Fix -H operations for multi-app case

Ralph Castain · Ralph Castain · commit f5a6b7f1e9b6 · 2018-10-11T09:30:01.000-07:00
Correctly aggregate slots across -H entries from each app. Take into
account any -H entry when computing nprocs when no value was given.

Signed-off-by: Ralph Castain &lt;rhc@open-mpi.org&gt;
diff --git a/orte/mca/rmaps/base/rmaps_base_map_job.c b/orte/mca/rmaps/base/rmaps_base_map_job.c
@@ -49,7 +49,7 @@ void orte_rmaps_base_map_job(int fd, short args, void *cbdata)
     orte_job_t *jdata;
     orte_node_t *node;
     int rc, i, ppx = 0;
-    bool did_map, given, pernode = false, persocket = false;
+    bool did_map, pernode = false, persocket = false;
     orte_rmaps_base_selected_module_t *mod;
     orte_job_t *parent;
     orte_vpid_t nprocs;
@@ -105,7 +105,6 @@ void orte_rmaps_base_map_job(int fd, short args, void *cbdata)
                 orte_std_cntr_t slots;
                 OBJ_CONSTRUCT(&nodes, opal_list_t);
                 orte_rmaps_base_get_target_nodes(&nodes, &slots, app, ORTE_MAPPING_BYNODE, true, true);
-                slots = 0;
                 if (pernode) {
                     slots = ppx * opal_list_get_size(&nodes);
                 } else if (persocket) {
@@ -115,34 +114,6 @@ void orte_rmaps_base_map_job(int fd, short args, void *cbdata)
                                                                           HWLOC_OBJ_SOCKET, 0,
                                                                           OPAL_HWLOC_AVAILABLE);
                     }
-                } else {
-                    /* if we are in a managed allocation, then all is good - otherwise,
-                     * we have to do a little more checking */
-                    if (!orte_managed_allocation) {
-                        /* if all the nodes have their slots given, then we are okay */
-                        given = true;
-                        OPAL_LIST_FOREACH(node, &nodes, orte_node_t) {
-                            if (!ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_SLOTS_GIVEN)) {
-                                given = false;
-                                break;
-                            }
-                        }
-                        /* if -host or -hostfile was given, and the slots were not,
-                         * then this is no longer allowed */
-                        if (!given &&
-                            (orte_get_attribute(&app->attributes, ORTE_APP_DASH_HOST, NULL, OPAL_STRING) ||
-                             orte_get_attribute(&app->attributes, ORTE_APP_HOSTFILE, NULL, OPAL_STRING))) {
-                            /* inform the user of the error */
-                            orte_show_help("help-orte-rmaps-base.txt", "num-procs-not-specified", true);
-                            OPAL_LIST_DESTRUCT(&nodes);
-                            OBJ_RELEASE(caddy);
-                            ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_MAP_FAILED);
-                            return;
-                        }
-                        OPAL_LIST_FOREACH(node, &nodes, orte_node_t) {
-                            slots += node->slots;
-                        }
-                    }
                 }
                 app->num_procs = slots;
                 OPAL_LIST_DESTRUCT(&nodes);
diff --git a/orte/mca/rmaps/base/rmaps_base_support_fns.c b/orte/mca/rmaps/base/rmaps_base_support_fns.c
@@ -469,12 +469,19 @@ int orte_rmaps_base_get_target_nodes(opal_list_t *allocated_nodes, orte_std_cntr
                 continue;
             }
             if (node->slots > node->slots_inuse) {
+                orte_std_cntr_t s;
+                /* check for any -host allocations */
+                if (orte_get_attribute(&app->attributes, ORTE_APP_DASH_HOST, (void**)&hosts, OPAL_STRING)) {
+                    s = orte_util_dash_host_compute_slots(node, hosts);
+                } else {
+                    s = node->slots - node->slots_inuse;
+                }
                 /* add the available slots */
                 OPAL_OUTPUT_VERBOSE((5, orte_rmaps_base_framework.framework_output,
                                      "%s node %s has %d slots available",
                                      ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
-                                     node->name, node->slots - node->slots_inuse));
-                num_slots += node->slots - node->slots_inuse;
+                                     node->name, s));
+                num_slots += s;
                 continue;
             }
             if (!(ORTE_MAPPING_NO_OVERSUBSCRIBE & ORTE_GET_MAPPING_DIRECTIVE(policy))) {
diff --git a/orte/util/dash_host/dash_host.c b/orte/util/dash_host/dash_host.c
@@ -10,7 +10,7 @@
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2013      Cisco Systems, Inc.  All rights reserved.
- * Copyright (c) 2014-2017 Intel, Inc.  All rights reserved.
+ * Copyright (c) 2014-2018 Intel, Inc.  All rights reserved.
  * Copyright (c) 2015      Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
  * Copyright (c) 2016      IBM Corporation.  All rights reserved.
@@ -42,6 +42,36 @@
 
 #include "dash_host.h"
 
+int orte_util_dash_host_compute_slots(orte_node_t *node, char *hosts)
+{
+    char **specs, *cptr;
+    int slots = 0;
+    int n;
+
+    specs = opal_argv_split(hosts, ',');
+
+    /* see if this node appears in the list */
+    for (n=0; NULL != specs[n]; n++) {
+        if (0 == strncmp(node->name, specs[n], strlen(node->name))) {
+            /* check if the #slots was specified */
+            if (NULL != (cptr = strchr(specs[n], ':'))) {
+                *cptr = '\0';
+                ++cptr;
+                if ('*' == *cptr || 0 == strcmp(cptr, "auto")) {
+                    slots += node->slots - node->slots_inuse;
+                } else {
+                    slots += strtol(cptr, NULL, 10);
+                }
+            } else {
+                ++slots;
+            }
+
+        }
+    }
+    opal_argv_free(specs);
+    return slots;
+}
+
 /* we can only enter this routine if no other allocation
  * was found, so we only need to know that finding any
  * relative node syntax should generate an immediate error
@@ -289,7 +319,7 @@ int orte_util_add_dash_host_nodes(opal_list_t *nodes,
                                      ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node->name));
                 if (ORTE_FLAG_TEST(nd, ORTE_NODE_FLAG_SLOTS_GIVEN)) {
                     /* transfer across the number of slots */
-                    node->slots = nd->slots;
+                    node->slots += nd->slots;
                     ORTE_FLAG_SET(node, ORTE_NODE_FLAG_SLOTS_GIVEN);
                 }
                 break;
diff --git a/orte/util/dash_host/dash_host.h b/orte/util/dash_host/dash_host.h
@@ -9,7 +9,7 @@
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
- * Copyright (c) 2014-2015 Intel, Inc. All rights reserved.
+ * Copyright (c) 2014-2018 Intel, Inc.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -27,6 +27,7 @@
 
 #include "opal/class/opal_list.h"
 
+#include "orte/runtime/orte_globals.h"
 
 BEGIN_C_DECLS
 
@@ -41,6 +42,8 @@ ORTE_DECLSPEC int orte_util_filter_dash_host_nodes(opal_list_t *nodes,
 ORTE_DECLSPEC int orte_util_get_ordered_dash_host_list(opal_list_t *nodes,
                                                        char *hosts);
 
+ORTE_DECLSPEC int orte_util_dash_host_compute_slots(orte_node_t *node, char *hosts);
+
 END_C_DECLS
 
 #endif
