https://github.com/python/cpython/commit/a6793662e71f07d7c0ee3872316d93f3e8c48b34
commit: a6793662e71f07d7c0ee3872316d93f3e8c48b34
branch: main
author: Mark Shannon <[email protected]>
committer: markshannon <[email protected]>
date: 2026-05-26T15:14:17+01:00
summary:
GH-126910: Make `_Py_get_machine_stack_pointer` return the actual stack pointer
(GH-149103)
* Make _Py_ReachedRecursionLimit inline again
* Remove _Py_MakeRecCheck replacing its use with _Py_ReachedRecursionLimit
* Move the check for C stack swtiching into _Py_CheckRecursiveCall
files:
M Include/internal/pycore_ceval.h
M Include/internal/pycore_pystate.h
M Include/internal/pycore_pythonrun.h
M Lib/test/test_pyexpat.py
M Python/ceval.c
M Python/jit.c
diff --git a/Include/internal/pycore_ceval.h b/Include/internal/pycore_ceval.h
index fd4221f0816d24b..06c4ca1619d7ce1 100644
--- a/Include/internal/pycore_ceval.h
+++ b/Include/internal/pycore_ceval.h
@@ -208,16 +208,16 @@ extern void _PyEval_DeactivateOpCache(void);
/* --- _Py_EnterRecursiveCall() ----------------------------------------- */
-static inline int _Py_MakeRecCheck(PyThreadState *tstate) {
+static inline int _Py_ReachedRecursionLimit(PyThreadState *tstate) {
uintptr_t here_addr = _Py_get_machine_stack_pointer();
_PyThreadStateImpl *_tstate = (_PyThreadStateImpl *)tstate;
- // Overflow if stack pointer is between soft limit and the base of the
hardware stack.
- // If it is below the hardware stack base, assume that we have the wrong
stack limits, and do nothing.
- // We could have the wrong stack limits because of limited platform
support, or user-space threads.
+ // Possible overflow if stack pointer is beyond the soft limit.
+ // _Py_CheckRecursiveCall will check for corner cases and
+ // report an error if there is an overflow.
#if _Py_STACK_GROWS_DOWN
- return here_addr < _tstate->c_stack_soft_limit && here_addr >=
_tstate->c_stack_soft_limit - 2 * _PyOS_STACK_MARGIN_BYTES;
+ return here_addr < _tstate->c_stack_soft_limit;
#else
- return here_addr > _tstate->c_stack_soft_limit && here_addr <=
_tstate->c_stack_soft_limit + 2 * _PyOS_STACK_MARGIN_BYTES;
+ return here_addr > _tstate->c_stack_soft_limit;
#endif
}
@@ -232,7 +232,7 @@ PyAPI_FUNC(int) _Py_CheckRecursiveCallPy(
static inline int _Py_EnterRecursiveCallTstate(PyThreadState *tstate,
const char *where) {
- return (_Py_MakeRecCheck(tstate) && _Py_CheckRecursiveCall(tstate, where));
+ return (_Py_ReachedRecursionLimit(tstate) &&
_Py_CheckRecursiveCall(tstate, where));
}
static inline int _Py_EnterRecursiveCall(const char *where) {
@@ -246,8 +246,6 @@ static inline void
_Py_LeaveRecursiveCallTstate(PyThreadState *tstate) {
PyAPI_FUNC(void) _Py_InitializeRecursionLimits(PyThreadState *tstate);
-PyAPI_FUNC(int) _Py_ReachedRecursionLimit(PyThreadState *tstate);
-
// Export for test_peg_generator
PyAPI_FUNC(int) _Py_ReachedRecursionLimitWithMargin(
PyThreadState *tstate,
diff --git a/Include/internal/pycore_pystate.h
b/Include/internal/pycore_pystate.h
