The multiple interpreter concept of mod_python is broken for Python extension modules since Python 2.3

• Key: MODPYTHON-77
• Type: Bug
• Status: Closed
• Resolution: Fixed
• Priority: Major
• Assignee: Graham Dumpleton
• Reporter: Boyan Boyadjiev
• Component/s: core
• Affects Version/s: 3.1.4
• Fix Version/s: 3.3.1

Description

The multiple interpreter concept of mod_python is broken for Python extension modules since Python 2.3 because of the PEP 311 (Simplified Global Interpreter Lock Acquisition for Extensions):
...
Limitations and Exclusions
    This proposal identifies a solution for extension authors with
    complex multi-threaded requirements, but that only require a
    single "PyInterpreterState". There is no attempt to cater for
    extensions that require multiple interpreter states. At the time
    of writing, no extension has been identified that requires
    multiple PyInterpreterStates, and indeed it is not clear if that
    facility works correctly in Python itself.
...

For mod_python this means, that complex Python extensions won't work any more with Python >= 2.3, because they are supposed to work only with the first interpreter state initialized for the current process (a problem we experienced). The first interpreter state is not used by mod_python after the python_init is called.

One solution, which works fine for me, is to save the first interpreter state into the "interpreters" dictionary in the function python_init (MAIN_INTERPRETER is used as a key):

static int python_init(apr_pool_t *p, apr_pool_t *ptemp,
                       apr_pool_t *plog, server_rec *s)
{

    ...

    /* initialize global Python interpreter if necessary */
    if (! Py_IsInitialized())
    {

        /* initialze the interpreter */
        Py_Initialize();

#ifdef WITH_THREAD
        /* create and acquire the interpreter lock */
        PyEval_InitThreads();
#endif
        /* create the obCallBack dictionary */
        interpreters = PyDict_New();
        if (! interpreters) {
            ap_log_error(APLOG_MARK, APLOG_NOERRNO|APLOG_ERR, 0, s,
                         "python_init: PyDict_New() failed! No more memory?");
            exit(1);
        }

        {
            /*
            Workaround PEP 311 - Simplified Global Interpreter Lock Acquisition for Extensions
            BEGIN
            */
            PyObject *p = 0;
            interpreterdata * idata = (interpreterdata *)malloc(sizeof(interpreterdata));
            PyThreadState* currentThreadState = PyThreadState_Get();
            PyInterpreterState *istate = currentThreadState->interp;
            idata->istate = istate;
            /* obcallback will be created on first use */
            idata->obcallback = NULL;
            p = PyCObject_FromVoidPtr((void ) idata, NULL); /*p->refcout = 1*/
            PyDict_SetItemString(interpreters, MAIN_INTERPRETER, p); /*p->refcout = 2*/
            Py_DECREF(p); /*p->refcout = 1*/
            /*
            END
            Workaround PEP 311 - Simplified Global Interpreter Lock Acquisition for Extensions
            */
        }

        /* Release the thread state because we will never use
         * the main interpreter, only sub interpreters created later. */
        PyThreadState_Swap(NULL);

#ifdef WITH_THREAD
        /* release the lock; now other threads can run */
        PyEval_ReleaseLock();
#endif

    }
    return OK;
}


Another change I've made in the attached file is to Py_DECREF(p) in get_interpreter, which will remove leaky reference to the PyCObject with the interpreter data. This was not a real problem, but now I see fewer leaks in BoundsChecker :-).