From 675bad714584939e688145ea5e0a66548695595b Mon Sep 17 00:00:00 2001
From: Julian Taylor <juliantaylor108@gmail.com>
Date: Mon, 17 Sep 2018 19:09:27 +0200
Subject: [PATCH] BUG: fix cached allocations without the GIL

There currently only is a global cache for small allocations so the
functions must be called while the GIL is held.
Ensure this by checking the GIL state in debug mode (which is tested via
a ci configuration).
Closes gh-11942
---
 numpy/core/src/multiarray/alloc.c          |  9 +++++++++
 numpy/core/src/multiarray/item_selection.c | 22 +++++++++++-----------
 2 files changed, 20 insertions(+), 11 deletions(-)

diff --git a/numpy/core/src/multiarray/alloc.c b/numpy/core/src/multiarray/alloc.c
index f8305d1150a4..97935f171e24 100644
--- a/numpy/core/src/multiarray/alloc.c
+++ b/numpy/core/src/multiarray/alloc.c
@@ -35,6 +35,13 @@ typedef struct {
 static cache_bucket datacache[NBUCKETS];
 static cache_bucket dimcache[NBUCKETS_DIM];
 
+/* as the cache is managed in global variables verify the GIL is held */
+#if defined(NPY_PY3K)
+#define NPY_CHECK_GIL_HELD() PyGILState_Check()
+#else
+#define NPY_CHECK_GIL_HELD() 1
+#endif
+
 /*
  * very simplistic small memory block cache to avoid more expensive libc
  * allocations
@@ -47,6 +54,7 @@ _npy_alloc_cache(npy_uintp nelem, npy_uintp esz, npy_uint msz,
 {
     assert((esz == 1 && cache == datacache) ||
            (esz == sizeof(npy_intp) && cache == dimcache));
+    assert(NPY_CHECK_GIL_HELD());
     if (nelem < msz) {
         if (cache[nelem].available > 0) {
             return cache[nelem].ptrs[--(cache[nelem].available)];
@@ -75,6 +83,7 @@ static NPY_INLINE void
 _npy_free_cache(void * p, npy_uintp nelem, npy_uint msz,
                 cache_bucket * cache, void (*dealloc)(void *))
 {
+    assert(NPY_CHECK_GIL_HELD());
     if (p != NULL && nelem < msz) {
         if (cache[nelem].available < NCACHE) {
             cache[nelem].ptrs[cache[nelem].available++] = p;
diff --git a/numpy/core/src/multiarray/item_selection.c b/numpy/core/src/multiarray/item_selection.c
index 486eb43ce4e5..c6dfcfdf5127 100644
--- a/numpy/core/src/multiarray/item_selection.c
+++ b/numpy/core/src/multiarray/item_selection.c
@@ -832,8 +832,6 @@ _new_sortlike(PyArrayObject *op, int axis, PyArray_SortFunc *sort,
     }
     size = it->size;
 
-    NPY_BEGIN_THREADS_DESCR(PyArray_DESCR(op));
-
     if (needcopy) {
         buffer = npy_alloc_cache(N * elsize);
         if (buffer == NULL) {
@@ -842,6 +840,8 @@ _new_sortlike(PyArrayObject *op, int axis, PyArray_SortFunc *sort,
         }
     }
 
+    NPY_BEGIN_THREADS_DESCR(PyArray_DESCR(op));
+
     while (size--) {
         char *bufptr = it->dataptr;
 
@@ -916,8 +916,8 @@ _new_sortlike(PyArrayObject *op, int axis, PyArray_SortFunc *sort,
     }
 
 fail:
-    npy_free_cache(buffer, N * elsize);
     NPY_END_THREADS_DESCR(PyArray_DESCR(op));
+    npy_free_cache(buffer, N * elsize);
     if (ret < 0 && !PyErr_Occurred()) {
         /* Out of memory during sorting or buffer creation */
         PyErr_NoMemory();
@@ -977,8 +977,6 @@ _new_argsortlike(PyArrayObject *op, int axis, PyArray_ArgSortFunc *argsort,
     }
     size = it->size;
 
-    NPY_BEGIN_THREADS_DESCR(PyArray_DESCR(op));
-
     if (needcopy) {
         valbuffer = npy_alloc_cache(N * elsize);
         if (valbuffer == NULL) {
@@ -995,6 +993,8 @@ _new_argsortlike(PyArrayObject *op, int axis, PyArray_ArgSortFunc *argsort,
         }
     }
 
+    NPY_BEGIN_THREADS_DESCR(PyArray_DESCR(op));
+
     while (size--) {
         char *valptr = it->dataptr;
         npy_intp *idxptr = (npy_intp *)rit->dataptr;
@@ -1078,9 +1078,9 @@ _new_argsortlike(PyArrayObject *op, int axis, PyArray_ArgSortFunc *argsort,
     }
 
 fail:
+    NPY_END_THREADS_DESCR(PyArray_DESCR(op));
     npy_free_cache(valbuffer, N * elsize);
     npy_free_cache(idxbuffer, N * sizeof(npy_intp));
-    NPY_END_THREADS_DESCR(PyArray_DESCR(op));
     if (ret < 0) {
         if (!PyErr_Occurred()) {
             /* Out of memory during sorting or buffer creation */
@@ -1495,13 +1495,13 @@ PyArray_LexSort(PyObject *sort_keys, int axis)
         char *valbuffer, *indbuffer;
         int *swaps;
 
-        valbuffer = npy_alloc_cache(N * maxelsize);
+        valbuffer = PyDataMem_NEW(N * maxelsize);
         if (valbuffer == NULL) {
             goto fail;
         }
-        indbuffer = npy_alloc_cache(N * sizeof(npy_intp));
+        indbuffer = PyDataMem_NEW(N * sizeof(npy_intp));
         if (indbuffer == NULL) {
-            npy_free_cache(indbuffer, N * sizeof(npy_intp));
+            PyDataMem_FREE(indbuffer);
             goto fail;
         }
         swaps = malloc(n*sizeof(int));
@@ -1544,8 +1544,8 @@ PyArray_LexSort(PyObject *sort_keys, int axis)
                                          sizeof(npy_intp), N, sizeof(npy_intp));
             PyArray_ITER_NEXT(rit);
         }
-        npy_free_cache(valbuffer, N * maxelsize);
-        npy_free_cache(indbuffer, N * sizeof(npy_intp));
+        PyDataMem_FREE(valbuffer);
+        PyDataMem_FREE(indbuffer);
         free(swaps);
     }
     else {