Cleanup singleton detection and data retrieval

Extend the PMIx modex recv macros to cover the full set of
immediate/optional combinations. If PMIx_Init cannot reach a server,
then declare the MPI proc to be a singleton.

Provide full support for info values via PMIx

Catch all the values used in the "info" area of OMPI using data
available from PMIx instead of via envars. Update PMIx and PRRTE to sync
with their capabilities.

PMIx
- ensure cleanup of fork/exec children
- fix bug in gds/hash that left app info off of list

PRRTE
- fix multi-app bugs
- port setup_child logic from orte
- OMPI env changes
- set app->first_rank
- ensure common hostname across prun, prte, and pmix
- Fix "nolocal" support

Silence a warning from btl/vader

Signed-off-by: Ralph Castain <[email protected]>

Author: rhc54

ompi/attribute/attribute_predefined.c

@@ -13,6 +13,7 @@
  * Copyright (c) 2007      Cisco Systems, Inc.  All rights reserved.
  * Copyright (c) 2017      Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2020      Intel, Inc.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -87,6 +88,7 @@
 #include "ompi/errhandler/errcode.h"
 #include "ompi/communicator/communicator.h"
 #include "ompi/mca/pml/pml.h"
+#include "ompi/runtime/ompi_rte.h"
 
 /*
  * Private functions
@@ -103,8 +105,6 @@ static int set_f(int keyval, MPI_Fint value);
 int ompi_attr_create_predefined(void)
 {
     int ret;
-    char *univ_size;
-    int usize;
 
     /* Create all the keyvals */
 
@@ -138,14 +138,8 @@ int ompi_attr_create_predefined(void)
         return ret;
     }
 
-    /* If the universe size is set, then use it. Otherwise default
-     * to the size of MPI_COMM_WORLD */
-    univ_size = getenv("OMPI_UNIVERSE_SIZE");
-    if (NULL == univ_size || (usize = strtol(univ_size, NULL, 0)) <= 0) {
-        ret = set_f(MPI_UNIVERSE_SIZE, ompi_comm_size(MPI_COMM_WORLD));
-    } else {
-        ret = set_f(MPI_UNIVERSE_SIZE, usize);
-    }
+    /* set the universe size */
+    ret = set_f(MPI_UNIVERSE_SIZE, ompi_process_info.univ_size);
     if (OMPI_SUCCESS != ret) {
         return ret;
     }

ompi/info/info.c

@@ -19,6 +19,7 @@
  * Copyright (c) 2016-2017 IBM Corporation. All rights reserved.
  * Copyright (c) 2019      Triad National Security, LLC. All rights
  *                         reserved.
+ * Copyright (c) 2020      Intel, Inc.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -46,11 +47,13 @@
 #include "opal/util/argv.h"
 #include "opal/util/opal_getcwd.h"
 #include "opal/util/output.h"
+#include "opal/util/printf.h"
 #include "opal/util/info.h"
 
 #include "ompi/info/info.h"
 #include "ompi/runtime/mpiruntime.h"
 #include "ompi/runtime/params.h"
+#include "ompi/runtime/ompi_rte.h"
 
 /*
  * Global variables
@@ -85,8 +88,7 @@ opal_pointer_array_t ompi_info_f_to_c_table = {{0}};
  */
 int ompi_mpiinfo_init(void)
 {
-    const char *val;
-    char *cptr;
+    char *cptr, **tmp;
 
     /* initialize table */
 
@@ -107,45 +109,43 @@ int ompi_mpiinfo_init(void)
     /* fill the env info object */
 
     /* command for this app_context */
-    if (NULL != (cptr = getenv("OMPI_COMMAND"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "command", cptr);
-    }
-
-    /* space-separated list of argv for this command */
-    if (NULL != (cptr = getenv("OMPI_ARGV"))) {
+    if (NULL != ompi_process_info.command) {
+        tmp = opal_argv_split(ompi_process_info.command, ' ');
+        opal_info_set(&ompi_mpi_info_env.info.super, "command", tmp[0]);
+
+        /* space-separated list of argv for this command */
+        if (1 < opal_argv_count(tmp)) {
+            cptr = opal_argv_join(&tmp[1], ' ');
+        } else {
+            cptr = strdup(tmp[0]);
+        }
+        opal_argv_free(tmp);
         opal_info_set(&ompi_mpi_info_env.info.super, "argv", cptr);
+        free(cptr);
     }
 
     /* max procs for the entire job */
-    if (NULL != (cptr = getenv("OMPI_MCA_num_procs"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "maxprocs", cptr);
-        /* Open MPI does not support the "soft" option, so set it to maxprocs */
-        opal_info_set(&ompi_mpi_info_env.info.super, "soft", cptr);
-    }
+    opal_asprintf(&cptr, "%u", ompi_process_info.num_procs);
+    opal_info_set(&ompi_mpi_info_env.info.super, "maxprocs", cptr);
+    /* Open MPI does not support the "soft" option, so set it to maxprocs */
+    opal_info_set(&ompi_mpi_info_env.info.super, "soft", cptr);
+    free(cptr);
 
