Attempt to reduce the RARP traffic during definition of allocations

opal/mca/pmix/native/pmix_native.c

@@ -101,7 +101,6 @@ const opal_pmix_base_module_t opal_pmix_native_module = {
 // local variables
 static int init_cntr = 0;
 opal_process_name_t native_pname;
-static char *local_uri = NULL;
 static uint32_t sm_flag;
 
 static void unpack_segment_info(opal_buffer_t *buf, opal_process_name_t *id, char** seg_info)
@@ -437,12 +436,6 @@ static int native_put(opal_pmix_scope_t scope,
         }
     }
 
-    /* if this is our uri, save it as we need to send it to our server
-     * as a special, separate item */
-    if (0 == strcmp(OPAL_DSTORE_URI, kv->key)) {
-        local_uri = strdup(kv->data.string);
-    }
-
     /* have to save a copy locally as some of our components will
      * look for it */
     (void)opal_dstore.store(opal_dstore_internal, &OPAL_PROC_MY_NAME, kv);
@@ -493,17 +486,6 @@ static int native_fence(opal_process_name_t *procs, size_t nprocs)
             return rc;
         }
     }
-    /* provide our URI */
-    if (OPAL_SUCCESS != (rc = opal_dss.pack(msg, &local_uri, 1, OPAL_STRING))) {
-        OPAL_ERROR_LOG(rc);
-        OBJ_RELEASE(msg);
-        return rc;
-    }
-    /* only do it once */
-    if (NULL != local_uri) {
-        free(local_uri);
-        local_uri = NULL;
-    }
 
     /* pack 1 if we have sm dstore enabled, 0 otherwise */
     if (OPAL_SUCCESS != (rc = opal_dss.pack(msg, &sm_flag, 1, OPAL_UINT32))) {
@@ -757,17 +739,6 @@ static int native_fence_nb(opal_process_name_t *procs, size_t nprocs,
             return rc;
         }
     }
-    /* provide our URI */
-    if (OPAL_SUCCESS != (rc = opal_dss.pack(msg, &local_uri, 1, OPAL_STRING))) {
-        OPAL_ERROR_LOG(rc);
-        OBJ_RELEASE(msg);
-        return rc;
-    }
-    /* only do it once */
-    if (NULL != local_uri) {
-        free(local_uri);
-        local_uri = NULL;
-    }
 
     /* pack 1 if we have sm dstore enabled, 0 otherwise */
     if (OPAL_SUCCESS != (rc = opal_dss.pack(msg, &sm_flag, 1, OPAL_UINT32))) {

orte/mca/dfs/app/dfs_app.c

@@ -1,7 +1,7 @@
 /*
  * Copyright (c) 2012-2013 Los Alamos National Security, LLC.
  *                         All rights reserved.
- * Copyright (c) 2014      Intel, Inc. All rights reserved
+ * Copyright (c) 2014-2015 Intel, Inc. All rights reserved
  * Copyright (c) 2014      Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
  * $COPYRIGHT$
@@ -33,6 +33,7 @@
 
 #include "orte/util/error_strings.h"
 #include "orte/util/name_fns.h"
+#include "orte/util/proc_info.h"
 #include "orte/util/show_help.h"
 #include "orte/runtime/orte_globals.h"
 #include "orte/mca/errmgr/errmgr.h"
@@ -537,9 +538,7 @@ static void process_opens(int fd, short args, void *cbdata)
     }
 
     /* if the host is our own, then treat it as a local file */
-    if (0 == strcmp(host, orte_process_info.nodename) ||
-        0 == strcmp(host, "localhost") ||
-        opal_ifislocal(host)) {
+    if (orte_ifislocal(host)) {
         opal_output_verbose(1, orte_dfs_base_framework.framework_output,
                             "%s file %s on local host",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),

