diff -r 9c5d2a0b51ce Lib/test/test_capi.py
--- a/Lib/test/test_capi.py	Mon Mar 14 15:40:14 2016 +0100
+++ b/Lib/test/test_capi.py	Mon Mar 14 16:56:30 2016 +0100
@@ -440,6 +440,7 @@ class EmbeddingTests(unittest.TestCase):
         self.maxDiff = None
         self.assertEqual(out.strip(), expected_output)
 
+
 class SkipitemTest(unittest.TestCase):
 
     def test_skipitem(self):
@@ -557,12 +558,13 @@ class Test_testcapi(unittest.TestCase):
                     test()
 
 
-class MallocTests(unittest.TestCase):
-    ENV = 'debug'
+class PyMemDebugTests(unittest.TestCase):
+    PYTHONMALLOC = 'debug'
 
     def check(self, code):
         with support.SuppressCrashReport():
-            out = assert_python_failure('-c', code, PYTHONMALLOC=self.ENV)
+            out = assert_python_failure('-c', code,
+                                        PYTHONMALLOC=self.PYTHONMALLOC)
         stderr = out.err
         return stderr.decode('ascii', 'replace')
 
@@ -596,20 +598,28 @@ class MallocTests(unittest.TestCase):
                  r"Fatal Python error: bad ID: Allocated using API 'm', verified using API 'r'\n")
         self.assertRegex(out, regex)
 
+    def test_pymem_without_gil(self):
+        code = 'import _testcapi; _testcapi.pymem_without_gil()'
+        out = self.check(code)
+        expected = ('Fatal Python error: Python memory allocator called '
+                    'without holding the GIL')
+        self.assertIn(expected, out)
 
-class MallocDebugTests(MallocTests):
-    ENV = 'malloc_debug'
+
+class PyMemMallocDebugTests(PyMemDebugTests):
+    PYTHONMALLOC = 'malloc_debug'
 
 
 @unittest.skipUnless(sysconfig.get_config_var('WITH_PYMALLOC') == 1,
                      'need pymalloc')
-class PymallocDebugTests(MallocTests):
-    ENV = 'pymalloc_debug'
+class PyMemPymallocDebugTests(PyMemDebugTests):
+    PYTHONMALLOC = 'pymalloc_debug'
 
 
 @unittest.skipUnless(Py_DEBUG, 'need Py_DEBUG')
-class DefaultMallocDebugTests(MallocTests):
-    ENV = ''
+class PyMemDefaultTests(PyMemDebugTests):
+    # test default allocator of Python compiled in debug mode
+    PYTHONMALLOC = ''
 
 
 if __name__ == "__main__":
diff -r 9c5d2a0b51ce Modules/_testcapimodule.c
--- a/Modules/_testcapimodule.c	Mon Mar 14 15:40:14 2016 +0100
+++ b/Modules/_testcapimodule.c	Mon Mar 14 16:56:30 2016 +0100
@@ -3643,6 +3643,20 @@ pymem_api_misuse(PyObject *self, PyObjec
     Py_RETURN_NONE;
 }
 
+static PyObject*
+pymem_without_gil(PyObject *self, PyObject *args)
+{
+    char *buffer;
+
+    Py_BEGIN_ALLOW_THREADS
+    buffer = PyMem_Malloc(10);
+    Py_END_ALLOW_THREADS
+
+    PyMem_Free(buffer);
+
+    Py_RETURN_NONE;
+}
+
 
 static PyMethodDef TestMethods[] = {
     {"raise_exception",         raise_exception,                 METH_VARARGS},
@@ -3827,6 +3841,7 @@ static PyMethodDef TestMethods[] = {
     {"get_recursion_depth", get_recursion_depth, METH_NOARGS},
     {"pymem_buffer_overflow", pymem_buffer_overflow, METH_NOARGS},
     {"pymem_api_misuse", pymem_api_misuse, METH_NOARGS},
+    {"pymem_without_gil", pymem_without_gil, METH_NOARGS},
     {NULL, NULL} /* sentinel */
 };
 
diff -r 9c5d2a0b51ce Objects/obmalloc.c
--- a/Objects/obmalloc.c	Mon Mar 14 15:40:14 2016 +0100
+++ b/Objects/obmalloc.c	Mon Mar 14 16:56:30 2016 +0100
@@ -16,10 +16,15 @@
 #define uptr    Py_uintptr_t
 
