bpo-33021: Release the GIL during fstat() calls (GH-6019)

fstat may block for long time if the file descriptor is on a
non-responsive NFS server, hanging all threads. Most fstat() calls are
handled by _Py_fstat(), releasing the GIL internally, but but
_Py_fstat_noraise() does not release the GIL, and most calls release the
GIL explicitly around it.

This patch fixes last 2 calls to _Py_fstat_no_raise(), avoiding hangs
when calling:
- mmap.mmap()
- os.urandom()
- random.seed()

Misc/NEWS.d/next/Library/2018-03-12-00-27-56.bpo-33021.m19B9T.rst

@@ -0,0 +1,2 @@
+Release the GIL during fstat() calls, avoiding hang of all threads when
+calling mmap.mmap(), os.urandom(), and random.seed().  Patch by Nir Soffer.

Modules/mmapmodule.c

@@ -1050,6 +1050,7 @@ static PyObject *
 new_mmap_object(PyTypeObject *type, PyObject *args, PyObject *kwdict)
 {
     struct _Py_stat_struct status;
+    int fstat_result;
     mmap_object *m_obj;
     Py_ssize_t map_size;
     off_t offset = 0;
@@ -1115,8 +1116,14 @@ new_mmap_object(PyTypeObject *type, PyObject *args, PyObject *kwdict)
     if (fd != -1)
         (void)fcntl(fd, F_FULLFSYNC);
 #endif
-    if (fd != -1 && _Py_fstat_noraise(fd, &status) == 0
-        && S_ISREG(status.st_mode)) {
+
+    if (fd != -1) {
+        Py_BEGIN_ALLOW_THREADS
+        fstat_result = _Py_fstat_noraise(fd, &status);
+        Py_END_ALLOW_THREADS
+    }
+
+    if (fd != -1 && fstat_result == 0 && S_ISREG(status.st_mode)) {
         if (map_size == 0) {
             if (status.st_size == 0) {
                 PyErr_SetString(PyExc_ValueError,

Python/bootstrap_hash.c

@@ -301,10 +301,15 @@ dev_urandom(char *buffer, Py_ssize_t size, int raise)
 
     if (raise) {
         struct _Py_stat_struct st;
+        int fstat_result;
 
         if (urandom_cache.fd >= 0) {
+            Py_BEGIN_ALLOW_THREADS
+            fstat_result = _Py_fstat_noraise(urandom_cache.fd, &st);
+            Py_END_ALLOW_THREADS
+
             /* Does the fd point to the same thing as before? (issue #21207) */
-            if (_Py_fstat_noraise(urandom_cache.fd, &st)
+            if (fstat_result
                 || st.st_dev != urandom_cache.st_dev
                 || st.st_ino != urandom_cache.st_ino) {
                 /* Something changed: forget the cached fd (but don't close it,