 100) if claimed > 0 else 0 + lines.append(f" Peak memory: {peak_mb:8.2f} MB (claimed: {claimed:,} MB, saved: {saved_mb:.2f} MB, {savings_pct:.1f}%)") + else: + lines.append(f" Peak memory: {peak_mb:8.2f} MB") + lines.append(f" Status: {data['error_type']}") + + lines.append("\n" + "=" * 100) + + # Calculate statistics by test type + truncated_claimed = 0 + truncated_peak = 0 + truncated_count = 0 + + sparse_peak_total = 0 + sparse_count = 0 + + for size_key, tests in results.items(): + for test_name, data in tests.items(): + expected_outcome = data.get('expected_outcome', 'failure') + + if expected_outcome == 'failure': + # Truncated data test + claimed = data.get('claimed_mb', 0) + if claimed != 'N/A' and claimed > 0: + truncated_claimed += claimed + truncated_peak += data['peak_memory_mb'] + truncated_count += 1 + else: + # Sparse memo test + sparse_peak_total += data['peak_memory_mb'] + sparse_count += 1 + + lines.append("\nSummary:") + lines.append("") + + if truncated_count > 0: + avg_claimed = truncated_claimed / truncated_count + avg_peak = truncated_peak / truncated_count + avg_saved = avg_claimed - avg_peak + avg_savings_pct = (avg_saved / avg_claimed * 100) if avg_claimed > 0 else 0 + + lines.append(" Truncated Data Protection (chunked reading):") + lines.append(f" Average claimed: {avg_claimed:,.1f} MB") + lines.append(f" Average peak: {avg_peak:,.2f} MB") + lines.append(f" Average saved: {avg_saved:,.2f} MB ({avg_savings_pct:.1f}% reduction)") + lines.append(f" Status: ✓ Fails fast with minimal memory") + lines.append("") + + if sparse_count > 0: + avg_sparse_peak = sparse_peak_total / sparse_count + lines.append(" Sparse Memo Protection (dict-based memo):") + lines.append(f" Average peak: {avg_sparse_peak:,.2f} MB") + lines.append(f" Status: ✓ Succeeds with dict (vs GB-sized arrays without PR)") + lines.append(f" Note: Compare with --baseline to see actual memory savings") + + lines.append("") + lines.append("=" * 100) + return "\n".join(lines) + + +def main(): + parser = argparse.ArgumentParser( + description="Benchmark pickle unpickling performance for large objects" + ) + parser.add_argument( + '--sizes', + type=int, + nargs='+', + default=None, + metavar='MiB', + help=f'Object sizes to test in MiB (default: {DEFAULT_SIZES_MIB})' + ) + parser.add_argument( + '--protocol', + type=int, + default=5, + choices=[0, 1, 2, 3, 4, 5], + help='Pickle protocol version (default: 5)' + ) + parser.add_argument( + '--iterations', + type=int, + default=3, + help='Number of benchmark iterations (default: 3)' + ) + parser.add_argument( + '--format', + choices=['text', 'markdown', 'json'], + default='text', + help='Output format (default: text)' + ) + parser.add_argument( + '--baseline', + type=str, + metavar='PYTHON', + help='Path to baseline Python interpreter for comparison (e.g., ../main-build/python)' + ) + parser.add_argument( + '--antagonistic', + action='store_true', + help='Run antagonistic/malicious pickle tests (DoS protection benchmark)' + ) + + args = parser.parse_args() + + # Handle antagonistic mode + if args.antagonistic: + # Antagonistic mode uses claimed sizes in MB, not actual data sizes + if args.sizes is None: + claimed_sizes_mb = AntagonisticTestSuite.DEFAULT_ANTAGONISTIC_SIZES_MB + else: + claimed_sizes_mb = args.sizes + + print(f"Running ANTAGONISTIC pickle benchmark (DoS protection test)...") + print(f"Claimed sizes: {claimed_sizes_mb} MiB (actual data: 1KB each)") + print(f"NOTE: These pickles will FAIL to unpickle (expected)") + print() + + # Run antagonistic benchmark suite + suite = AntagonisticTestSuite(claimed_sizes_mb) + results = suite.run_all_tests() + + # Format and display results + if args.baseline: + # Verify baseline Python exists + baseline_path = Path(args.baseline) + if not baseline_path.exists(): + print(f"Error: Baseline Python not found: {args.baseline}", file=sys.stderr) + return 1 + + # Run baseline benchmark + baseline_results = Comparator.run_baseline_benchmark(args.baseline, args) + + # Show comparison + comparison_output = Comparator.format_antagonistic_comparison(results, baseline_results) + print(comparison_output) + else: + # Format and display results + output = _format_output(results, args.format, is_antagonistic=True) + print(output) + + else: + # Normal mode: legitimate pickle benchmarks + # Convert sizes from MiB to bytes + if args.sizes is None: + sizes_bytes = DEFAULT_SIZES + else: + sizes_bytes = [size * (1 << 20) for size in args.sizes] + + print(f"Running pickle benchmark with protocol {args.protocol}...") + print(f"Test sizes: {[f'{s/(1<<20):.2f}MiB' for s in sizes_bytes]}") + print(f"Iterations per test: {args.iterations}") + print() + + # Run benchmark suite + suite = TestSuite(sizes_bytes, args.protocol, args.iterations) + results = suite.run_all_tests() + + # If baseline comparison requested, run baseline and compare + if args.baseline: + # Verify baseline Python exists + baseline_path = Path(args.baseline) + if not baseline_path.exists(): + print(f"Error: Baseline Python not found: {args.baseline}", file=sys.stderr) + return 1 + + # Run baseline benchmark + baseline_results = Comparator.run_baseline_benchmark(args.baseline, args) + + # Show comparison + comparison_output = Comparator.format_comparison(results, baseline_results) + print(comparison_output) + + else: + # Format and display results + output = _format_output(results, args.format, is_antagonistic=False) + print(output) + + return 0 + + +if __name__ == '__main__': + sys.exit(main())