 /* Forward declaration */
+static void* _PyMem_DebugRawMalloc(void *ctx, size_t size);
+static void* _PyMem_DebugRawCalloc(void *ctx, size_t nelem, size_t elsize);
+static void* _PyMem_DebugRawRealloc(void *ctx, void *ptr, size_t size);
+static void _PyMem_DebugRawFree(void *ctx, void *p);
+
 static void* _PyMem_DebugMalloc(void *ctx, size_t size);
 static void* _PyMem_DebugCalloc(void *ctx, size_t nelem, size_t elsize);
+static void* _PyMem_DebugRealloc(void *ctx, void *ptr, size_t size);
 static void _PyMem_DebugFree(void *ctx, void *p);
-static void* _PyMem_DebugRealloc(void *ctx, void *ptr, size_t size);
 
 static void _PyObject_DebugDumpAddress(const void *p);
 static void _PyMem_DebugCheckAddress(char api_id, const void *p);
@@ -173,11 +178,14 @@ static struct {
     {'o', {NULL, PYOBJ_FUNCS}}
     };
 
-#define PYDBG_FUNCS _PyMem_DebugMalloc, _PyMem_DebugCalloc, _PyMem_DebugRealloc, _PyMem_DebugFree
+#define PYRAWDBG_FUNCS \
+    _PyMem_DebugRawMalloc, _PyMem_DebugRawCalloc, _PyMem_DebugRawRealloc, _PyMem_DebugRawFree
+#define PYDBG_FUNCS \
+    _PyMem_DebugMalloc, _PyMem_DebugCalloc, _PyMem_DebugRealloc, _PyMem_DebugFree
 
 static PyMemAllocatorEx _PyMem_Raw = {
 #ifdef Py_DEBUG
-    &_PyMem_Debug.raw, PYDBG_FUNCS
+    &_PyMem_Debug.raw, PYRAWDBG_FUNCS
 #else
     NULL, PYRAW_FUNCS
 #endif
@@ -260,6 +268,7 @@ int
 #undef PYRAW_FUNCS
 #undef PYMEM_FUNCS
 #undef PYOBJ_FUNCS
+#undef PYRAWDBG_FUNCS
 #undef PYDBG_FUNCS
 
 static PyObjectArenaAllocator _PyObject_Arena = {NULL,
@@ -300,17 +309,22 @@ PyMem_SetupDebugHooks(void)
     if (_PyMem_DebugEnabled())
         return;
 
+    alloc.malloc = _PyMem_DebugRawMalloc;
+    alloc.calloc = _PyMem_DebugRawCalloc;
+    alloc.realloc = _PyMem_DebugRawRealloc;
+    alloc.free = _PyMem_DebugRawFree;
+
+    if (_PyMem_Raw.malloc != _PyMem_DebugRawMalloc) {
+        alloc.ctx = &_PyMem_Debug.raw;
+        PyMem_GetAllocator(PYMEM_DOMAIN_RAW, &_PyMem_Debug.raw.alloc);
+        PyMem_SetAllocator(PYMEM_DOMAIN_RAW, &alloc);
+    }
+
     alloc.malloc = _PyMem_DebugMalloc;
     alloc.calloc = _PyMem_DebugCalloc;
     alloc.realloc = _PyMem_DebugRealloc;
     alloc.free = _PyMem_DebugFree;
 
