bpo-39877: Fix PyEval_RestoreThread() for daemon threads (GH-18811)
* exit_thread_if_finalizing() does now access directly _PyRuntime variable, rather than using tstate->interp->runtime since tstate can be a dangling pointer after Py_Finalize() has been called. * exit_thread_if_finalizing() is now called *before* calling take_gil(). _PyRuntime.finalizing is an atomic variable, we don't need to hold the GIL to access it. * Add ensure_tstate_not_null() function to check that tstate is not NULL at runtime. Check tstate earlier. take_gil() does not longer check if tstate is NULL. Cleanup: * PyEval_RestoreThread() no longer saves/restores errno: it's already done inside take_gil(). * PyEval_AcquireLock(), PyEval_AcquireThread(), PyEval_RestoreThread() and _PyEval_EvalFrameDefault() now check if tstate is valid with the new is_tstate_valid() function which uses _PyMem_IsPtrFreed().
This commit is contained in:
parent
d5aa2e941c
commit
eb4e2ae2b8
|
@ -0,0 +1,5 @@
|
|||
Fix :c:func:`PyEval_RestoreThread` random crash at exit with daemon threads.
|
||||
It now accesses the ``_PyRuntime`` variable directly instead of using
|
||||
``tstate->interp->runtime``, since ``tstate`` can be a dangling pointer after
|
||||
:c:func:`Py_Finalize` has been called. Moreover, the daemon thread now exits
|
||||
before trying to take the GIL.
|
|
@ -188,6 +188,25 @@ static size_t opcache_global_misses = 0;
|
|||
#include "pythread.h"
|
||||
#include "ceval_gil.h"
|
||||
|
||||
static void
|
||||
ensure_tstate_not_null(const char *func, PyThreadState *tstate)
|
||||
{
|
||||
if (tstate == NULL) {
|
||||
_Py_FatalErrorFunc(func, "current thread state is NULL");
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
#ifndef NDEBUG
|
||||
static int is_tstate_valid(PyThreadState *tstate)
|
||||
{
|
||||
assert(!_PyMem_IsPtrFreed(tstate));
|
||||
assert(!_PyMem_IsPtrFreed(tstate->interp));
|
||||
return 1;
|
||||
}
|
||||
#endif
|
||||
|
||||
|
||||
int
|
||||
PyEval_ThreadsInitialized(void)
|
||||
{
|
||||
|
@ -208,6 +227,7 @@ PyEval_InitThreads(void)
|
|||
PyThread_init_thread();
|
||||
create_gil(gil);
|
||||
PyThreadState *tstate = _PyRuntimeState_GetThreadState(runtime);
|
||||
ensure_tstate_not_null(__func__, tstate);
|
||||
take_gil(ceval, tstate);
|
||||
|
||||
struct _pending_calls *pending = &ceval->pending;
|
||||
|
@ -235,14 +255,26 @@ _PyEval_FiniThreads(struct _ceval_runtime_state *ceval)
|
|||
}
|
||||
}
|
||||
|
||||
/* This function is designed to exit daemon threads immediately rather than
|
||||
taking the GIL if Py_Finalize() has been called.
|
||||
|
||||
The caller must *not* hold the GIL, since this function does not release
|
||||
the GIL before exiting the thread.
|
||||
|
||||
When this function is called by a daemon thread after Py_Finalize() has been
|
||||
called, the GIL does no longer exist.
|
||||
|
||||
tstate must be non-NULL. */
|
||||
static inline void
|
||||
exit_thread_if_finalizing(PyThreadState *tstate)
|
||||
{
|
||||
_PyRuntimeState *runtime = tstate->interp->runtime;
|
||||
/* _Py_Finalizing is protected by the GIL */
|
||||
/* bpo-39877: Access _PyRuntime directly rather than using
|
||||
tstate->interp->runtime to support calls from Python daemon threads.
