orte: use a unique top_session_dir directory when possible.

currently, top_session_dir is based on hostname, uid, HNP pid and job family.
there is a risk a top_session_dir exists and contains some old data when
a job starts, leading to undefined behavior.

if the app is started via mpirun, use a unique top_session_dir
mkdtemp("$TMP/ompi.<hostname>.<uid>/XXXXXX") that is passed to fork'ed MPI tasks
via the OPAL_MCA_PREFIX"orte_top_session_dir" environment variable.

if the app is direct launched, then the current behavior is unchanged.
direct launch behavior will be enhanced when PMIx is able to pass
a per-node directory (PMIX_NSDIR ?) to a direct launched task.

Author: ggouaillardet

orte/orted/orted_main.c

@@ -615,6 +615,11 @@ int orte_daemon(int argc, char *argv[])
         opal_argv_append_nosize(&singenv, env_str);
         free(env_str);
 
+        /* append the top session dir to the envars needed by the singleton */
+        asprintf(&env_str, OPAL_MCA_PREFIX"orte_top_session_dir=%s", orte_process_info.top_session_dir);
+        opal_argv_append_nosize(&singenv, env_str);
+        free(env_str);
+
         nptr = opal_argv_join(singenv, ',');
         opal_argv_free(singenv);
         /* create a string that contains our uri + sysinfo + PMIx server URI envars */

orte/util/session_dir.c

@@ -10,7 +10,7 @@
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2014      Cisco Systems, Inc.  All rights reserved.
- * Copyright (c) 2015      Research Organization for Information Science
+ * Copyright (c) 2015-2016 Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
  * Copyright (c) 2015      Intel, Inc. All rights reserved.
  * $COPYRIGHT$
@@ -134,13 +134,24 @@ static int _setup_tmpdir_base(void)
 static int _setup_top_session_dir(void)
 {
     int rc = ORTE_SUCCESS;
+    char *env;
     /* get the effective uid */
     uid_t uid = geteuid();
 
+
     /* construct the top_session_dir if we need */
     if (NULL == orte_process_info.top_session_dir) {
+        env = getenv(OPAL_MCA_PREFIX"orte_top_session_dir");
+        if (NULL != env) {
+            orte_process_info.tmpdir_base = strdup(env);
+            orte_process_info.top_session_dir = strdup(env);
+            return ORTE_SUCCESS;
+        }
+
+        assert(!ORTE_PROC_IS_APP || (NULL == getenv(OPAL_MCA_PREFIX"orte_launch")));
+
         if (ORTE_SUCCESS != (rc = _setup_tmpdir_base())) {
-            return rc;
+            goto exit;
         }
         if( NULL == orte_process_info.nodename ||
                 NULL == orte_process_info.tmpdir_base ){
@@ -156,6 +167,24 @@ static int _setup_top_session_dir(void)
             rc = ORTE_ERR_OUT_OF_RESOURCE;
             goto exit;
         }
+        if (!ORTE_PROC_IS_APP) {
+            char *dir = orte_process_info.top_session_dir;
+            if (ORTE_SUCCESS != (rc = orte_create_dir(dir))) {
+                goto exit;
+            }
+            if (0 > asprintf(&orte_process_info.top_session_dir, "%s/XXXXXX", dir)) {
+                free(dir);
+                orte_process_info.top_session_dir = NULL;
+                rc = ORTE_ERR_OUT_OF_RESOURCE;
+                goto exit;
+            }
+            free(dir);
+
+            if (NULL == mkdtemp(orte_process_info.top_session_dir)) {
+                rc = ORTE_ERROR;
+                goto exit;
+            }
+        }
     }
 exit:
     if( ORTE_SUCCESS != rc ){