-    if (_PyMem_Raw.malloc != _PyMem_DebugMalloc) {
-        alloc.ctx = &_PyMem_Debug.raw;
-        PyMem_GetAllocator(PYMEM_DOMAIN_RAW, &_PyMem_Debug.raw.alloc);
-        PyMem_SetAllocator(PYMEM_DOMAIN_RAW, &alloc);
-    }
-
     if (_PyMem.malloc != _PyMem_DebugMalloc) {
         alloc.ctx = &_PyMem_Debug.mem;
         PyMem_GetAllocator(PYMEM_DOMAIN_MEM, &_PyMem_Debug.mem.alloc);
@@ -1869,7 +1883,7 @@ p[2*S+n+S: 2*S+n+2*S]
 */
 
 static void *
-_PyMem_DebugAlloc(int use_calloc, void *ctx, size_t nbytes)
+_PyMem_DebugRawAlloc(int use_calloc, void *ctx, size_t nbytes)
 {
     debug_alloc_api_t *api = (debug_alloc_api_t *)ctx;
     uchar *p;           /* base address of malloc'ed block */
@@ -1906,18 +1920,18 @@ static void *
 }
 
 static void *
-_PyMem_DebugMalloc(void *ctx, size_t nbytes)
+_PyMem_DebugRawMalloc(void *ctx, size_t nbytes)
 {
-    return _PyMem_DebugAlloc(0, ctx, nbytes);
+    return _PyMem_DebugRawAlloc(0, ctx, nbytes);
 }
 
 static void *
-_PyMem_DebugCalloc(void *ctx, size_t nelem, size_t elsize)
+_PyMem_DebugRawCalloc(void *ctx, size_t nelem, size_t elsize)
 {
     size_t nbytes;
     assert(elsize == 0 || nelem <= PY_SSIZE_T_MAX / elsize);
     nbytes = nelem * elsize;
-    return _PyMem_DebugAlloc(1, ctx, nbytes);
+    return _PyMem_DebugRawAlloc(1, ctx, nbytes);
 }
 
 /* The debug free first checks the 2*SST bytes on each end for sanity (in
@@ -1926,7 +1940,7 @@ static void *
    Then calls the underlying free.
 */
 static void
-_PyMem_DebugFree(void *ctx, void *p)
+_PyMem_DebugRawFree(void *ctx, void *p)
 {
     debug_alloc_api_t *api = (debug_alloc_api_t *)ctx;
     uchar *q = (uchar *)p - 2*SST;  /* address returned from malloc */
@@ -1943,7 +1957,7 @@ static void
 }
 
 static void *
-_PyMem_DebugRealloc(void *ctx, void *p, size_t nbytes)
+_PyMem_DebugRawRealloc(void *ctx, void *p, size_t nbytes)
 {
     debug_alloc_api_t *api = (debug_alloc_api_t *)ctx;
     uchar *q = (uchar *)p, *oldq;
@@ -1953,7 +1967,7 @@ static void *
     int i;
 
     if (p == NULL)
-        return _PyMem_DebugAlloc(0, ctx, nbytes);
+        return _PyMem_DebugRawAlloc(0, ctx, nbytes);
 
     _PyMem_DebugCheckAddress(api->api_id, p);
     bumpserialno();
@@ -1996,6 +2010,44 @@ static void *
     return q;
 }
 
+static void
+_PyMem_DebugCheckGIL(void)
+{
+#ifdef WITH_THREAD
+    if (!PyGILState_Check())
+        Py_FatalError("Python memory allocator called "
+                      "without holding the GIL");
+#endif
+}
+
+static void *
+_PyMem_DebugMalloc(void *ctx, size_t nbytes)
+{
+    _PyMem_DebugCheckGIL();
+    return _PyMem_DebugRawMalloc(ctx, nbytes);
+}
+
+static void *
+_PyMem_DebugCalloc(void *ctx, size_t nelem, size_t elsize)
+{
+    _PyMem_DebugCheckGIL();
+    return _PyMem_DebugRawCalloc(ctx, nelem, elsize);
+}
+
+static void
+_PyMem_DebugFree(void *ctx, void *ptr)
+{
+    _PyMem_DebugCheckGIL();
+    return _PyMem_DebugRawFree(ctx, ptr);
+}
+
+static void *
+_PyMem_DebugRealloc(void *ctx, void *ptr, size_t nbytes)
+{
+    _PyMem_DebugCheckGIL();
+    return _PyMem_DebugRawRealloc(ctx, ptr, nbytes);
+}
+
 /* Check the forbidden bytes on both ends of the memory allocated for p.
  * If anything is wrong, print info to stderr via _PyObject_DebugDumpAddress,
  * and call Py_FatalError to kill the program.