Commit 0507bf56 authored by Victor Stinner's avatar Victor Stinner

Issue #3329: Implement the PEP 445

Add new enum:

* PyMemAllocatorDomain

Add new structures:

* PyMemAllocator
* PyObjectArenaAllocator

Add new functions:

* PyMem_RawMalloc(), PyMem_RawRealloc(), PyMem_RawFree()
* PyMem_GetAllocator(), PyMem_SetAllocator()
* PyObject_GetArenaAllocator(), PyObject_SetArenaAllocator()
* PyMem_SetupDebugHooks()

Changes:

* PyMem_Malloc()/PyObject_Realloc() now always call malloc()/realloc(), instead
  of calling PyObject_Malloc()/PyObject_Realloc() in debug mode.
* PyObject_Malloc()/PyObject_Realloc() now falls back to
  PyMem_Malloc()/PyMem_Realloc() for allocations larger than 512 bytes.
* Redesign debug checks on memory block allocators as hooks, instead of using C
  macros
parent df715ba5
......@@ -84,6 +84,48 @@ the C library allocator as shown in the previous example, the allocated memory
for the I/O buffer escapes completely the Python memory manager.
Raw Memory Interface
====================
The following function sets are wrappers to the system allocator. These
functions are thread-safe, the :term:`GIL <global interpreter lock>` does not
need to be held.
The default raw memory block allocator uses the following functions:
:c:func:`malloc`, :c:func:`realloc` and :c:func:`free`; call ``malloc(1)`` when
requesting zero bytes.
.. versionadded:: 3.4
.. c:function:: void* PyMem_RawMalloc(size_t n)
Allocates *n* bytes and returns a pointer of type :c:type:`void\*` to the
allocated memory, or *NULL* if the request fails. Requesting zero bytes
returns a distinct non-*NULL* pointer if possible, as if
``PyMem_RawMalloc(1)`` had been called instead. The memory will not have
been initialized in any way.
.. c:function:: void* PyMem_RawRealloc(void *p, size_t n)
Resizes the memory block pointed to by *p* to *n* bytes. The contents will
be unchanged to the minimum of the old and the new sizes. If *p* is *NULL*,
the call is equivalent to ``PyMem_RawMalloc(n)``; else if *n* is equal to
zero, the memory block is resized but is not freed, and the returned pointer
is non-*NULL*. Unless *p* is *NULL*, it must have been returned by a
previous call to :c:func:`PyMem_RawMalloc` or :c:func:`PyMem_RawRealloc`. If
the request fails, :c:func:`PyMem_RawRealloc` returns *NULL* and *p* remains
a valid pointer to the previous memory area.
.. c:function:: void PyMem_RawFree(void *p)
Frees the memory block pointed to by *p*, which must have been returned by a
previous call to :c:func:`PyMem_RawMalloc` or :c:func:`PyMem_RawRealloc`.
Otherwise, or if ``PyMem_Free(p)`` has been called before, undefined
behavior occurs. If *p* is *NULL*, no operation is performed.
.. _memoryinterface:
Memory Interface
......@@ -91,8 +133,16 @@ Memory Interface
The following function sets, modeled after the ANSI C standard, but specifying
behavior when requesting zero bytes, are available for allocating and releasing
memory from the Python heap:
memory from the Python heap.
The default memory block allocator uses the following functions:
:c:func:`malloc`, :c:func:`realloc` and :c:func:`free`; call ``malloc(1)`` when
requesting zero bytes.
.. warning::
The :term:`GIL <global interpreter lock>` must be held when using these
functions.
.. c:function:: void* PyMem_Malloc(size_t n)
......@@ -155,6 +205,125 @@ versions and is therefore deprecated in extension modules.
:c:func:`PyMem_NEW`, :c:func:`PyMem_RESIZE`, :c:func:`PyMem_DEL`.
