8.1 Thread State and the Global Interpreter Lock

The Python interpreter is not fully thread safe. In order to support multi-threaded Python programs, there's a global lock that must be held by the current thread before it can safely access Python objects. Without the lock, even the simplest operations could cause problems in a multi-threaded program: for example, when two threads simultaneously increment the reference count of the same object, the reference count could end up being incremented only once instead of twice.

Therefore, the rule exists that only the thread that has acquired the global interpreter lock may operate on Python objects or call Python/C API functions. In order to support multi-threaded Python programs, the interpreter regularly release and reacquires the lock -- by default, every ten bytecode instructions (this can be changed with sys.setcheckinterval()). The lock is also released and reacquired around potentially blocking I/O operations like reading or writing a file, so that other threads can run while the thread that requests the I/O is waiting for the I/O operation to complete.

The Python interpreter needs to keep some bookkeeping information separate per thread -- for this it uses a data structure called PyThreadState. This is new in Python 1.5; in earlier versions, such state was stored in global variables, and switching threads could cause problems. In particular, exception handling is now thread safe, when the application uses sys.exc_info() to access the exception last raised in the current thread.

There's one global variable left, however: the pointer to the current PyThreadState structure. While most thread packages have a way to store ``per-thread global data,'' Python's internal platform independent thread abstraction doesn't support this yet. Therefore, the current thread state must be manipulated explicitly.

This is easy enough in most cases. Most code manipulating the global interpreter lock has the following simple structure:

Save the thread state in a local variable.
Release the interpreter lock.
...Do some blocking I/O operation...
Reacquire the interpreter lock.
Restore the thread state from the local variable.

This is so common that a pair of macros exists to simplify it:

Py_BEGIN_ALLOW_THREADS
...Do some blocking I/O operation...
Py_END_ALLOW_THREADS

The Py_BEGIN_ALLOW_THREADS macro opens a new block and declares a hidden local variable; the Py_END_ALLOW_THREADS macro closes the block. Another advantage of using these two macros is that when Python is compiled without thread support, they are defined empty, thus saving the thread state and lock manipulations.

When thread support is enabled, the block above expands to the following code:

{
    PyThreadState *_save;
    _save = PyEval_SaveThread();
    ...Do some blocking I/O operation...
    PyEval_RestoreThread(_save);
}

Using even lower level primitives, we can get roughly the same effect as follows:

{
    PyThreadState *_save;
    _save = PyThreadState_Swap(NULL);
    PyEval_ReleaseLock();
    ...Do some blocking I/O operation...
    PyEval_AcquireLock();
    PyThreadState_Swap(_save);
}

There are some subtle differences; in particular, PyEval_RestoreThread() saves and restores the value of the global variable errno, since the lock manipulation does not guarantee that errno is left alone. Also, when thread support is disabled, PyEval_SaveThread() and PyEval_RestoreThread() don't manipulate the lock; in this case, PyEval_ReleaseLock() and PyEval_AcquireLock() are not available. This is done so that dynamically loaded extensions compiled with thread support enabled can be loaded by an interpreter that was compiled with disabled thread support.

The global interpreter lock is used to protect the pointer to the current thread state. When releasing the lock and saving the thread state, the current thread state pointer must be retrieved before the lock is released (since another thread could immediately acquire the lock and store its own thread state in the global variable). Reversely, when acquiring the lock and restoring the thread state, the lock must be acquired before storing the thread state pointer.

Why am I going on with so much detail about this? Because when threads are created from C, they don't have the global interpreter lock, nor is there a thread state data structure for them. Such threads must bootstrap themselves into existence, by first creating a thread state data structure, then acquiring the lock, and finally storing their thread state pointer, before they can start using the Python/C API. When they are done, they should reset the thread state pointer, release the lock, and finally free their thread state data structure.

When creating a thread data structure, you need to provide an interpreter state data structure. The interpreter state data structure hold global data that is shared by all threads in an interpreter, for example the module administration (sys.modules). Depending on your needs, you can either create a new interpreter state data structure, or share the interpreter state data structure used by the Python main thread (to access the latter, you must obtain the thread state and access its interp member; this must be done by a thread that is created by Python or by the main thread after Python is initialized).

XXX More?

PyInterpreterState
This data structure represents the state shared by a number of cooperating threads. Threads belonging to the same interpreter share their module administration and a few other internal items. There are no public members in this structure.

Threads belonging to different interpreters initially share nothing, except process state like available memory, open file descriptors and such. The global interpreter lock is also shared by all threads, regardless of to which interpreter they belong.

PyThreadState
This data structure represents the state of a single thread. The only public data member is PyInterpreterState *interp, which points to this thread's interpreter state.

void PyEval_InitThreads()
Initialize and acquire the global interpreter lock. It should be called in the main thread before creating a second thread or engaging in any other thread operations such as PyEval_ReleaseLock() or PyEval_ReleaseThread(tstate). It is not needed before calling PyEval_SaveThread() or PyEval_RestoreThread().

