diff -r ff5648697ac7 Lib/test/test_file2k.py
--- a/Lib/test/test_file2k.py	Tue Apr 15 08:01:27 2014 -0700
+++ b/Lib/test/test_file2k.py	Wed Apr 16 19:01:50 2014 -0400
@@ -831,6 +831,23 @@
                 'sys.stdout.flush()')
             self.assertEqual(stdout, "\xe9")
 
+    def testFileReadIsEightBytes(self):
+        f = open(TESTFN, 'wt')
+        f.write('a' * 1000)
+        f.close()
+
+        f = open(TESTFN, 'rt')
+        # file.read() should always accept
+        # an eight-byte int as its buffersize parameter.
+        try:
+            data = f.read(2 ** 33)
+        except OverflowError:
+            self.fail("should never get overflow error!")
+        except MemoryError:
+            # this is fine, this means parsing succeeded
+            # but allocating the buffer failed
+            pass
+
 
 def test_main():
     # Historically, these tests have been sloppy about removing TESTFN.
diff -r ff5648697ac7 Objects/fileobject.c
--- a/Objects/fileobject.c	Tue Apr 15 08:01:27 2014 -0700
+++ b/Objects/fileobject.c	Wed Apr 16 19:01:50 2014 -0400
@@ -1051,7 +1051,7 @@
 static PyObject *
 file_read(PyFileObject *f, PyObject *args)
 {
-    long bytesrequested = -1;
+    Py_ssize_t bytesrequested = -1;
     size_t bytesread, buffersize, chunksize;
     PyObject *v;
 
@@ -1064,17 +1064,12 @@
         (f->f_bufend - f->f_bufptr) > 0 &&
         f->f_buf[0] != '\0')
         return err_iterbuffered();
-    if (!PyArg_ParseTuple(args, "|l:read", &bytesrequested))
+    if (!PyArg_ParseTuple(args, "|n:read", &bytesrequested))
         return NULL;
     if (bytesrequested < 0)
         buffersize = new_buffersize(f, (size_t)0);
     else
         buffersize = bytesrequested;
-    if (buffersize > PY_SSIZE_T_MAX) {
-        PyErr_SetString(PyExc_OverflowError,
-    "requested number of bytes is more than a Python string can hold");
-        return NULL;
-    }
     v = PyString_FromStringAndSize((char *)NULL, buffersize);
     if (v == NULL)
         return NULL;