Customize Memory Allocators
===========================
.. versionadded:: 3.4
.. c:type:: PyMemAllocator
Structure used to describe a memory block allocator. The structure has
four fields:
+----------------------------------------------------------+---------------------------------------+
| Field | Meaning |
+==========================================================+=======================================+
| ``void *ctx`` | user context passed as first argument |
+----------------------------------------------------------+---------------------------------------+
| ``void* malloc(void *ctx, size_t size)`` | allocate a memory block |
+----------------------------------------------------------+---------------------------------------+
| ``void* realloc(void *ctx, void *ptr, size_t new_size)`` | allocate or resize a memory block |
+----------------------------------------------------------+---------------------------------------+
| ``void free(void *ctx, void *ptr)`` | free a memory block |
+----------------------------------------------------------+---------------------------------------+
.. c:type:: PyMemAllocatorDomain
Enum used to identify an allocator domain. Domains:
* :c:data:`PYMEM_DOMAIN_RAW`: functions :c:func:`PyMem_RawMalloc`,
:c:func:`PyMem_RawRealloc` and :c:func:`PyMem_RawFree`
* :c:data:`PYMEM_DOMAIN_MEM`: functions :c:func:`PyMem_Malloc`,
:c:func:`PyMem_Realloc` and :c:func:`PyMem_Free`
* :c:data:`PYMEM_DOMAIN_OBJ`: functions :c:func:`PyObject_Malloc`,
:c:func:`PyObject_Realloc` and :c:func:`PyObject_Free`
.. c:function:: void PyMem_GetAllocator(PyMemAllocatorDomain domain, PyMemAllocator *allocator)
Get the memory block allocator of the specified domain.
.. c:function:: void PyMem_SetAllocator(PyMemAllocatorDomain domain, PyMemAllocator *allocator)
Set the memory block allocator of the specified domain.
The new allocator must return a distinct non-NULL pointer when requesting
zero bytes.
For the :c:data:`PYMEM_DOMAIN_RAW` domain, the allocator must be
thread-safe: the :term:`GIL <global interpreter lock>` is not held when the
allocator is called.
If the new allocator is not a hook (does not call the previous allocator),
the :c:func:`PyMem_SetupDebugHooks` function must be called to reinstall the
debug hooks on top on the new allocator.
.. c:function:: void PyMem_SetupDebugHooks(void)
Setup hooks to detect bugs in the following Python memory allocator
functions:
- :c:func:`PyMem_RawMalloc`, :c:func:`PyMem_RawRealloc`,
:c:func:`PyMem_RawFree`
- :c:func:`PyMem_Malloc`, :c:func:`PyMem_Realloc`, :c:func:`PyMem_Free`
- :c:func:`PyObject_Malloc`, :c:func:`PyObject_Realloc`,
:c:func:`PyObject_Free`
Newly allocated memory is filled with the byte ``0xCB``, freed memory is
filled with the byte ``0xDB``. Additionnal checks:
- detect API violations, ex: :c:func:`PyObject_Free` called on a buffer
allocated by :c:func:`PyMem_Malloc`
- detect write before the start of the buffer (buffer underflow)
- detect write after the end of the buffer (buffer overflow)
The function does nothing if Python is not compiled is debug mode.
Customize PyObject Arena Allocator
==================================
Python has a *pymalloc* allocator for allocations smaller than 512 bytes. This
allocator is optimized for small objects with a short lifetime. It uses memory
mappings called "arenas" with a fixed size of 256 KB. It falls back to
:c:func:`PyMem_Malloc` and :c:func:`PyMem_Realloc` for allocations larger than
512 bytes. *pymalloc* is the default allocator used by
:c:func:`PyObject_Malloc`.
The default arena allocator uses the following functions:
* :c:func:`VirtualAlloc` and :c:func:`VirtualFree` on Windows,
* :c:func:`mmap` and :c:func:`munmap` if available,
* :c:func:`malloc` and :c:func:`free` otherwise.
.. versionadded:: 3.4
.. c:type:: PyObjectArenaAllocator
Structure used to describe an arena allocator. The structure has
three fields:
+--------------------------------------------------+---------------------------------------+
| Field | Meaning |
+==================================================+=======================================+
| ``void *ctx`` | user context passed as first argument |
+--------------------------------------------------+---------------------------------------+
| ``void* alloc(void *ctx, size_t size)`` | allocate an arena of size bytes |
+--------------------------------------------------+---------------------------------------+
| ``void free(void *ctx, size_t size, void *ptr)`` | free an arena |
+--------------------------------------------------+---------------------------------------+
.. c:function:: PyObject_GetArenaAllocator(PyObjectArenaAllocator *allocator)
Get the arena allocator.