index c9e918bceda9fc7..ca6819d2cd44730 100644
--- a/Include/internal/pycore_pystate.h
+++ b/Include/internal/pycore_pystate.h
@@ -316,15 +316,20 @@ static uintptr_t return_pointer_as_int(char* p) {
static inline uintptr_t
_Py_get_machine_stack_pointer(void) {
-#if _Py__has_builtin(__builtin_frame_address) || defined(__GNUC__)
- return (uintptr_t)__builtin_frame_address(0);
-#elif defined(_MSC_VER)
- return (uintptr_t)_AddressOfReturnAddress();
+ uintptr_t result;
+#if defined(_M_ARM64)
+ result = __getReg(31);
+#elif defined(_M_X64) || defined(_M_IX86)
+ result = (uintptr_t)_AddressOfReturnAddress();
+#elif defined(__aarch64__)
+ __asm__ ("mov %0, sp" : "=r" (result));
+#elif defined(__x86_64__)
+ __asm__("{movq %%rsp, %0" : "=r" (result));
#else
char here;
- /* Avoid compiler warning about returning stack address */
- return return_pointer_as_int(&here);
+ result = (uintptr_t)&here;
#endif
+ return result;
}
static inline intptr_t
diff --git a/Include/internal/pycore_pythonrun.h
b/Include/internal/pycore_pythonrun.h
index 2a544edc431e6b1..66dd7cd843b04fc 100644
--- a/Include/internal/pycore_pythonrun.h
+++ b/Include/internal/pycore_pythonrun.h
@@ -46,7 +46,8 @@ extern PyObject * _Py_CompileStringObjectWithModule(
* stack consumption of PyEval_EvalDefault */
#if (defined(Py_DEBUG) \
|| defined(_Py_ADDRESS_SANITIZER) \
- || defined(_Py_THREAD_SANITIZER))
+ || defined(_Py_THREAD_SANITIZER)) \
+ || defined(_Py_UNDEFINED_BEHAVIOR_SANITIZER)
# define _PyOS_LOG2_STACK_MARGIN 12
#else
# define _PyOS_LOG2_STACK_MARGIN 11
diff --git a/Lib/test/test_pyexpat.py b/Lib/test/test_pyexpat.py
index 10dca684accee3c..cddf52d569bcd31 100644
--- a/Lib/test/test_pyexpat.py
+++ b/Lib/test/test_pyexpat.py
@@ -861,7 +861,7 @@ def test_trigger_leak(self):
def test_deeply_nested_content_model(self):
# This should raise a RecursionError and not crash.
# See https://github.com/python/cpython/issues/145986.
- N = 500_000
+ N = 800_000
data = (
b'<!DOCTYPE root [\n<!ELEMENT root '
+ b'(a, ' * N + b'a' + b')' * N
diff --git a/Python/ceval.c b/Python/ceval.c
index 5661200e74d0a55..5dbb116ded165ec 100644
--- a/Python/ceval.c
+++ b/Python/ceval.c
@@ -49,20 +49,6 @@ _Py_ReachedRecursionLimitWithMargin(PyThreadState *tstate,
int margin_count)
#endif
}
-void
-_Py_EnterRecursiveCallUnchecked(PyThreadState *tstate)
-{
- uintptr_t here_addr = _Py_get_machine_stack_pointer();
- _PyThreadStateImpl *_tstate = (_PyThreadStateImpl *)tstate;
-#if _Py_STACK_GROWS_DOWN
- if (here_addr < _tstate->c_stack_hard_limit) {
-#else
- if (here_addr > _tstate->c_stack_hard_limit) {
-#endif
- Py_FatalError("Unchecked stack overflow.");
- }
-}
-
#if defined(__s390x__)
# define Py_C_STACK_SIZE 320000
#elif defined(_WIN32)
@@ -278,7 +264,7 @@ PyUnstable_ThreadState_ResetStackProtection(PyThreadState
*tstate)
/* The function _Py_EnterRecursiveCallTstate() only calls
_Py_CheckRecursiveCall()
- if the stack pointer is between the stack base and c_stack_hard_limit. */