|
||||
After Py_Finalize() has been called, tstate can be a dangling pointer:
|
||||
point to PyThreadState freed memory. */
|
||||
_PyRuntimeState *runtime = &_PyRuntime;
|
||||
PyThreadState *finalizing = _PyRuntimeState_GetFinalizing(runtime);
|
||||
if (finalizing != NULL && finalizing != tstate) {
|
||||
drop_gil(&runtime->ceval, tstate);
|
||||
PyThread_exit_thread();
|
||||
}
|
||||
}
|
||||
|
@ -280,13 +312,14 @@ void
|
|||
PyEval_AcquireLock(void)
|
||||
{
|
||||
_PyRuntimeState *runtime = &_PyRuntime;
|
||||
struct _ceval_runtime_state *ceval = &runtime->ceval;
|
||||
PyThreadState *tstate = _PyRuntimeState_GetThreadState(runtime);
|
||||
if (tstate == NULL) {
|
||||
Py_FatalError("current thread state is NULL");
|
||||
}
|
||||
take_gil(ceval, tstate);
|
||||
ensure_tstate_not_null(__func__, tstate);
|
||||
|
||||
exit_thread_if_finalizing(tstate);
|
||||
assert(is_tstate_valid(tstate));
|
||||
|
||||
struct _ceval_runtime_state *ceval = &runtime->ceval;
|
||||
take_gil(ceval, tstate);
|
||||
}
|
||||
|
||||
void
|
||||
|
@ -304,15 +337,18 @@ PyEval_ReleaseLock(void)
|
|||
void
|
||||
PyEval_AcquireThread(PyThreadState *tstate)
|
||||
{
|
||||
assert(tstate != NULL);
|
||||
ensure_tstate_not_null(__func__, tstate);
|
||||
|
||||
exit_thread_if_finalizing(tstate);
|
||||
assert(is_tstate_valid(tstate));
|
||||
|
||||
_PyRuntimeState *runtime = tstate->interp->runtime;
|
||||
struct _ceval_runtime_state *ceval = &runtime->ceval;
|
||||
|
||||
/* Check someone has called PyEval_InitThreads() to create the lock */
|
||||
assert(gil_created(&ceval->gil));
|
||||
|
||||
take_gil(ceval, tstate);
|
||||
exit_thread_if_finalizing(tstate);
|
||||
if (_PyThreadState_Swap(&runtime->gilstate, tstate) != NULL) {
|
||||
Py_FatalError("non-NULL old thread state");
|
||||
}
|
||||
|
@ -344,8 +380,9 @@ _PyEval_ReInitThreads(_PyRuntimeState *runtime)
|
|||
return;
|
||||
}
|
||||
recreate_gil(&ceval->gil);
|
||||
PyThreadState *current_tstate = _PyRuntimeState_GetThreadState(runtime);
|
||||
take_gil(ceval, current_tstate);
|
||||
PyThreadState *tstate = _PyRuntimeState_GetThreadState(runtime);
|
||||
ensure_tstate_not_null(__func__, tstate);
|
||||
take_gil(ceval, tstate);
|
||||
|
||||
struct _pending_calls *pending = &ceval->pending;
|
||||
pending->lock = PyThread_allocate_lock();
|
||||
|
@ -354,7 +391,7 @@ _PyEval_ReInitThreads(_PyRuntimeState *runtime)
|
|||
}
|
||||
|
||||
/* Destroy all threads except the current one */
|
||||
_PyThreadState_DeleteExcept(runtime, current_tstate);
|
||||
_PyThreadState_DeleteExcept(runtime, tstate);
|
||||
}
|
||||
|
||||
/* This function is used to signal that async exceptions are waiting to be
|
||||
|
@ -383,16 +420,16 @@ PyEval_SaveThread(void)
|
|||
void
|
||||
PyEval_RestoreThread(PyThreadState *tstate)
|
||||
{
|
||||
assert(tstate != NULL);
|
||||
ensure_tstate_not_null(__func__, tstate);
|
||||
|
||||
exit_thread_if_finalizing(tstate);