orte/mca/dfs/orted/dfs_orted.c

@@ -1,7 +1,8 @@
 /*
  * Copyright (c) 2012-2013 Los Alamos National Security, LLC.
  *                         All rights reserved.
- * Copyright (c) 2013 Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2013      Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2015      Intel, Inc. All rights reserved
  * $COPYRIGHT$
  * 
  * Additional copyrights may follow
@@ -329,10 +330,7 @@ static void process_opens(int fd, short args, void *cbdata)
         goto complete;
     }
     /* if the host is our own, then treat it as a local file */
-    if (NULL == host ||
-        0 == strcmp(host, orte_process_info.nodename) ||
-        0 == strcmp(host, "localhost") ||
-        opal_ifislocal(host)) {
+    if (NULL == host || orte_ifislocal(host)) {
         opal_output_verbose(1, orte_dfs_base_framework.framework_output,
                             "%s file %s on local host",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),

orte/mca/ess/pmi/ess_pmi_module.c

@@ -232,28 +232,31 @@ static int rte_init(void)
 
     /***  PUSH DATA FOR OTHERS TO FIND   ***/
 
-    /* if our URI was not provided by the system, then
-     * push our URI so others can find us */
-    OBJ_CONSTRUCT(&vals, opal_list_t);
-    if (OPAL_SUCCESS != opal_dstore.fetch(opal_dstore_internal, &OPAL_PROC_MY_NAME,
-                                          OPAL_DSTORE_URI, &vals)) {
-        /* construct the RTE string */
-        rmluri = orte_rml.get_contact_info();
-        /* push it out for others to use */
-        OBJ_CONSTRUCT(&kvn, opal_value_t);
-        kvn.key = strdup(OPAL_DSTORE_URI);
-        kvn.type = OPAL_STRING;
-        kvn.data.string = strdup(rmluri);
-        if (ORTE_SUCCESS != (ret = opal_pmix.put(PMIX_GLOBAL, &kvn))) {
-            error = "db store uri";
+    /* if we are direct launched, then push our RML URI - there
+     * is no need to do so when launched by mpirun as all apps
+     * communicate thru their local daemon */
+    if (orte_standalone_operation) {
+        OBJ_CONSTRUCT(&vals, opal_list_t);
+        if (OPAL_SUCCESS != opal_dstore.fetch(opal_dstore_internal, &OPAL_PROC_MY_NAME,
+                                              OPAL_DSTORE_URI, &vals)) {
+            /* construct the RTE string */
+            rmluri = orte_rml.get_contact_info();
+            /* push it out for others to use */
+            OBJ_CONSTRUCT(&kvn, opal_value_t);
+            kvn.key = strdup(OPAL_DSTORE_URI);
+            kvn.type = OPAL_STRING;
+            kvn.data.string = strdup(rmluri);
+            if (ORTE_SUCCESS != (ret = opal_pmix.put(PMIX_GLOBAL, &kvn))) {
+                error = "db store uri";
+                OBJ_DESTRUCT(&kvn);
+                goto error;
+            }
             OBJ_DESTRUCT(&kvn);
-            goto error;
+            free(rmluri);
         }
-        OBJ_DESTRUCT(&kvn);
-        free(rmluri);
+        OPAL_LIST_DESTRUCT(&vals);
     }
-    OPAL_LIST_DESTRUCT(&vals);
-
+
     /* push our hostname so others can find us, if they need to */
     OBJ_CONSTRUCT(&kvn, opal_value_t);
     kvn.key = strdup(OPAL_DSTORE_HOSTNAME);

orte/mca/oob/tcp/oob_tcp.c

@@ -216,16 +216,14 @@ static void accept_connection(const int accepted_fd,
     ORTE_ACTIVATE_TCP_ACCEPT_STATE(accepted_fd, addr, recv_handler);
 }
 