+ if the stack pointer is beyond c_stack_soft_limit. */
int
_Py_CheckRecursiveCall(PyThreadState *tstate, const char *where)
{
@@ -287,16 +273,21 @@ _Py_CheckRecursiveCall(PyThreadState *tstate, const char
*where)
assert(_tstate->c_stack_soft_limit != 0);
assert(_tstate->c_stack_hard_limit != 0);
#if _Py_STACK_GROWS_DOWN
- assert(here_addr >= _tstate->c_stack_hard_limit -
_PyOS_STACK_MARGIN_BYTES);
if (here_addr < _tstate->c_stack_hard_limit) {
- /* Overflowing while handling an overflow. Give up. */
+ if (here_addr < _tstate->c_stack_hard_limit -
_PyOS_STACK_MARGIN_BYTES) {
+ // Far out of bounds -- Assume stack switching has occurred
+ return 0;
+ }
int kbytes_used = (int)(_tstate->c_stack_top - here_addr)/1024;
#else
- assert(here_addr <= _tstate->c_stack_hard_limit +
_PyOS_STACK_MARGIN_BYTES);
if (here_addr > _tstate->c_stack_hard_limit) {
- /* Overflowing while handling an overflow. Give up. */
+ if (here_addr > _tstate->c_stack_hard_limit +
_PyOS_STACK_MARGIN_BYTES) {
+ // Far out of bounds -- Assume stack switching has occurred
+ return 0;
+ }
int kbytes_used = (int)(here_addr - _tstate->c_stack_top)/1024;
#endif
+ /* Too much stack used to safely raise an exception. Give up. */
char buffer[80];
snprintf(buffer, 80, "Unrecoverable stack overflow (used %d kB)%s",
kbytes_used, where);
Py_FatalError(buffer);
@@ -1146,19 +1137,6 @@ _PyEval_GetIter(_PyStackRef iterable, _PyStackRef
*index_or_null, int yield_from
return PyStackRef_FromPyObjectSteal(iter_o);
}
-Py_NO_INLINE int
-_Py_ReachedRecursionLimit(PyThreadState *tstate) {
- uintptr_t here_addr = _Py_get_machine_stack_pointer();
- _PyThreadStateImpl *_tstate = (_PyThreadStateImpl *)tstate;
- assert(_tstate->c_stack_hard_limit != 0);
-#if _Py_STACK_GROWS_DOWN
- return here_addr <= _tstate->c_stack_soft_limit;
-#else
- return here_addr >= _tstate->c_stack_soft_limit;
-#endif
-}
-
-
#if (defined(__GNUC__) && __GNUC__ >= 10 && !defined(__clang__)) &&
defined(__x86_64__)
/*
* gh-129987: The SLP autovectorizer can cause poor code generation for
@@ -1169,7 +1147,7 @@ _Py_ReachedRecursionLimit(PyThreadState *tstate) {
* (prior to GCC 9, 40% performance drop), so we have to selectively disable
* it.
*/
-#define DONT_SLP_VECTORIZE __attribute__((optimize ("no-tree-slp-vectorize")))
+#define DONT_SLP_VECTORIZE __attribute__((optimize ("no-tree-slp-vectorize",
"no-omit-frame-pointer")))
#else
#define DONT_SLP_VECTORIZE
#endif
diff --git a/Python/jit.c b/Python/jit.c
index 67dd88f510040eb..ade8b8d36e789d4 100644
--- a/Python/jit.c
+++ b/Python/jit.c
@@ -33,7 +33,7 @@
#include "pycore_jit.h"
-// Memory management stuff:
////////////////////////////////////////////////////
+// Memory management stuff: ///////////////////////////////////////////////////
#ifndef MS_WINDOWS
#include <sys/mman.h>
_______________________________________________
Python-checkins mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3//lists/python-checkins.python.org
Member address: [email protected]