.. c:function:: PyObject_SetArenaAllocator(PyObjectArenaAllocator *allocator)
Set the arena allocator.
.. _memoryexamples:
Examples
......
......@@ -112,21 +112,11 @@ Security improvements:
Please read on for a comprehensive list of user-facing changes.
.. PEP-sized items next.
.. _pep-4XX:
.. PEP 4XX: Example PEP
.. ====================
.. (Implemented by Foo Bar.)
.. .. seealso::
:pep:`4XX` - Example PEP
PEP written by Example Author
PEP 445: Add new APIs to customize Python memory allocators
===========================================================
The :pep:`445` adds new Application Programming Interfaces (API) to customize
Python memory allocators.
......
......@@ -94,9 +94,9 @@ PyObject_{New, NewVar, Del}.
the object gets initialized via PyObject_{Init, InitVar} after obtaining
the raw memory.
*/
PyAPI_FUNC(void *) PyObject_Malloc(size_t);
PyAPI_FUNC(void *) PyObject_Realloc(void *, size_t);
PyAPI_FUNC(void) PyObject_Free(void *);
PyAPI_FUNC(void *) PyObject_Malloc(size_t size);
PyAPI_FUNC(void *) PyObject_Realloc(void *ptr, size_t new_size);
PyAPI_FUNC(void) PyObject_Free(void *ptr);
/* This function returns the number of allocated memory blocks, regardless of size */
PyAPI_FUNC(Py_ssize_t) _Py_GetAllocatedBlocks(void);
......@@ -106,41 +106,15 @@ PyAPI_FUNC(Py_ssize_t) _Py_GetAllocatedBlocks(void);
#ifndef Py_LIMITED_API
PyAPI_FUNC(void) _PyObject_DebugMallocStats(FILE *out);
#endif /* #ifndef Py_LIMITED_API */
#ifdef PYMALLOC_DEBUG /* WITH_PYMALLOC && PYMALLOC_DEBUG */
PyAPI_FUNC(void *) _PyObject_DebugMalloc(size_t nbytes);
PyAPI_FUNC(void *) _PyObject_DebugRealloc(void *p, size_t nbytes);
PyAPI_FUNC(void) _PyObject_DebugFree(void *p);
PyAPI_FUNC(void) _PyObject_DebugDumpAddress(const void *p);
PyAPI_FUNC(void) _PyObject_DebugCheckAddress(const void *p);
PyAPI_FUNC(void *) _PyObject_DebugMallocApi(char api, size_t nbytes);
PyAPI_FUNC(void *) _PyObject_DebugReallocApi(char api, void *p, size_t nbytes);
PyAPI_FUNC(void) _PyObject_DebugFreeApi(char api, void *p);
PyAPI_FUNC(void) _PyObject_DebugCheckAddressApi(char api, const void *p);
PyAPI_FUNC(void *) _PyMem_DebugMalloc(size_t nbytes);
PyAPI_FUNC(void *) _PyMem_DebugRealloc(void *p, size_t nbytes);
PyAPI_FUNC(void) _PyMem_DebugFree(void *p);
#define PyObject_MALLOC _PyObject_DebugMalloc
#define PyObject_Malloc _PyObject_DebugMalloc
#define PyObject_REALLOC _PyObject_DebugRealloc
#define PyObject_Realloc _PyObject_DebugRealloc
#define PyObject_FREE _PyObject_DebugFree
#define PyObject_Free _PyObject_DebugFree
#else /* WITH_PYMALLOC && ! PYMALLOC_DEBUG */
#endif
/* Macros */
#define PyObject_MALLOC PyObject_Malloc
#define PyObject_REALLOC PyObject_Realloc
#define PyObject_FREE PyObject_Free
#endif
#else /* ! WITH_PYMALLOC */
#define PyObject_MALLOC PyMem_MALLOC
#define PyObject_REALLOC PyMem_REALLOC
#define PyObject_FREE PyMem_FREE
#endif /* WITH_PYMALLOC */
#define PyObject_Del PyObject_Free
#define PyObject_DEL PyObject_FREE
#define PyObject_DEL PyObject_Free
/*
* Generic object allocator interface
......@@ -224,6 +198,26 @@ PyAPI_FUNC(PyVarObject *) _PyObject_NewVar(PyTypeObject *, Py_ssize_t);
constructor you would start directly with PyObject_Init/InitVar
*/
#ifndef Py_LIMITED_API
typedef struct {
/* user context passed as the first argument to the 2 functions */
void *ctx;
/* allocate an arena of size bytes */
void* (*alloc) (void *ctx, size_t size);
/* free an arena */
void (*free) (void *ctx, void *ptr, size_t size);
} PyObjectArenaAllocator;
/* Get the arena allocator. */
PyAPI_FUNC(void) PyObject_GetArenaAllocator(PyObjectArenaAllocator *allocator);
/* Set the arena allocator. */
PyAPI_FUNC(void) PyObject_SetArenaAllocator(PyObjectArenaAllocator *allocator);
#endif
/*
* Garbage Collection Support
* ==========================
......