|
||||
assert(is_tstate_valid(tstate));
|
||||
|
||||
_PyRuntimeState *runtime = tstate->interp->runtime;
|
||||
struct _ceval_runtime_state *ceval = &runtime->ceval;
|
||||
assert(gil_created(&ceval->gil));
|
||||
|
||||
int err = errno;
|
||||
take_gil(ceval, tstate);
|
||||
exit_thread_if_finalizing(tstate);
|
||||
errno = err;
|
||||
|
||||
_PyThreadState_Swap(&runtime->gilstate, tstate);
|
||||
}
|
||||
|
@ -750,11 +787,14 @@ _PyEval_EvalFrameDefault(PyFrameObject *f, int throwflag)
|
|||
PyObject **fastlocals, **freevars;
|
||||
PyObject *retval = NULL; /* Return value */
|
||||
_PyRuntimeState * const runtime = &_PyRuntime;
|
||||
PyThreadState * const tstate = _PyRuntimeState_GetThreadState(runtime);
|
||||
struct _ceval_runtime_state * const ceval = &runtime->ceval;
|
||||
_Py_atomic_int * const eval_breaker = &ceval->eval_breaker;
|
||||
PyCodeObject *co;
|
||||
|
||||
PyThreadState * const tstate = _PyRuntimeState_GetThreadState(runtime);
|
||||
ensure_tstate_not_null(__func__, tstate);
|
||||
assert(is_tstate_valid(tstate));
|
||||
|
||||
/* when tracing we set things up so that
|
||||
|
||||
not (instr_lb <= current_bytecode_offset < instr_ub)
|
||||
|
@ -1242,11 +1282,11 @@ main_loop:
|
|||
|
||||
/* Other threads may run now */
|
||||
|
||||
take_gil(ceval, tstate);
|
||||
|
||||
/* Check if we should make a quick exit. */
|
||||
exit_thread_if_finalizing(tstate);
|
||||
|
||||
take_gil(ceval, tstate);
|
||||
|
||||
if (_PyThreadState_Swap(&runtime->gilstate, tstate) != NULL) {
|
||||
Py_FatalError("orphan tstate");
|
||||
}
|
||||
|
|
|
@ -180,15 +180,17 @@ drop_gil(struct _ceval_runtime_state *ceval, PyThreadState *tstate)
|
|||
#endif
|
||||
}
|
||||
|
||||
/* Take the GIL.
|
||||
|
||||
The function saves errno at entry and restores its value at exit.
|
||||
|
||||
tstate must be non-NULL. */
|
||||
static void
|
||||
take_gil(struct _ceval_runtime_state *ceval, PyThreadState *tstate)
|
||||
{
|
||||
if (tstate == NULL) {
|
||||
Py_FatalError("take_gil: NULL tstate");
|
||||
}
|
||||
int err = errno;
|
||||
|
||||
struct _gil_runtime_state *gil = &ceval->gil;
|
||||
int err = errno;
|
||||
MUTEX_LOCK(gil->mutex);
|
||||
|
||||
if (!_Py_atomic_load_relaxed(&gil->locked)) {
|
||||
|
@ -240,6 +242,7 @@ _ready:
|
|||
}
|
||||
|
||||
MUTEX_UNLOCK(gil->mutex);
|
||||
|
||||
errno = err;
|
||||
}
|
||||
|
||||
|
|
|
@ -1364,8 +1364,8 @@ Py_FinalizeEx(void)
|
|||
int malloc_stats = interp->config.malloc_stats;
|
||||
#endif
|
||||
|
||||
/* Remaining threads (e.g. daemon threads) will automatically exit
|
||||
after taking the GIL (in PyEval_RestoreThread()). */
|
||||
/* Remaining daemon threads will automatically exit
|
||||
when they attempt to take the GIL (ex: PyEval_RestoreThread()). */
|
||||
_PyRuntimeState_SetFinalizing(runtime, tstate);
|
||||
runtime->initialized = 0;
|
||||
runtime->core_initialized = 0;
|
||||
|
|
Loading…
Reference in New Issue