if (!new_page) {
                page_cache_release(old_page);
-
-               /* Logically this is OOM, not a SIGBUS, but an OOM
-                * could cause the kernel to go killing other
-                * processes which won't help the hugepage situation
-                * at all (?) */
-               return VM_FAULT_SIGBUS;
+               return VM_FAULT_OOM;
        }
 
        spin_unlock(&mm->page_table_lock);
                page = alloc_huge_page(vma, address);
                if (!page) {
                        hugetlb_put_quota(mapping);
-                       /*
-                        * No huge pages available. So this is an OOM
-                        * condition but we do not want to trigger the OOM
-                        * killer, so we return VM_FAULT_SIGBUS.
-                        *
-                        * A program using hugepages may fault with Bus Error
-                        * because no huge pages are available in the cpuset, per
-                        * memory policy or because all are in use!
-                        */
+                       ret = VM_FAULT_OOM;
                        goto out;
                }