......@@ -11,6 +11,11 @@
extern "C" {
#endif
PyAPI_FUNC(void *) PyMem_RawMalloc(size_t size);
PyAPI_FUNC(void *) PyMem_RawRealloc(void *ptr, size_t new_size);
PyAPI_FUNC(void) PyMem_RawFree(void *ptr);
/* BEWARE:
Each interface exports both functions and macros. Extension modules should
......@@ -49,21 +54,11 @@ extern "C" {
performed on failure (no exception is set, no warning is printed, etc).
*/
PyAPI_FUNC(void *) PyMem_Malloc(size_t);
PyAPI_FUNC(void *) PyMem_Realloc(void *, size_t);
PyAPI_FUNC(void) PyMem_Free(void *);
/* Starting from Python 1.6, the wrappers Py_{Malloc,Realloc,Free} are
no longer supported. They used to call PyErr_NoMemory() on failure. */
PyAPI_FUNC(void *) PyMem_Malloc(size_t size);
PyAPI_FUNC(void *) PyMem_Realloc(void *ptr, size_t new_size);
PyAPI_FUNC(void) PyMem_Free(void *ptr);
/* Macros. */
#ifdef PYMALLOC_DEBUG
/* Redirect all memory operations to Python's debugging allocator. */
#define PyMem_MALLOC _PyMem_DebugMalloc
#define PyMem_REALLOC _PyMem_DebugRealloc
#define PyMem_FREE _PyMem_DebugFree
#else /* ! PYMALLOC_DEBUG */
/* PyMem_MALLOC(0) means malloc(1). Some systems would return NULL
for malloc(0), which would be treated as an error. Some platforms
......@@ -71,13 +66,9 @@ PyAPI_FUNC(void) PyMem_Free(void *);
pymalloc. To solve these problems, allocate an extra byte. */
/* Returns NULL to indicate error if a negative size or size larger than
Py_ssize_t can represent is supplied. Helps prevents security holes. */
#define PyMem_MALLOC(n) ((size_t)(n) > (size_t)PY_SSIZE_T_MAX ? NULL \
: malloc((n) ? (n) : 1))
#define PyMem_REALLOC(p, n) ((size_t)(n) > (size_t)PY_SSIZE_T_MAX ? NULL \
: realloc((p), (n) ? (n) : 1))
#define PyMem_FREE free
#endif /* PYMALLOC_DEBUG */
#define PyMem_MALLOC(n) PyMem_Malloc(n)
#define PyMem_REALLOC(p, n) PyMem_Realloc(p, n)
#define PyMem_FREE(p) PyMem_Free(p)
/*
* Type-oriented memory interface
......@@ -115,6 +106,69 @@ PyAPI_FUNC(void) PyMem_Free(void *);
#define PyMem_Del PyMem_Free
#define PyMem_DEL PyMem_FREE
#ifndef Py_LIMITED_API
typedef enum {
/* PyMem_RawMalloc(), PyMem_RawRealloc() and PyMem_RawFree() */
PYMEM_DOMAIN_RAW,
/* PyMem_Malloc(), PyMem_Realloc() and PyMem_Free() */
PYMEM_DOMAIN_MEM,
/* PyObject_Malloc(), PyObject_Realloc() and PyObject_Free() */
PYMEM_DOMAIN_OBJ
} PyMemAllocatorDomain;
typedef struct {
/* user context passed as the first argument to the 3 functions */
void *ctx;
/* allocate a memory block */
void* (*malloc) (void *ctx, size_t size);
/* allocate or resize a memory block */
void* (*realloc) (void *ctx, void *ptr, size_t new_size);
/* release a memory block */
void (*free) (void *ctx, void *ptr);
} PyMemAllocator;
/* Get the memory block allocator of the specified domain. */
PyAPI_FUNC(void) PyMem_GetAllocator(PyMemAllocatorDomain domain,
PyMemAllocator *allocator);
/* Set the memory block allocator of the specified domain.
