Default the OOB to the "ud" component IFF the HNP finds itself on a node with a supported Infiniband device. Ensure that the daemons all pick the matching component by dictating the selection via mca param on the orted cmd line.

This commit was SVN r26582.

orte/mca/oob/base/oob_base_init.c

@@ -10,6 +10,8 @@
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2007      Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2012      Los Alamos National Security, LLC. All rights
+ *                         reserved.
  * $COPYRIGHT$
  * 
  * Additional copyrights may follow
@@ -27,6 +29,7 @@
 #include "opal/util/output.h"
 #include "opal/mca/base/base.h"
 
+#include "orte/runtime/orte_globals.h"
 #include "orte/mca/oob/oob.h"
 #include "orte/mca/oob/base/base.h"
 
@@ -56,7 +59,7 @@ int mca_oob_base_init(void)
 {
     opal_list_item_t *item;
     mca_base_component_list_item_t *cli;
-    mca_oob_base_component_t *component;
+    mca_oob_base_component_t *component, *s_component;
     mca_oob_t *module;
     mca_oob_t *s_module = NULL;
     int  s_priority = -1;
@@ -85,6 +88,7 @@ int mca_oob_base_init(void)
                 if(priority > s_priority) {
                     s_priority = priority;
                     s_module = module;
+                    s_component = component;
                 }
             }
         }
@@ -97,6 +101,10 @@ int mca_oob_base_init(void)
     }
 
    mca_oob = *s_module;
+
+   /* record the name of the selected component */
+   orte_selected_oob_component = strdup(s_component->oob_base.mca_component_name);
+   opal_output_verbose(10, mca_oob_base_output, "mca_oob_base_init: %s module selected\n", orte_selected_oob_component);
    return ORTE_SUCCESS;
 }
 

orte/mca/oob/ud/oob_ud_component.c

@@ -212,7 +212,14 @@ static mca_oob_t *mca_oob_ud_component_init(int *priority)
     struct ibv_device **devices;
     int num_devices, i, rc;
 
-    *priority = 0;
+    /* set the priority so that we will select this component
+     * IF devices are available. Note that only the HNP has
+     * the freedom to make this decision as we will use the
+     * mca param on the orted cmd line to dictate that it
+     * match the HNP's choice. The value gets ignored if
+     * the NULL module is returned
+     */
+    *priority = 100;
 
     opal_hash_table_init (&mca_oob_ud_component.ud_peers, 1024);
 

orte/mca/plm/base/plm_base_launch_support.c

@@ -992,6 +992,13 @@ int orte_plm_base_orted_append_basic_args(int *argc, char ***argv,
         }
     }
 
+    if (NULL != orte_selected_oob_component) {
+        /* ensure we all use the same OOB component */
+        opal_argv_append(argc, argv, "-mca");
+        opal_argv_append(argc, argv, "oob");
+        opal_argv_append(argc, argv, orte_selected_oob_component);
+    }
+
     return ORTE_SUCCESS;
 }
 

orte/runtime/orte_globals.c

@@ -195,6 +195,8 @@ opal_thread_t orte_progress_thread;
 opal_event_t orte_finalize_event;
 #endif
 
+char *orte_selected_oob_component = NULL;
+
 #endif /* !ORTE_DISABLE_FULL_RTE */
 
 int orte_debug_output = -1;

orte/runtime/orte_globals.h

@@ -720,6 +720,9 @@ ORTE_DECLSPEC extern bool orte_map_stddiag_to_stderr;
 /* maximum size of virtual machine - used to subdivide allocation */
 ORTE_DECLSPEC extern int orte_max_vm_size;
 
+/* record the selected oob component */
+ORTE_DECLSPEC extern char *orte_selected_oob_component;
+
 #endif /* ORTE_DISABLE_FULL_SUPPORT */
 
 END_C_DECLS