+/* the host in this case is always in "dot" notation, and
+ * thus we do not need to do a DNS lookup to convert it */
 static int parse_uri(const uint16_t af_family,
                      const char* host,
                      const char *port,
                      struct sockaddr* inaddr)
 {
     struct sockaddr_in *in;
-#if OPAL_ENABLE_IPV6
-    struct addrinfo hints, *res;
-    int ret;
-#endif
 
     if (AF_INET == af_family) {
         memset(inaddr, 0, sizeof(struct sockaddr_in));
@@ -239,21 +237,14 @@ static int parse_uri(const uint16_t af_family,
     }
 #if OPAL_ENABLE_IPV6
     else if (AF_INET6 == af_family) {
-        size_t len;
+        struct sockaddr_in6 *in6;
         memset(inaddr, 0, sizeof(struct sockaddr_in6));
-        memset(&hints, 0, sizeof(hints));
-        hints.ai_family = af_family;
-        hints.ai_socktype = SOCK_STREAM;
-        ret = getaddrinfo(host, NULL, &hints, &res);
+        in6 = (struct sockaddr_in6*) inaddr;
 
-        if (ret) {
-            opal_output (0, "oob_tcp_parse_uri: Could not resolve %s. [Error: %s]\n",
-                         host, gai_strerror (ret));
+        if (0 == inet_pton(AF_INET6, host, (void*)&in6->sin6_addr)) {
+            opal_output (0, "oob_tcp_parse_uri: Could not convert %s\n", host);
             return ORTE_ERR_BAD_PARAM;
         }
-        len = (res->ai_addrlen < sizeof(struct sockaddr_in6)) ? res->ai_addrlen : sizeof(struct sockaddr_in6);
-        memcpy(inaddr, res->ai_addr, len);
-        freeaddrinfo(res);
     }
 #endif
     else {

orte/mca/oob/tcp/oob_tcp_component.c

@@ -451,6 +451,12 @@ static bool component_available(void)
     opal_output_verbose(5, orte_oob_base_framework.framework_output,
                         "oob:tcp: component_available called");
 
+    /* if we are an APP and we are not direct launched,
+     * then we don't want to be considered */
+    if (ORTE_PROC_IS_APP && !orte_standalone_operation) {
+        return false;
+    }
+
     /* if interface include was given, construct a list
      * of those interfaces which match the specifications - remember,
      * the includes could be given as named interfaces, IP addrs, or

orte/mca/ras/base/ras_base_allocate.c

@@ -11,7 +11,7 @@
  *                         All rights reserved.
  * Copyright (c) 2011-2012 Los Alamos National Security, LLC.  All rights
  *                         reserved. 
- * Copyright (c) 2014      Intel, Inc. All rights reserved.
+ * Copyright (c) 2014-2015 Intel, Inc. All rights reserved.
  * $COPYRIGHT$
  * 
  * Additional copyrights may follow
@@ -34,6 +34,7 @@
 #include "opal/util/output.h"
 #include "opal/dss/dss.h"
 #include "opal/util/argv.h"
+#include "opal/mca/if/if.h"
 
 #include "orte/util/show_help.h"
 #include "orte/mca/errmgr/errmgr.h"
@@ -195,6 +196,11 @@ void orte_ras_base_allocate(int fd, short args, void *cbdata)
     } 
     /* If something came back, save it and we are done */
     if (!opal_list_is_empty(&nodes)) {
+        /* flag that the allocation is managed */
+        orte_managed_allocation = true;
+        /* since it is managed, we do not attempt to resolve
+         * the nodenames */
+        opal_if_do_not_resolve = true;
         /* store the results in the global resource pool - this removes the
          * list items
          */
@@ -210,8 +216,6 @@ void orte_ras_base_allocate(int fd, short args, void *cbdata)
         if (!(ORTE_MAPPING_SUBSCRIBE_GIVEN & ORTE_GET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping))) {
             ORTE_SET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping, ORTE_MAPPING_NO_OVERSUBSCRIBE);
         }
-        /* flag that the allocation is managed */
-        orte_managed_allocation = true;
         goto DISPLAY;
     } else if (orte_allocation_required) {
         /* if nothing was found, and an allocation is