     /* local host name */
-    val = opal_gethostname();
-    opal_info_set(&ompi_mpi_info_env.info.super, "host", val);
+    opal_info_set(&ompi_mpi_info_env.info.super, "host", ompi_process_info.nodename);
 
-    /* architecture name */
-    if (NULL != (cptr = getenv("OMPI_MCA_cpu_type"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "arch", cptr);
-    }
 #ifdef HAVE_SYS_UTSNAME_H
-    else {
+    {
         struct utsname sysname;
         uname(&sysname);
         cptr = sysname.machine;
         opal_info_set(&ompi_mpi_info_env.info.super, "arch", cptr);
     }
 #endif
 
-    /* initial working dir of this process - only set when
-     * run by mpiexec as we otherwise have no reliable way
-     * of determining the value
-     */
-    if (NULL != (cptr = getenv("OMPI_MCA_initial_wdir"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "wdir", cptr);
+    /* initial working dir of this process, if provided */
+    if (NULL != ompi_process_info.initial_wdir) {
+        opal_info_set(&ompi_mpi_info_env.info.super, "wdir", ompi_process_info.initial_wdir);
     }
 
     /* provide the REQUESTED thread level - may be different
@@ -172,25 +172,25 @@ int ompi_mpiinfo_init(void)
     /**** now some OMPI-specific values that other MPIs may not provide ****/
 
     /* the number of app_contexts in this job */
-    if (NULL != (cptr = getenv("OMPI_NUM_APP_CTX"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "ompi_num_apps", cptr);
-    }
+    opal_asprintf(&cptr, "%u", ompi_process_info.num_apps);
+    opal_info_set(&ompi_mpi_info_env.info.super, "ompi_num_apps", cptr);
+    free(cptr);
 
     /* space-separated list of first MPI rank of each app_context */
-    if (NULL != (cptr = getenv("OMPI_FIRST_RANKS"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "ompi_first_rank", cptr);
+    if (NULL != ompi_process_info.app_ldrs) {
+        opal_info_set(&ompi_mpi_info_env.info.super, "ompi_first_rank", ompi_process_info.app_ldrs);
     }
 
     /* space-separated list of num procs for each app_context */
-    if (NULL != (cptr = getenv("OMPI_APP_CTX_NUM_PROCS"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "ompi_np", cptr);
+    if (NULL != ompi_process_info.app_sizes) {
+        opal_info_set(&ompi_mpi_info_env.info.super, "ompi_np", ompi_process_info.app_sizes);
     }
 
     /* location of the directory containing any prepositioned files
      * the user may have requested
      */
-    if (NULL != (cptr = getenv("OMPI_FILE_LOCATION"))) {
-        opal_info_set(&ompi_mpi_info_env.info.super, "ompi_positioned_file_dir", cptr);
+    if (NULL != ompi_process_info.proc_session_dir) {
+        opal_info_set(&ompi_mpi_info_env.info.super, "ompi_positioned_file_dir", ompi_process_info.proc_session_dir);
     }
 
     /* All done */
@@ -334,9 +334,9 @@ static void info_constructor(ompi_info_t *info)
                                                   info);
     info->i_freed = false;
 
-/* 
+/*
  * If the user doesn't want us to ever free it, then add an extra
- * RETAIN here 
+ * RETAIN here
  */
     if (ompi_debug_no_free_handles) {
         OBJ_RETAIN(&(info->super));

ompi/runtime/ompi_mpi_init.c

@@ -504,23 +504,8 @@ int ompi_mpi_init(int argc, char **argv, int requested, int *provided,
 
     OMPI_TIMING_NEXT("initialization");
 
-    /* if we were not externally started, then we need to setup
-     * some envars so the MPI_INFO_ENV can get the cmd name
-     * and argv (but only if the user supplied a non-NULL argv!), and
-     * the requested thread level
-     */
-    if (NULL == getenv("OMPI_COMMAND") && NULL != argv && NULL != argv[0]) {
-        opal_setenv("OMPI_COMMAND", argv[0], true, &environ);
-    }
-    if (NULL == getenv("OMPI_ARGV") && 1 < argc) {
-        char *tmp;
-        tmp = opal_argv_join(&argv[1], ' ');
-        opal_setenv("OMPI_ARGV", tmp, true, &environ);
-        free(tmp);
-    }
-
     /* Setup RTE */
-    if (OMPI_SUCCESS != (ret = ompi_rte_init(NULL, NULL))) {
+    if (OMPI_SUCCESS != (ret = ompi_rte_init(&argc, &argv))) {
         error = "ompi_mpi_init: ompi_rte_init failed";
         goto error;
     }