This is a no-op when called for a second time. It is safe to call this function before calling Py_Initialize().

When only the main thread exists, no lock operations are needed. This is a common situation (most Python programs do not use threads), and the lock operations slow the interpreter down a bit. Therefore, the lock is not created initially. This situation is equivalent to having acquired the lock: when there is only a single thread, all object accesses are safe. Therefore, when this function initializes the lock, it also acquires it. Before the Python thread  module creates a new thread, knowing that either it has the lock or the lock hasn't been created yet, it calls PyEval_InitThreads(). When this call returns, it is guaranteed that the lock has been created and that it has acquired it.

It is not safe to call this function when it is unknown which thread (if any) currently has the global interpreter lock.

This function is not available when thread support is disabled at compile time.

void PyEval_AcquireLock()
Acquire the global interpreter lock. The lock must have been created earlier. If this thread already has the lock, a deadlock ensues. This function is not available when thread support is disabled at compile time.

void PyEval_ReleaseLock()
Release the global interpreter lock. The lock must have been created earlier. This function is not available when thread support is disabled at compile time.

void PyEval_AcquireThread(PyThreadState *tstate)
Acquire the global interpreter lock and then set the current thread state to tstate, which should not be NULL. The lock must have been created earlier. If this thread already has the lock, deadlock ensues. This function is not available when thread support is disabled at compile time.

void PyEval_ReleaseThread(PyThreadState *tstate)
Reset the current thread state to NULL and release the global interpreter lock. The lock must have been created earlier and must be held by the current thread. The tstate argument, which must not be NULL, is only used to check that it represents the current thread state -- if it isn't, a fatal error is reported. This function is not available when thread support is disabled at compile time.

PyThreadState* PyEval_SaveThread()
Release the interpreter lock (if it has been created and thread support is enabled) and reset the thread state to NULL, returning the previous thread state (which is not NULL). If the lock has been created, the current thread must have acquired it. (This function is available even when thread support is disabled at compile time.)

void PyEval_RestoreThread(PyThreadState *tstate)
Acquire the interpreter lock (if it has been created and thread support is enabled) and set the thread state to tstate, which must not be NULL. If the lock has been created, the current thread must not have acquired it, otherwise deadlock ensues. (This function is available even when thread support is disabled at compile time.)

Py_BEGIN_ALLOW_THREADS
This macro expands to "{ PyThreadState *_save; _save = PyEval_SaveThread();". Note that it contains an opening brace; it must be matched with a following Py_END_ALLOW_THREADS macro. See above for further discussion of this macro. It is a no-op when thread support is disabled at compile time.

Py_END_ALLOW_THREADS
This macro expands to "PyEval_RestoreThread(_save); }". Note that it contains a closing brace; it must be matched with an earlier Py_BEGIN_ALLOW_THREADS macro. See above for further discussion of this macro. It is a no-op when thread support is disabled at compile time.

Py_BEGIN_BLOCK_THREADS
This macro expands to "PyEval_RestoreThread(_save);" i.e. it is equivalent to Py_END_ALLOW_THREADS without the closing brace. It is a no-op when thread support is disabled at compile time.

Py_BEGIN_UNBLOCK_THREADS
This macro expands to "_save = PyEval_SaveThread();" i.e. it is equivalent to Py_BEGIN_ALLOW_THREADS without the opening brace and variable declaration. It is a no-op when thread support is disabled at compile time.

All of the following functions are only available when thread support is enabled at compile time, and must be called only when the interpreter lock has been created.

PyInterpreterState* PyInterpreterState_New()
Create a new interpreter state object. The interpreter lock must be held.

void PyInterpreterState_Clear(PyInterpreterState *interp)
Reset all information in an interpreter state object. The interpreter lock must be held.

void PyInterpreterState_Delete(PyInterpreterState *interp)
Destroy an interpreter state object. The interpreter lock need not be held. The interpreter state must have been reset with a previous call to PyInterpreterState_Clear().

PyThreadState* PyThreadState_New(PyInterpreterState *interp)
Create a new thread state object belonging to the given interpreter object. The interpreter lock must be held.

void PyThreadState_Clear(PyThreadState *tstate)
Reset all information in a thread state object. The interpreter lock must be held.

void PyThreadState_Delete(PyThreadState *tstate)
Destroy a thread state object. The interpreter lock need not be held. The thread state must have been reset with a previous call to PyThreadState_Clear().

PyThreadState* PyThreadState_Get()
Return the current thread state. The interpreter lock must be held. When the current thread state is NULL, this issues a fatal error (so that the caller needn't check for NULL).

PyThreadState* PyThreadState_Swap(PyThreadState *tstate)
Swap the current thread state with the thread state given by the argument tstate, which may be NULL. The interpreter lock must be held.

guido@python.org