orte/mca/ras/base/ras_base_node.c

@@ -11,7 +11,7 @@
  *                         All rights reserved.
  * Copyright (c) 2011-2012 Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2014      Intel, Inc. All rights reserved.
+ * Copyright (c) 2014-2015 Intel, Inc. All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -103,10 +103,7 @@ int orte_ras_base_node_insert(opal_list_t* nodes, orte_job_t *jdata)
          * first position since it is the first one entered. We need to check to see
          * if this node is the same as the HNP's node so we don't double-enter it
          */
-        if (NULL != hnp_node &&
-            (0 == strcmp(node->name, hnp_node->name) ||
-             0 == strcmp(node->name, "localhost") ||
-             opal_ifislocal(node->name))) {
+        if (NULL != hnp_node && orte_ifislocal(node->name)) {
             OPAL_OUTPUT_VERBOSE((5, orte_ras_base_framework.framework_output,
                                  "%s ras:base:node_insert updating HNP [%s] info to %ld slots",
                                  ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),

orte/mca/rmaps/rank_file/rmaps_rank_file.c

@@ -14,7 +14,7 @@
  *                         All rights reserved.
  * Copyright (c) 2008      Voltaire. All rights reserved
  * Copyright (c) 2010      Oracle and/or its affiliates.  All rights reserved.
- * Copyright (c) 2014      Intel, Inc. All rights reserved.
+ * Copyright (c) 2014-2015 Intel, Inc. All rights reserved.
  *  
  * $COPYRIGHT$
  * 
@@ -484,8 +484,7 @@ static int orte_rmaps_rank_file_parse(const char *rankfile)
                             goto unlock;
                         }
                         /* check if this is the local node */
-                        if (0 == strcmp(node_name, hnp_node->name) ||
-                            opal_ifislocal(node_name)) {
+                        if (orte_ifislocal(node_name)) {
                             rfmap->node_name = strdup(hnp_node->name);
                         } else {
                             rfmap->node_name = strdup(node_name);

orte/mca/rmaps/seq/rmaps_seq.c

@@ -12,7 +12,7 @@
  * Copyright (c) 2006-2011 Cisco Systems, Inc.  All rights reserved.
  * Copyright (c) 2011      Los Alamos National Security, LLC.
  *                         All rights reserved.
- * Copyright (c) 2014      Intel, Inc. All rights reserved.
+ * Copyright (c) 2014-2015 Intel, Inc. All rights reserved.
  * Copyright (c) 2015      Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
  * $COPYRIGHT$
@@ -298,7 +298,7 @@ static int orte_rmaps_seq_map(orte_job_t *jdata)
                  * hostfile may not have been FQDN, while name returned
                  * by gethostname may have been (or vice versa)
                  */
-                if (opal_ifislocal(seq->hostname)) {
+                if (orte_ifislocal(seq->hostname)) {
                     opal_output_verbose(5, orte_rmaps_base_framework.framework_output,
                                         "mca:rmaps:seq: removing head node %s", seq->hostname);
                     opal_list_remove_item(seq_list, item);

orte/orted/pmix/pmix_server_process_msgs.c

@@ -100,7 +100,6 @@ void pmix_server_process_message(pmix_server_peer_t *peer)
     pmix_server_dmx_req_t *req, *nextreq;
     bool found;
     orte_grpcomm_signature_t *sig;
-    char *local_uri;
     uint32_t sm_flag;
 
     /* xfer the message to a buffer for unpacking */
@@ -207,19 +206,6 @@ void pmix_server_process_message(pmix_server_peer_t *peer)
                         (PMIX_FENCENB_CMD == cmd) ? "FENCE_NB" : "FENCE", tmp);
             free(tmp);
         }
-        /* get the URI for this process */
-        cnt = 1;
-        if (OPAL_SUCCESS != (rc = opal_dss.unpack(&xfer, &local_uri, &cnt, OPAL_STRING))) {
-            ORTE_ERROR_LOG(rc);
-            OBJ_RELEASE(sig);
-            goto reply_fence;
-        }
-        /* if not NULL, then update our connection info as we might need
-         * to send this proc a message at some point */
-        if (NULL != local_uri) {
-            orte_rml.set_contact_info(local_uri);
-            free(local_uri);
-        }
         /* unpack flag if sm dstore is supported by the client */
         cnt = 1;
         if (OPAL_SUCCESS != (rc = opal_dss.unpack(&xfer, &sm_flag, &cnt, OPAL_UINT32))) {