Fix for bug #370

The POE ras did not correctly enter the number of slots per node. This fixes that.

This commit was SVN r11716.

orte/mca/ras/poe/ras_poe.h

@@ -34,7 +34,6 @@ extern "C" {
  */
 struct orte_ras_poe_component_t {
     orte_ras_base_component_t super;
-    int debug;
     int priority;
 };
 typedef struct orte_ras_poe_component_t orte_ras_poe_component_t;

orte/mca/ras/poe/ras_poe_component.c

@@ -20,6 +20,8 @@
 #include "orte/orte_constants.h"
 #include "opal/mca/base/base.h"
 #include "opal/mca/base/mca_base_param.h"
+#include "orte/mca/ras/ras.h"
+#include "orte/mca/ras/base/base.h"
 #include "orte/util/proc_info.h"
 #include "opal/util/output.h"
 #include "ras_poe.h"
@@ -68,10 +70,6 @@ static int orte_ras_poe_open(void)
 {
     mca_base_component_t *c = &mca_ras_poe_component.super.ras_version;
 
-    mca_base_param_reg_int(c, "debug",
-                           "Whether or not to enable debugging output for the poe ras component (0 or 1)",
-                           false, false, 0, &mca_ras_poe_component.debug);
-
     mca_base_param_reg_int(c, "priority",
                            "Priority of the poe ras component",
                            false , false, 100, &mca_ras_poe_component.priority);
@@ -88,9 +86,15 @@ static orte_ras_base_module_t *orte_ras_poe_init(int* priority)
         
     *priority = mca_ras_poe_component.priority;
 
-    if ( NULL != getenv("LOADL_PID") ) {
+    if ( NULL != getenv("LOADL_PROCESSOR_LIST") ) {
+        opal_output(orte_ras_base.ras_output,
+                    "ras:poe: available for selection with priority %d",
+                    mca_ras_poe_component.priority);
         return &orte_ras_poe_module;
     }
+
+    opal_output(orte_ras_base.ras_output,
+                "ras:poe: NOT available for selection");
     return NULL;
 }
 

orte/mca/ras/poe/ras_poe_module.c

@@ -38,26 +38,42 @@ static int orte_ras_poe_allocate(orte_jobid_t jobid)
     if (NULL == poe_node_str) {
         return ORTE_ERR_NOT_FOUND;
     }
-
+    
     /* poe_node_str is something like "nodeA nodeA nodeB nodeB" */
 
     names = opal_argv_copy(opal_argv_split(poe_node_str, ' '));
     nnode = opal_argv_count(names);
 
+    /* iterate through all the nodes listed. If a single node is listed more than once,
+     * it means that we have multiple slots allocated on that node */ 
     OBJ_CONSTRUCT(&nodes_list, opal_list_t);
     for (i = 0; i < nnode; i++) {
-        node = OBJ_NEW(orte_ras_node_t);
-        if (NULL == node) {
-            return ORTE_ERR_OUT_OF_RESOURCE;
+        /* check for duplicated nodes */
+        for (item = opal_list_get_first(&nodes_list); 
+             opal_list_get_end(&nodes_list) != item;
+             item = opal_list_get_next(item)) {
+             node = (orte_ras_node_t*) item;
+             if (0 == strcmp(node->node_name, names[i])) {
+                ++node->node_slots;
+                break;
+            }
+        }
+        
+        if(opal_list_get_end(&nodes_list) == item) {
+            /* we did not find a duplicate, so add a new item to the list */
+            node = OBJ_NEW(orte_ras_node_t);
+            if (NULL == node) {
+                return ORTE_ERR_OUT_OF_RESOURCE;
+            }
+            node->node_name = strdup(names[i]);
+            node->node_arch = NULL;
+            node->node_state = ORTE_NODE_STATE_UP;
+            node->node_cellid = 0;
+            node->node_slots_inuse = 0;
+            node->node_slots_max = 0;
+            node->node_slots = 1;
+            opal_list_append(&nodes_list, &node->super);
         }
-        node->node_name = strdup(names[i]);
-        node->node_arch = NULL;
-        node->node_state = ORTE_NODE_STATE_UP;
-        node->node_cellid = 0;
-        node->node_slots_inuse = 0;
-        node->node_slots_max = 0;
-        node->node_slots = 1;
-        opal_list_append(&nodes_list, &node->super);
     }
     ret = orte_ras_base_node_insert(&nodes_list);
     ret = orte_ras_base_allocate_nodes(jobid, &nodes_list);
@@ -66,7 +82,6 @@ static int orte_ras_poe_allocate(orte_jobid_t jobid)
         OBJ_RELEASE(item);    
     }
     OBJ_DESTRUCT(&nodes_list);
-
     return ret;
 }