The new allocator must return a distinct non-NULL pointer when requesting
zero bytes.
For the PYMEM_DOMAIN_RAW domain, the allocator must be thread-safe: the GIL
is not held when the allocator is called.
If the new allocator is not a hook (don't call the previous allocator), the
PyMem_SetupDebugHooks() function must be called to reinstall the debug hooks
on top on the new allocator. */
PyAPI_FUNC(void) PyMem_SetAllocator(PyMemAllocatorDomain domain,
PyMemAllocator *allocator);
/* Setup hooks to detect bugs in the following Python memory allocator
functions:
- PyMem_RawMalloc(), PyMem_RawRealloc(), PyMem_RawFree()
- PyMem_Malloc(), PyMem_Realloc(), PyMem_Free()
- PyObject_Malloc(), PyObject_Realloc() and PyObject_Free()
Newly allocated memory is filled with the byte 0xCB, freed memory is filled
with the byte 0xDB. Additionnal checks:
- detect API violations, ex: PyObject_Free() called on a buffer allocated
by PyMem_Malloc()
- detect write before the start of the buffer (buffer underflow)
- detect write after the end of the buffer (buffer overflow)
The function does nothing if Python is not compiled is debug mode. */
PyAPI_FUNC(void) PyMem_SetupDebugHooks(void);
#endif
#ifdef __cplusplus
}
#endif
......
......@@ -10,6 +10,9 @@ What's New in Python 3.4.0 Alpha 1?
Core and Builtins
-----------------
- Issue #3329: Implement the PEP 445: Add new APIs to customize Python memory
allocators.
- Issue #18328: Reorder ops in PyThreadState_Delete*() functions. Now the
tstate is first removed from TLS and then deallocated.
......
......@@ -2511,6 +2511,161 @@ test_decref_doesnt_leak(PyObject *ob)
Py_RETURN_NONE;
}
static PyObject *
test_pymem_alloc0(PyObject *self)
{
void *ptr;
ptr = PyMem_Malloc(0);
if (ptr == NULL) {
PyErr_SetString(PyExc_RuntimeError, "PyMem_Malloc(0) returns NULL");
return NULL;
}
PyMem_Free(ptr);
ptr = PyObject_Malloc(0);
if (ptr == NULL) {
PyErr_SetString(PyExc_RuntimeError, "PyObject_Malloc(0) returns NULL");
return NULL;
}
PyObject_Free(ptr);
Py_RETURN_NONE;
}
typedef struct {
PyMemAllocator alloc;
size_t malloc_size;
void *realloc_ptr;
size_t realloc_new_size;
void *free_ptr;
} alloc_hook_t;
static void* hook_malloc (void* ctx, size_t size)
{
alloc_hook_t *hook = (alloc_hook_t *)ctx;
hook->malloc_size = size;
return hook->alloc.malloc(hook->alloc.ctx, size);
}
static void* hook_realloc (void* ctx, void* ptr, size_t new_size)
{
alloc_hook_t *hook = (alloc_hook_t *)ctx;
hook->realloc_ptr = ptr;
hook->realloc_new_size = new_size;
return hook->alloc.realloc(hook->alloc.ctx, ptr, new_size);
}
static void hook_free (void *ctx, void *ptr)
{
alloc_hook_t *hook = (alloc_hook_t *)ctx;
hook->free_ptr = ptr;
hook->alloc.free(hook->alloc.ctx, ptr);
}
static PyObject *
test_setallocators(PyMemAllocatorDomain domain)
{
PyObject *res = NULL;
const char *error_msg;
alloc_hook_t hook;
PyMemAllocator alloc;
size_t size, size2;
void *ptr, *ptr2;
hook.malloc_size = 0;
hook.realloc_ptr = NULL;
hook.realloc_new_size = 0;
hook.free_ptr = NULL;
alloc.ctx = &hook;
alloc.malloc = &hook_malloc;
alloc.realloc = &hook_realloc;
alloc.free = &hook_free;
PyMem_GetAllocator(domain, &hook.alloc);
PyMem_SetAllocator(domain, &alloc);
size = 42;
switch(domain)
{
case PYMEM_DOMAIN_RAW: ptr = PyMem_RawMalloc(size); break;
case PYMEM_DOMAIN_MEM: ptr = PyMem_Malloc(size); break;
case PYMEM_DOMAIN_OBJ: ptr = PyObject_Malloc(size); break;
default: ptr = NULL; break;
}
if (ptr == NULL) {
error_msg = "malloc failed";
goto fail;
}
if (hook.malloc_size != size) {
error_msg = "malloc invalid size";
goto fail;
}
size2 = 200;
switch(domain)
{
case PYMEM_DOMAIN_RAW: ptr2 = PyMem_RawRealloc(ptr, size2); break;
case PYMEM_DOMAIN_MEM: ptr2 = PyMem_Realloc(ptr, size2); break;
case PYMEM_DOMAIN_OBJ: ptr2 = PyObject_Realloc(ptr, size2); break;
}
if (ptr2 == NULL) {
error_msg = "realloc failed";
goto fail;
}
if (hook.realloc_ptr != ptr
|| hook.realloc_new_size != size2) {
error_msg = "realloc invalid parameters";
goto fail;
}
switch(domain)
{
case PYMEM_DOMAIN_RAW: PyMem_RawFree(ptr2); break;
case PYMEM_DOMAIN_MEM: PyMem_Free(ptr2); break;
case PYMEM_DOMAIN_OBJ: PyObject_Free(ptr2); break;
}
if (hook.free_ptr != ptr2) {
error_msg = "free invalid pointer";
goto fail;
}
Py_INCREF(Py_None);
res = Py_None;
goto finally;
fail:
PyErr_SetString(PyExc_RuntimeError, error_msg);
finally:
PyMem_SetAllocator(domain, &hook.alloc);
return res;
}
static PyObject *
test_pymem_setrawallocators(PyObject *self)
{
return test_setallocators(PYMEM_DOMAIN_RAW);
}
static PyObject *
test_pymem_setallocators(PyObject *self)
{
return test_setallocators(PYMEM_DOMAIN_MEM);
}
static PyObject *
test_pyobject_setallocators(PyObject *self)
{
return test_setallocators(PYMEM_DOMAIN_OBJ);
}
static PyMethodDef TestMethods[] = {
{"raise_exception", raise_exception, METH_VARARGS},
{"raise_memoryerror", (PyCFunction)raise_memoryerror, METH_NOARGS},
......@@ -2611,6 +2766,14 @@ static PyMethodDef TestMethods[] = {
{"pytime_object_to_time_t", test_pytime_object_to_time_t, METH_VARARGS},
{"pytime_object_to_timeval", test_pytime_object_to_timeval, METH_VARARGS},
{"pytime_object_to_timespec", test_pytime_object_to_timespec, METH_VARARGS},
{"test_pymem",
(PyCFunction)test_pymem_alloc0, METH_NOARGS},
{"test_pymem_alloc0",
(PyCFunction)test_pymem_setrawallocators, METH_NOARGS},
{"test_pymem_setallocators",
(PyCFunction)test_pymem_setallocators, METH_NOARGS},
{"test_pyobject_setallocators",
(PyCFunction)test_pyobject_setallocators, METH_NOARGS},
{NULL, NULL} /* sentinel */
};
......
......@@ -1859,26 +1859,6 @@ PyTypeObject *_PyCapsule_hack = &PyCapsule_Type;
Py_ssize_t (*_Py_abstract_hack)(PyObject *) = PyObject_Size;
/* Python's malloc wrappers (see pymem.h) */
void *
PyMem_Malloc(size_t nbytes)
{
return PyMem_MALLOC(nbytes);
}
void *
PyMem_Realloc(void *p, size_t nbytes)
{
return PyMem_REALLOC(p, nbytes);
}
void
PyMem_Free(void *p)
{
PyMem_FREE(p);
}
void
_PyObject_DebugTypeStats(FILE *out)
{
......
This diff is collapsed.
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment