Example #1
0
static int
hwloc_solaris_get_sth_membind(hwloc_topology_t topology, idtype_t idtype, id_t id, hwloc_nodeset_t nodeset, hwloc_membind_policy_t *policy, int flags __hwloc_attribute_unused)
{
  int depth = hwloc_get_type_depth(topology, HWLOC_OBJ_NUMANODE);
  int n;
  int i;

  if (depth < 0) {
    errno = ENOSYS;
    return -1;
  }

  hwloc_bitmap_zero(nodeset);
  n = hwloc_get_nbobjs_by_depth(topology, depth);

  for (i = 0; i < n; i++) {
    hwloc_obj_t obj = hwloc_get_obj_by_depth(topology, depth, i);
    lgrp_affinity_t aff = lgrp_affinity_get(idtype, id, obj->os_index);

    if (aff == LGRP_AFF_STRONG)
      hwloc_bitmap_set(nodeset, obj->os_index);
  }

  if (hwloc_bitmap_iszero(nodeset))
    hwloc_bitmap_copy(nodeset, hwloc_topology_get_complete_nodeset(topology));

  *policy = HWLOC_MEMBIND_BIND;
  return 0;
}
Example #2
0
static int
hwloc_solaris_get_sth_cpubind(hwloc_topology_t topology, idtype_t idtype, id_t id, hwloc_bitmap_t hwloc_set, int flags __hwloc_attribute_unused)
{
  processorid_t binding;
  int depth = hwloc_get_type_depth(topology, HWLOC_OBJ_NUMANODE);
  int n;
  int i;

  if (depth < 0) {
    errno = ENOSYS;
    return -1;
  }

  /* first check if processor_bind() was used to bind to a single processor rather than to an lgroup */
  if ( processor_bind(idtype, id, PBIND_QUERY, &binding) == 0 && binding != PBIND_NONE ) {
    hwloc_bitmap_only(hwloc_set, binding);
    return 0;
  }

  /* if not, check lgroups */
  hwloc_bitmap_zero(hwloc_set);
  n = hwloc_get_nbobjs_by_depth(topology, depth);
  for (i = 0; i < n; i++) {
    hwloc_obj_t obj = hwloc_get_obj_by_depth(topology, depth, i);
    lgrp_affinity_t aff = lgrp_affinity_get(idtype, id, obj->os_index);

    if (aff == LGRP_AFF_STRONG)
      hwloc_bitmap_or(hwloc_set, hwloc_set, obj->cpuset);
  }

  if (hwloc_bitmap_iszero(hwloc_set))
    hwloc_bitmap_copy(hwloc_set, hwloc_topology_get_complete_cpuset(topology));

  return 0;
}
Example #3
0
static int dontset_return_complete_cpuset(hwloc_topology_t topology, hwloc_cpuset_t set)
{
  hwloc_const_cpuset_t cpuset = hwloc_topology_get_complete_cpuset(topology);
  if (cpuset) {
    hwloc_bitmap_copy(set, hwloc_topology_get_complete_cpuset(topology));
    return 0;
  } else
    return -1;
}
Example #4
0
static int dontset_return_complete_nodeset(hwloc_topology_t topology, hwloc_nodeset_t set, hwloc_membind_policy_t *policy)
{
  hwloc_const_nodeset_t nodeset = hwloc_topology_get_complete_nodeset(topology);
  if (nodeset) {
    hwloc_bitmap_copy(set, hwloc_topology_get_complete_nodeset(topology));
    *policy = HWLOC_MEMBIND_DEFAULT;
    return 0;
  } else
    return -1;
}
Example #5
0
static int
hwloc_fix_membind_cpuset(hwloc_topology_t topology, hwloc_nodeset_t nodeset, hwloc_const_cpuset_t cpuset)
{
  hwloc_const_bitmap_t topology_set = hwloc_topology_get_topology_cpuset(topology);
  hwloc_const_bitmap_t complete_set = hwloc_topology_get_complete_cpuset(topology);
  hwloc_const_bitmap_t complete_nodeset = hwloc_topology_get_complete_nodeset(topology);

  if (!topology_set) {
    /* The topology is composed of several systems, the cpuset is thus
     * ambiguous. */
    errno = EXDEV;
    return -1;
  }

  if (!complete_nodeset) {
    /* There is no NUMA node */
    errno = ENODEV;
    return -1;
  }

  if (hwloc_bitmap_iszero(cpuset)) {
    errno = EINVAL;
    return -1;
  }

  if (!hwloc_bitmap_isincluded(cpuset, complete_set)) {
    errno = EINVAL;
    return -1;
  }

  if (hwloc_bitmap_isincluded(topology_set, cpuset)) {
    hwloc_bitmap_copy(nodeset, complete_nodeset);
    return 0;
  }

  hwloc_cpuset_to_nodeset(topology, cpuset, nodeset);
  return 0;
}
Example #6
0
int orte_daemon(int argc, char *argv[])
{
    int ret = 0;
    opal_cmd_line_t *cmd_line = NULL;
    char *rml_uri;
    int i;
    opal_buffer_t *buffer;
    char hostname[100];
#if OPAL_ENABLE_FT_CR == 1
    char *tmp_env_var = NULL;
#endif
    
    /* initialize the globals */
    memset(&orted_globals, 0, sizeof(orted_globals));
    /* initialize the singleton died pipe to an illegal value so we can detect it was set */
    orted_globals.singleton_died_pipe = -1;
    /* init the failure orted vpid to an invalid value */
    orted_globals.fail = ORTE_VPID_INVALID;
    
    /* setup to check common command line options that just report and die */
    cmd_line = OBJ_NEW(opal_cmd_line_t);
    if (OPAL_SUCCESS != opal_cmd_line_create(cmd_line, orte_cmd_line_opts)) {
        OBJ_RELEASE(cmd_line);
        exit(1);
    }
    mca_base_cmd_line_setup(cmd_line);
    if (ORTE_SUCCESS != (ret = opal_cmd_line_parse(cmd_line, false,
                                                   argc, argv))) {
        char *args = NULL;
        args = opal_cmd_line_get_usage_msg(cmd_line);
        fprintf(stderr, "Usage: %s [OPTION]...\n%s\n", argv[0], args);
        free(args);
        OBJ_RELEASE(cmd_line);
        return ret;
    }
    
    /*
     * Since this process can now handle MCA/GMCA parameters, make sure to
     * process them.
     */
    mca_base_cmd_line_process_args(cmd_line, &environ, &environ);
    
    /* Ensure that enough of OPAL is setup for us to be able to run */
    /*
     * NOTE: (JJH)
     *  We need to allow 'mca_base_cmd_line_process_args()' to process command
     *  line arguments *before* calling opal_init_util() since the command
     *  line could contain MCA parameters that affect the way opal_init_util()
     *  functions. AMCA parameters are one such option normally received on the
     *  command line that affect the way opal_init_util() behaves.
     *  It is "safe" to call mca_base_cmd_line_process_args() before 
     *  opal_init_util() since mca_base_cmd_line_process_args() does *not*
     *  depend upon opal_init_util() functionality.
     */
    if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
        fprintf(stderr, "OPAL failed to initialize -- orted aborting\n");
        exit(1);
    }

    /* save the environment for launch purposes. This MUST be
     * done so that we can pass it to any local procs we
     * spawn - otherwise, those local procs won't see any
     * non-MCA envars that were set in the enviro when the
     * orted was executed - e.g., by .csh
     */
    orte_launch_environ = opal_argv_copy(environ);
    
    /* purge any ess flag set in the environ when we were launched */
    opal_unsetenv(OPAL_MCA_PREFIX"ess", &orte_launch_environ);
    
    /* if orte_daemon_debug is set, let someone know we are alive right
     * away just in case we have a problem along the way
     */
    if (orted_globals.debug) {
        gethostname(hostname, 100);
        fprintf(stderr, "Daemon was launched on %s - beginning to initialize\n", hostname);
    }
    
    /* check for help request */
    if (orted_globals.help) {
        char *args = NULL;
        args = opal_cmd_line_get_usage_msg(cmd_line);
        orte_show_help("help-orted.txt", "orted:usage", false,
                       argv[0], args);
        free(args);
        return 1;
    }
#if defined(HAVE_SETSID)
    /* see if we were directed to separate from current session */
    if (orted_globals.set_sid) {
        setsid();
    }
#endif
    /* see if they want us to spin until they can connect a debugger to us */
    i=0;
    while (orted_spin_flag) {
        i++;
        if (1000 < i) i=0;        
    }

#if OPAL_ENABLE_FT_CR == 1
    /* Mark as a tool program */
    (void) mca_base_var_env_name ("opal_cr_is_tool", &tmp_env_var);
    opal_setenv(tmp_env_var,
                "1",
                true, &environ);
    free(tmp_env_var);
#endif

    /* if mapreduce set, flag it */
    if (orted_globals.mapreduce) {
        orte_map_reduce = true;
    }

    /* detach from controlling terminal
     * otherwise, remain attached so output can get to us
     */
    if(!orte_debug_flag &&
       !orte_debug_daemons_flag &&
       orted_globals.daemonize) {
        opal_daemon_init(NULL);
    }
    
    /* Set the flag telling OpenRTE that I am NOT a
     * singleton, but am "infrastructure" - prevents setting
     * up incorrect infrastructure that only a singleton would
     * require.
     */
    if (orted_globals.hnp) {
        if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_HNP))) {
            ORTE_ERROR_LOG(ret);
            return ret;
        }
    } else {
        if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_DAEMON))) {
            ORTE_ERROR_LOG(ret);
            return ret;
        }
    }
    /* finalize the OPAL utils. As they are opened again from orte_init->opal_init
     * we continue to have a reference count on them. So we have to finalize them twice...
     */
    opal_finalize_util();

#if OPAL_HAVE_HWLOC
    /* bind ourselves if so directed */
    if (NULL != orte_daemon_cores) {
        char **cores=NULL, tmp[128];
        hwloc_obj_t pu;
        hwloc_cpuset_t ours, pucpus, res;
        int core;

        /* could be a collection of comma-delimited ranges, so
         * use our handy utility to parse it
         */
        orte_util_parse_range_options(orte_daemon_cores, &cores);
        if (NULL != cores) {
            ours = hwloc_bitmap_alloc();
            hwloc_bitmap_zero(ours);
            pucpus = hwloc_bitmap_alloc();
            res = hwloc_bitmap_alloc();
            for (i=0; NULL != cores[i]; i++) {
                core = strtoul(cores[i], NULL, 10);
                if (NULL == (pu = opal_hwloc_base_get_pu(opal_hwloc_topology, core, OPAL_HWLOC_LOGICAL))) {
                    /* turn off the show help forwarding as we won't
                     * be able to cycle the event library to send
                     */
                    orte_show_help_finalize();
                    /* the message will now come out locally */
                    orte_show_help("help-orted.txt", "orted:cannot-bind",
                                   true, orte_process_info.nodename,
                                   orte_daemon_cores);
                    ret = ORTE_ERR_NOT_SUPPORTED;
                    goto DONE;
                }
                hwloc_bitmap_and(pucpus, pu->online_cpuset, pu->allowed_cpuset);
                hwloc_bitmap_or(res, ours, pucpus);
                hwloc_bitmap_copy(ours, res);
            }
            /* if the result is all zeros, then don't bind */
            if (!hwloc_bitmap_iszero(ours)) {
                (void)hwloc_set_cpubind(opal_hwloc_topology, ours, 0);
                if (opal_hwloc_report_bindings) {
                    opal_hwloc_base_cset2mapstr(tmp, sizeof(tmp), opal_hwloc_topology, ours);
                    opal_output(0, "Daemon %s is bound to cores %s",
                                ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), tmp);
                }
            }
            /* cleanup */
            hwloc_bitmap_free(ours);
            hwloc_bitmap_free(pucpus);
            hwloc_bitmap_free(res);
            opal_argv_free(cores);
        }
    }
#endif

    if ((int)ORTE_VPID_INVALID != orted_globals.fail) {
        orted_globals.abort=false;
        /* some vpid was ordered to fail. The value can be positive
         * or negative, depending upon the desired method for failure,
         * so need to check both here
         */
        if (0 > orted_globals.fail) {
            orted_globals.fail = -1*orted_globals.fail;
            orted_globals.abort = true;
        }
        /* are we the specified vpid? */
        if ((int)ORTE_PROC_MY_NAME->vpid == orted_globals.fail) {
            /* if the user specified we delay, then setup a timer
             * and have it kill us
             */
            if (0 < orted_globals.fail_delay) {
                ORTE_TIMER_EVENT(orted_globals.fail_delay, 0, shutdown_callback, ORTE_SYS_PRI);
                
            } else {
                opal_output(0, "%s is executing clean %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                            orted_globals.abort ? "abort" : "abnormal termination");

                /* do -not- call finalize as this will send a message to the HNP
                 * indicating clean termination! Instead, just forcibly cleanup
                 * the local session_dir tree and exit
                 */
                orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);
                
                /* if we were ordered to abort, do so */
                if (orted_globals.abort) {
                    abort();
                }
                
                /* otherwise, return with non-zero status */
                ret = ORTE_ERROR_DEFAULT_EXIT_CODE;
                goto DONE;
            }
        }
    }

    /* insert our contact info into our process_info struct so we
     * have it for later use and set the local daemon field to our name
     */
    orte_process_info.my_daemon_uri = orte_rml.get_contact_info();
    ORTE_PROC_MY_DAEMON->jobid = ORTE_PROC_MY_NAME->jobid;
    ORTE_PROC_MY_DAEMON->vpid = ORTE_PROC_MY_NAME->vpid;
    
    /* if I am also the hnp, then update that contact info field too */
    if (ORTE_PROC_IS_HNP) {
        orte_process_info.my_hnp_uri = orte_rml.get_contact_info();
        ORTE_PROC_MY_HNP->jobid = ORTE_PROC_MY_NAME->jobid;
        ORTE_PROC_MY_HNP->vpid = ORTE_PROC_MY_NAME->vpid;
    }
    
    /* setup the primary daemon command receive function */
    orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DAEMON,
                            ORTE_RML_PERSISTENT, orte_daemon_recv, NULL);
    
    /* output a message indicating we are alive, our name, and our pid
     * for debugging purposes
     */
    if (orte_debug_daemons_flag) {
        fprintf(stderr, "Daemon %s checking in as pid %ld on host %s\n",
                ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), (long)orte_process_info.pid,
                orte_process_info.nodename);
    }

    /* We actually do *not* want the orted to voluntarily yield() the
       processor more than necessary.  The orted already blocks when
       it is doing nothing, so it doesn't use any more CPU cycles than
       it should; but when it *is* doing something, we do not want it
       to be unnecessarily delayed because it voluntarily yielded the
       processor in the middle of its work.

       For example: when a message arrives at the orted, we want the
       OS to wake up the orted in a timely fashion (which most OS's
       seem good about doing) and then we want the orted to process
       the message as fast as possible.  If the orted yields and lets
       aggressive MPI applications get the processor back, it may be a
       long time before the OS schedules the orted to run again
       (particularly if there is no IO event to wake it up).  Hence,
       routed OOB messages (for example) may be significantly delayed
       before being delivered to MPI processes, which can be
       problematic in some scenarios (e.g., COMM_SPAWN, BTL's that
       require OOB messages for wireup, etc.). */
    opal_progress_set_yield_when_idle(false);

    /* Change the default behavior of libevent such that we want to
       continually block rather than blocking for the default timeout
       and then looping around the progress engine again.  There
       should be nothing in the orted that cannot block in libevent
       until "something" happens (i.e., there's no need to keep
       cycling through progress because the only things that should
       happen will happen in libevent).  This is a minor optimization,
       but what the heck... :-) */
    opal_progress_set_event_flag(OPAL_EVLOOP_ONCE);

    /* if requested, report my uri to the indicated pipe */
    if (orted_globals.uri_pipe > 0) {
        orte_job_t *jdata;
        orte_proc_t *proc;
        orte_node_t *node;
        orte_app_context_t *app;
        char *tmp, *nptr, *sysinfo;
        int32_t ljob;

        /* setup the singleton's job */
        jdata = OBJ_NEW(orte_job_t);
        orte_plm_base_create_jobid(jdata);
        ljob = ORTE_LOCAL_JOBID(jdata->jobid);
        opal_pointer_array_set_item(orte_job_data, ljob, jdata);

        /* must create a map for it (even though it has no
         * info in it) so that the job info will be picked
         * up in subsequent pidmaps or other daemons won't
         * know how to route
         */
        jdata->map = OBJ_NEW(orte_job_map_t);

        /* setup an app_context for the singleton */
        app = OBJ_NEW(orte_app_context_t);
        app->app = strdup("singleton");
        app->num_procs = 1;
        opal_pointer_array_add(jdata->apps, app);
        
        /* setup a proc object for the singleton - since we
         * -must- be the HNP, and therefore we stored our
         * node on the global node pool, and since the singleton
         * -must- be on the same node as us, indicate that
         */
        proc = OBJ_NEW(orte_proc_t);
        proc->name.jobid = jdata->jobid;
        proc->name.vpid = 0;
        ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_ALIVE);
        proc->state = ORTE_PROC_STATE_RUNNING;
        proc->app_idx = 0;
        /* obviously, it is on my node */
        node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, 0);
        proc->node = node;
        OBJ_RETAIN(node);  /* keep accounting straight */
        opal_pointer_array_add(jdata->procs, proc);
        jdata->num_procs = 1;
        /* and it obviously is on the node */
        OBJ_RETAIN(proc);
        opal_pointer_array_add(node->procs, proc);
        node->num_procs++;
        /* and obviously it is one of my local procs */
        OBJ_RETAIN(proc);
        opal_pointer_array_add(orte_local_children, proc);
        jdata->num_local_procs = 1;
        /* set the trivial */
        proc->local_rank = 0;
        proc->node_rank = 0;
        proc->app_rank = 0;
        proc->state = ORTE_PROC_STATE_RUNNING;
        proc->app_idx = 0;
        ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_LOCAL);

        /* create a string that contains our uri + sysinfo + PMIx server URI */
        orte_util_convert_sysinfo_to_string(&sysinfo, orte_local_cpu_type, orte_local_cpu_model);
        asprintf(&tmp, "%s[%s]%s", orte_process_info.my_daemon_uri, sysinfo, pmix_server_uri);
	free(sysinfo);

        /* pass that info to the singleton */
        write(orted_globals.uri_pipe, tmp, strlen(tmp)+1); /* need to add 1 to get the NULL */

        /* cleanup */
        free(tmp);

        /* since a singleton spawned us, we need to harvest
         * any MCA params from the local environment so
         * we can pass them along to any subsequent daemons
         * we may start as the result of a comm_spawn
         */
        for (i=0; NULL != environ[i]; i++) {
            if (0 == strncmp(environ[i], OPAL_MCA_PREFIX, 9)) {
                /* make a copy to manipulate */
                tmp = strdup(environ[i]);
                /* find the equal sign */
                nptr = strchr(tmp, '=');
                *nptr = '\0';
                nptr++;
                /* add the mca param to the orted cmd line */
                opal_argv_append_nosize(&orted_cmd_line, "-"OPAL_MCA_CMD_LINE_ID);
                opal_argv_append_nosize(&orted_cmd_line, &tmp[9]);
                opal_argv_append_nosize(&orted_cmd_line, nptr);
                free(tmp);
            }
        }
    }

    /* if we were given a pipe to monitor for singleton termination, set that up */
    if (orted_globals.singleton_died_pipe > 0) {
        /* register shutdown handler */
        pipe_handler = (opal_event_t*)malloc(sizeof(opal_event_t));
        opal_event_set(orte_event_base, pipe_handler,
                       orted_globals.singleton_died_pipe,
                       OPAL_EV_READ,
                       pipe_closed,
                       pipe_handler);
        opal_event_add(pipe_handler, NULL);
    }

    /* If I have a parent, then save his contact info so
     * any messages we send can flow thru him.
     */

    orte_parent_uri = NULL;
    (void) mca_base_var_register ("orte", "orte", NULL, "parent_uri",
                                  "URI for the parent if tree launch is enabled.",
                                  MCA_BASE_VAR_TYPE_STRING, NULL, 0,
                                  MCA_BASE_VAR_FLAG_INTERNAL,
                                  OPAL_INFO_LVL_9,
                                  MCA_BASE_VAR_SCOPE_CONSTANT,
                                  &orte_parent_uri);
    if (NULL != orte_parent_uri) {
        orte_process_name_t parent;

        /* set the contact info into the hash table */
        orte_rml.set_contact_info(orte_parent_uri);
        ret = orte_rml_base_parse_uris(orte_parent_uri, &parent, NULL);
        if (ORTE_SUCCESS != ret) {
            ORTE_ERROR_LOG(ret);
            free (orte_parent_uri);
            orte_parent_uri = NULL;
            goto DONE;
        }

        /* don't need this value anymore */
        free(orte_parent_uri);
        orte_parent_uri = NULL;

        /* tell the routed module that we have a path
         * back to the HNP
         */
        if (ORTE_SUCCESS != (ret = orte_routed.update_route(ORTE_PROC_MY_HNP, &parent))) {
            ORTE_ERROR_LOG(ret);
            goto DONE;
        }
        /* set the lifeline to point to our parent so that we
         * can handle the situation if that lifeline goes away
         */
        if (ORTE_SUCCESS != (ret = orte_routed.set_lifeline(&parent))) {
            ORTE_ERROR_LOG(ret);
            goto DONE;
        }
    }

    /* if we are not the HNP...the only time we will be an HNP
     * is if we are launched by a singleton to provide support
     * for it
     */
    if (!ORTE_PROC_IS_HNP) {
        /* send the information to the orted report-back point - this function
         * will process the data, but also counts the number of
         * orteds that reported back so the launch procedure can continue.
         * We need to do this at the last possible second as the HNP
         * can turn right around and begin issuing orders to us
         */

        buffer = OBJ_NEW(opal_buffer_t);
        /* insert our name for rollup purposes */
        if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, ORTE_PROC_MY_NAME, 1, ORTE_NAME))) {
            ORTE_ERROR_LOG(ret);
            OBJ_RELEASE(buffer);
            goto DONE;
        }
        /* for now, always include our contact info, even if we are using
         * static ports. Eventually, this will be removed
         */
        rml_uri = orte_rml.get_contact_info();
        if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &rml_uri, 1, OPAL_STRING))) {
            ORTE_ERROR_LOG(ret);
            OBJ_RELEASE(buffer);
            goto DONE;
        }

        /* include our node name */
        opal_dss.pack(buffer, &orte_process_info.nodename, 1, OPAL_STRING);

        /* if requested, include any non-loopback aliases for this node */
        if (orte_retain_aliases) {
            char **aliases=NULL;
            uint8_t naliases, ni;
            char hostname[ORTE_MAX_HOSTNAME_SIZE];

            /* if we stripped the prefix or removed the fqdn,
             * include full hostname as an alias
             */
            gethostname(hostname, ORTE_MAX_HOSTNAME_SIZE);
            if (strlen(orte_process_info.nodename) < strlen(hostname)) {
                opal_argv_append_nosize(&aliases, hostname);
            }
            opal_ifgetaliases(&aliases);
            naliases = opal_argv_count(aliases);
            if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &naliases, 1, OPAL_UINT8))) {
                ORTE_ERROR_LOG(ret);
                OBJ_RELEASE(buffer);
                goto DONE;
            }
            for (ni=0; ni < naliases; ni++) {
                if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &aliases[ni], 1, OPAL_STRING))) {
                    ORTE_ERROR_LOG(ret);
                    OBJ_RELEASE(buffer);
                    goto DONE;
                }
            }
            opal_argv_free(aliases);
        }

#if OPAL_HAVE_HWLOC
        {
            char *coprocessors;
            /* add the local topology */
            if (NULL != opal_hwloc_topology &&
                (1 == ORTE_PROC_MY_NAME->vpid || orte_hetero_nodes)) {
                if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &opal_hwloc_topology, 1, OPAL_HWLOC_TOPO))) {
                    ORTE_ERROR_LOG(ret);
                }
            }
            /* detect and add any coprocessors */
            coprocessors = opal_hwloc_base_find_coprocessors(opal_hwloc_topology);
            if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &coprocessors, 1, OPAL_STRING))) {
                ORTE_ERROR_LOG(ret);
            }
            /* see if I am on a coprocessor */
            coprocessors = opal_hwloc_base_check_on_coprocessor();
            if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &coprocessors, 1, OPAL_STRING))) {
                ORTE_ERROR_LOG(ret);
            }
        }
#endif

        /* send to the HNP's callback - will be routed if routes are available */
        if (0 > (ret = orte_rml.send_buffer_nb(ORTE_PROC_MY_HNP, buffer,
                                               ORTE_RML_TAG_ORTED_CALLBACK,
                                               orte_rml_send_callback, NULL))) {
            ORTE_ERROR_LOG(ret);
            OBJ_RELEASE(buffer);
            goto DONE;
        }
    }

    /* if we are tree-spawning, then we need to capture the MCA params
     * from our cmd line so we can pass them along to the daemons we spawn -
     * otherwise, only the first layer of daemons will ever see them
     */
    if (orted_globals.tree_spawn) {
        int j, k;
        bool ignore;
        char *no_keep[] = {
            "orte_hnp_uri",
            "orte_ess_jobid",
            "orte_ess_vpid",
            "orte_ess_num_procs",
            "orte_parent_uri",
            "mca_base_env_list",
            NULL
        };
        for (i=0; i < argc; i++) {
            if (0 == strcmp("-"OPAL_MCA_CMD_LINE_ID,  argv[i]) ||
                0 == strcmp("--"OPAL_MCA_CMD_LINE_ID, argv[i]) ) {
                ignore = false;
                /* see if this is something we cannot pass along */
                for (k=0; NULL != no_keep[k]; k++) {
                    if (0 == strcmp(no_keep[k], argv[i+1])) {
                        ignore = true;
                        break;
                    }
                }
                if (!ignore) {
                    /* see if this is already present so we at least can
                     * avoid growing the cmd line with duplicates
                     */
                    if (NULL != orted_cmd_line) {
                        for (j=0; NULL != orted_cmd_line[j]; j++) {
                            if (0 == strcmp(argv[i+1], orted_cmd_line[j])) {
                                /* already here - ignore it */
                                ignore = true;
                                break;
                            }
                        }
                    }
                    if (!ignore) {
                        opal_argv_append_nosize(&orted_cmd_line, argv[i]);
                        opal_argv_append_nosize(&orted_cmd_line, argv[i+1]);
                        opal_argv_append_nosize(&orted_cmd_line, argv[i+2]);
                    }
                }
                i += 2;
            }
        }
    }
            
    if (orte_debug_daemons_flag) {
        opal_output(0, "%s orted: up and running - waiting for commands!", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
    }
    ret = ORTE_SUCCESS;

    /* loop the event lib until an exit event is detected */
    while (orte_event_base_active) {
        opal_event_loop(orte_event_base, OPAL_EVLOOP_ONCE);
    }

    /* ensure all local procs are dead */
    orte_odls.kill_local_procs(NULL);

 DONE:
    /* update the exit status, in case it wasn't done */
    ORTE_UPDATE_EXIT_STATUS(ret);

    /* cleanup and leave */
    orte_finalize();

    if (orte_debug_flag) {
        fprintf(stderr, "exiting with status %d\n", orte_exit_status);
    }
    exit(orte_exit_status);
}
void output_console(struct lstopo_output *loutput, const char *filename)
{
  hwloc_topology_t topology = loutput->topology;
  unsigned topodepth;
  int verbose_mode = loutput->verbose_mode;
  int logical = loutput->logical;
  FILE *output;

  output = open_output(filename, loutput->overwrite);
  if (!output) {
    fprintf(stderr, "Failed to open %s for writing (%s)\n", filename, strerror(errno));
    return;
  }

  topodepth = hwloc_topology_get_depth(topology);

  /*
   * if verbose_mode == 0, only print the summary.
   * if verbose_mode == 1, only print the topology tree.
   * if verbose_mode > 1, print both.
   */

  if (lstopo_show_only != (hwloc_obj_type_t)-1) {
    if (verbose_mode > 1)
      fprintf(output, "Only showing %s objects\n", hwloc_obj_type_string(lstopo_show_only));
    output_only (topology, hwloc_get_root_obj(topology), output, logical, verbose_mode);
  } else if (verbose_mode >= 1) {
    output_topology (topology, hwloc_get_root_obj(topology), NULL, output, 0, logical, verbose_mode);
    fprintf(output, "\n");
  }

  if ((verbose_mode > 1 || !verbose_mode) && lstopo_show_only == (hwloc_obj_type_t)-1) {
    hwloc_lstopo_show_summary(output, topology);
 }

  if (verbose_mode > 1 && lstopo_show_only == (hwloc_obj_type_t)-1) {
    const struct hwloc_distances_s * distances;
    unsigned depth;

    for (depth = 0; depth < topodepth; depth++) {
      distances = hwloc_get_whole_distance_matrix_by_depth(topology, depth);
      if (!distances || !distances->latency)
        continue;
      fprintf(output, "relative latency matrix between %ss (depth %u) by %s indexes:\n",
	      hwloc_obj_type_string(hwloc_get_depth_type(topology, depth)),
	      depth,
	      logical ? "logical" : "physical");
      hwloc_utils_print_distance_matrix(output, topology, hwloc_get_root_obj(topology), distances->nbobjs, depth, distances->latency, logical);
    }
  }

  if (verbose_mode > 1 && lstopo_show_only == (hwloc_obj_type_t)-1) {
    hwloc_const_bitmap_t complete = hwloc_topology_get_complete_cpuset(topology);
    hwloc_const_bitmap_t topo = hwloc_topology_get_topology_cpuset(topology);
    hwloc_const_bitmap_t allowed = hwloc_topology_get_allowed_cpuset(topology);

    if (!hwloc_bitmap_isequal(topo, complete)) {
      hwloc_bitmap_t unknown = hwloc_bitmap_alloc();
      char *unknownstr;
      hwloc_bitmap_copy(unknown, complete);
      hwloc_bitmap_andnot(unknown, unknown, topo);
      hwloc_bitmap_asprintf(&unknownstr, unknown);
      fprintf (output, "%d processors not represented in topology: %s\n", hwloc_bitmap_weight(unknown), unknownstr);
      free(unknownstr);
      hwloc_bitmap_free(unknown);
    }
    if (!hwloc_bitmap_isequal(topo, allowed)) {
      hwloc_bitmap_t disallowed = hwloc_bitmap_alloc();
      char *disallowedstr;
      hwloc_bitmap_copy(disallowed, topo);
      hwloc_bitmap_andnot(disallowed, disallowed, allowed);
      hwloc_bitmap_asprintf(&disallowedstr, disallowed);
      fprintf(output, "%d processors represented but not allowed: %s\n", hwloc_bitmap_weight(disallowed), disallowedstr);
      free(disallowedstr);
      hwloc_bitmap_free(disallowed);
    }
    if (!hwloc_topology_is_thissystem(topology))
      fprintf (output, "Topology not from this system\n");
  }

  if (output != stdout)
    fclose(output);
}
Example #8
0
int orte_daemon(int argc, char *argv[])
{
    int ret = 0;
    opal_cmd_line_t *cmd_line = NULL;
    int i;
    opal_buffer_t *buffer;
    char hostname[OPAL_MAXHOSTNAMELEN];
#if OPAL_ENABLE_FT_CR == 1
    char *tmp_env_var = NULL;
#endif

    /* initialize the globals */
    memset(&orted_globals, 0, sizeof(orted_globals));
    /* initialize the singleton died pipe to an illegal value so we can detect it was set */
    orted_globals.singleton_died_pipe = -1;
    bucket = OBJ_NEW(opal_buffer_t);

    /* setup to check common command line options that just report and die */
    cmd_line = OBJ_NEW(opal_cmd_line_t);
    if (OPAL_SUCCESS != opal_cmd_line_create(cmd_line, orte_cmd_line_opts)) {
        OBJ_RELEASE(cmd_line);
        exit(1);
    }
    mca_base_cmd_line_setup(cmd_line);
    if (ORTE_SUCCESS != (ret = opal_cmd_line_parse(cmd_line, false, false,
                                                   argc, argv))) {
        char *args = NULL;
        args = opal_cmd_line_get_usage_msg(cmd_line);
        fprintf(stderr, "Usage: %s [OPTION]...\n%s\n", argv[0], args);
        free(args);
        OBJ_RELEASE(cmd_line);
        return ret;
    }

    /*
     * Since this process can now handle MCA/GMCA parameters, make sure to
     * process them.
     */
    mca_base_cmd_line_process_args(cmd_line, &environ, &environ);

    /* Ensure that enough of OPAL is setup for us to be able to run */
    /*
     * NOTE: (JJH)
     *  We need to allow 'mca_base_cmd_line_process_args()' to process command
     *  line arguments *before* calling opal_init_util() since the command
     *  line could contain MCA parameters that affect the way opal_init_util()
     *  functions. AMCA parameters are one such option normally received on the
     *  command line that affect the way opal_init_util() behaves.
     *  It is "safe" to call mca_base_cmd_line_process_args() before
     *  opal_init_util() since mca_base_cmd_line_process_args() does *not*
     *  depend upon opal_init_util() functionality.
     */
    if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
        fprintf(stderr, "OPAL failed to initialize -- orted aborting\n");
        exit(1);
    }

    /* save the environment for launch purposes. This MUST be
     * done so that we can pass it to any local procs we
     * spawn - otherwise, those local procs won't see any
     * non-MCA envars that were set in the enviro when the
     * orted was executed - e.g., by .csh
     */
    orte_launch_environ = opal_argv_copy(environ);

    /* purge any ess/pmix flags set in the environ when we were launched */
    opal_unsetenv(OPAL_MCA_PREFIX"ess", &orte_launch_environ);
    opal_unsetenv(OPAL_MCA_PREFIX"pmix", &orte_launch_environ);

    /* if orte_daemon_debug is set, let someone know we are alive right
     * away just in case we have a problem along the way
     */
    if (orted_globals.debug) {
        gethostname(hostname, sizeof(hostname));
        fprintf(stderr, "Daemon was launched on %s - beginning to initialize\n", hostname);
    }

    /* check for help request */
    if (orted_globals.help) {
        char *args = NULL;
        args = opal_cmd_line_get_usage_msg(cmd_line);
        orte_show_help("help-orted.txt", "orted:usage", false,
                       argv[0], args);
        free(args);
        return 1;
    }
#if defined(HAVE_SETSID)
    /* see if we were directed to separate from current session */
    if (orted_globals.set_sid) {
        setsid();
    }
#endif
    /* see if they want us to spin until they can connect a debugger to us */
    i=0;
    while (orted_spin_flag) {
        i++;
        if (1000 < i) i=0;
    }

#if OPAL_ENABLE_FT_CR == 1
    /* Mark as a tool program */
    (void) mca_base_var_env_name ("opal_cr_is_tool", &tmp_env_var);
    opal_setenv(tmp_env_var,
                "1",
                true, &environ);
    free(tmp_env_var);
#endif

    /* detach from controlling terminal
     * otherwise, remain attached so output can get to us
     */
    if(!orte_debug_flag &&
       !orte_debug_daemons_flag &&
       orted_globals.daemonize) {
        opal_daemon_init(NULL);
    }

    /* Set the flag telling OpenRTE that I am NOT a
     * singleton, but am "infrastructure" - prevents setting
     * up incorrect infrastructure that only a singleton would
     * require.
     */
    if (orted_globals.hnp) {
        if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_HNP))) {
            ORTE_ERROR_LOG(ret);
            return ret;
        }
    } else {
        if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_DAEMON))) {
            ORTE_ERROR_LOG(ret);
            return ret;
        }
    }

    /* finalize the OPAL utils. As they are opened again from orte_init->opal_init
     * we continue to have a reference count on them. So we have to finalize them twice...
     */
    opal_finalize_util();

    /* bind ourselves if so directed */
    if (NULL != orte_daemon_cores) {
        char **cores=NULL, tmp[128];
        hwloc_obj_t pu;
        hwloc_cpuset_t ours, res;
        int core;

        /* could be a collection of comma-delimited ranges, so
         * use our handy utility to parse it
         */
        orte_util_parse_range_options(orte_daemon_cores, &cores);
        if (NULL != cores) {
            ours = hwloc_bitmap_alloc();
            hwloc_bitmap_zero(ours);
            res = hwloc_bitmap_alloc();
            for (i=0; NULL != cores[i]; i++) {
                core = strtoul(cores[i], NULL, 10);
                if (NULL == (pu = opal_hwloc_base_get_pu(opal_hwloc_topology, core, OPAL_HWLOC_LOGICAL))) {
                    /* turn off the show help forwarding as we won't
                     * be able to cycle the event library to send
                     */
                    orte_show_help_finalize();
                    /* the message will now come out locally */
                    orte_show_help("help-orted.txt", "orted:cannot-bind",
                                   true, orte_process_info.nodename,
                                   orte_daemon_cores);
                    ret = ORTE_ERR_NOT_SUPPORTED;
                    hwloc_bitmap_free(ours);
                    hwloc_bitmap_free(res);
                    goto DONE;
                }
                hwloc_bitmap_or(res, ours, pu->cpuset);
                hwloc_bitmap_copy(ours, res);
            }
            /* if the result is all zeros, then don't bind */
            if (!hwloc_bitmap_iszero(ours)) {
                (void)hwloc_set_cpubind(opal_hwloc_topology, ours, 0);
                if (opal_hwloc_report_bindings) {
                    opal_hwloc_base_cset2mapstr(tmp, sizeof(tmp), opal_hwloc_topology, ours);
                    opal_output(0, "Daemon %s is bound to cores %s",
                                ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), tmp);
                }
            }
            /* cleanup */
            hwloc_bitmap_free(ours);
            hwloc_bitmap_free(res);
            opal_argv_free(cores);
        }
    }

    if ((int)ORTE_VPID_INVALID != orted_debug_failure) {
        orted_globals.abort=false;
        /* some vpid was ordered to fail. The value can be positive
         * or negative, depending upon the desired method for failure,
         * so need to check both here
         */
        if (0 > orted_debug_failure) {
            orted_debug_failure = -1*orted_debug_failure;
            orted_globals.abort = true;
        }
        /* are we the specified vpid? */
        if ((int)ORTE_PROC_MY_NAME->vpid == orted_debug_failure) {
            /* if the user specified we delay, then setup a timer
             * and have it kill us
             */
            if (0 < orted_debug_failure_delay) {
                ORTE_TIMER_EVENT(orted_debug_failure_delay, 0, shutdown_callback, ORTE_SYS_PRI);

            } else {
                opal_output(0, "%s is executing clean %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                            orted_globals.abort ? "abort" : "abnormal termination");

                /* do -not- call finalize as this will send a message to the HNP
                 * indicating clean termination! Instead, just forcibly cleanup
                 * the local session_dir tree and exit
                 */
                orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);

                /* if we were ordered to abort, do so */
                if (orted_globals.abort) {
                    abort();
                }

                /* otherwise, return with non-zero status */
                ret = ORTE_ERROR_DEFAULT_EXIT_CODE;
                goto DONE;
            }
        }
    }

    /* insert our contact info into our process_info struct so we
     * have it for later use and set the local daemon field to our name
     */
    orte_oob_base_get_addr(&orte_process_info.my_daemon_uri);
    if (NULL == orte_process_info.my_daemon_uri) {
        /* no way to communicate */
        ret = ORTE_ERROR;
        goto DONE;
    }
    ORTE_PROC_MY_DAEMON->jobid = ORTE_PROC_MY_NAME->jobid;
    ORTE_PROC_MY_DAEMON->vpid = ORTE_PROC_MY_NAME->vpid;

    /* if I am also the hnp, then update that contact info field too */
    if (ORTE_PROC_IS_HNP) {
        orte_process_info.my_hnp_uri = strdup(orte_process_info.my_daemon_uri);
        ORTE_PROC_MY_HNP->jobid = ORTE_PROC_MY_NAME->jobid;
        ORTE_PROC_MY_HNP->vpid = ORTE_PROC_MY_NAME->vpid;
    }

    /* setup the primary daemon command receive function */
    orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DAEMON,
                            ORTE_RML_PERSISTENT, orte_daemon_recv, NULL);

    /* output a message indicating we are alive, our name, and our pid
     * for debugging purposes
     */
    if (orte_debug_daemons_flag) {
        fprintf(stderr, "Daemon %s checking in as pid %ld on host %s\n",
                ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), (long)orte_process_info.pid,
                orte_process_info.nodename);
    }

    /* We actually do *not* want the orted to voluntarily yield() the
       processor more than necessary.  The orted already blocks when
       it is doing nothing, so it doesn't use any more CPU cycles than
       it should; but when it *is* doing something, we do not want it
       to be unnecessarily delayed because it voluntarily yielded the
       processor in the middle of its work.

       For example: when a message arrives at the orted, we want the
       OS to wake up the orted in a timely fashion (which most OS's
       seem good about doing) and then we want the orted to process
       the message as fast as possible.  If the orted yields and lets
       aggressive MPI applications get the processor back, it may be a
       long time before the OS schedules the orted to run again
       (particularly if there is no IO event to wake it up).  Hence,
       routed OOB messages (for example) may be significantly delayed
       before being delivered to MPI processes, which can be
       problematic in some scenarios (e.g., COMM_SPAWN, BTL's that
       require OOB messages for wireup, etc.). */
    opal_progress_set_yield_when_idle(false);

    /* Change the default behavior of libevent such that we want to
       continually block rather than blocking for the default timeout
       and then looping around the progress engine again.  There
       should be nothing in the orted that cannot block in libevent
       until "something" happens (i.e., there's no need to keep
       cycling through progress because the only things that should
       happen will happen in libevent).  This is a minor optimization,
       but what the heck... :-) */
    opal_progress_set_event_flag(OPAL_EVLOOP_ONCE);

    /* if requested, report my uri to the indicated pipe */
    if (orted_globals.uri_pipe > 0) {
        orte_job_t *jdata;
        orte_proc_t *proc;
        orte_node_t *node;
        orte_app_context_t *app;
        char *tmp, *nptr, *sysinfo;
        char **singenv=NULL, *string_key, *env_str;

        /* setup the singleton's job */
        jdata = OBJ_NEW(orte_job_t);
        /* default to ompi for now */
        opal_argv_append_nosize(&jdata->personality, "ompi");
        orte_plm_base_create_jobid(jdata);
        opal_hash_table_set_value_uint32(orte_job_data, jdata->jobid, jdata);

        /* must create a map for it (even though it has no
         * info in it) so that the job info will be picked
         * up in subsequent pidmaps or other daemons won't
         * know how to route
         */
        jdata->map = OBJ_NEW(orte_job_map_t);

        /* setup an app_context for the singleton */
        app = OBJ_NEW(orte_app_context_t);
        app->app = strdup("singleton");
        app->num_procs = 1;
        opal_pointer_array_add(jdata->apps, app);
        jdata->num_apps = 1;

        /* setup a proc object for the singleton - since we
         * -must- be the HNP, and therefore we stored our
         * node on the global node pool, and since the singleton
         * -must- be on the same node as us, indicate that
         */
        proc = OBJ_NEW(orte_proc_t);
        proc->name.jobid = jdata->jobid;
        proc->name.vpid = 0;
        proc->parent = 0;
        ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_ALIVE);
        proc->state = ORTE_PROC_STATE_RUNNING;
        proc->app_idx = 0;
        /* obviously, it is on my node */
        node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, 0);
        proc->node = node;
        OBJ_RETAIN(node);  /* keep accounting straight */
        opal_pointer_array_add(jdata->procs, proc);
        jdata->num_procs = 1;
        /* add the node to the job map */
        OBJ_RETAIN(node);
        opal_pointer_array_add(jdata->map->nodes, node);
        jdata->map->num_nodes++;
        /* and it obviously is on the node */
        OBJ_RETAIN(proc);
        opal_pointer_array_add(node->procs, proc);
        node->num_procs++;
        /* and obviously it is one of my local procs */
        OBJ_RETAIN(proc);
        opal_pointer_array_add(orte_local_children, proc);
        jdata->num_local_procs = 1;
        /* set the trivial */
        proc->local_rank = 0;
        proc->node_rank = 0;
        proc->app_rank = 0;
        proc->state = ORTE_PROC_STATE_RUNNING;
        proc->app_idx = 0;
        ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_LOCAL);

        /* set the ORTE_JOB_TRANSPORT_KEY from the environment */
        orte_pre_condition_transports(jdata, NULL);

        /* register the singleton's nspace with our PMIx server */
        if (ORTE_SUCCESS != (ret = orte_pmix_server_register_nspace(jdata, false))) {
          ORTE_ERROR_LOG(ret);
          goto DONE;
        }
        /* use setup fork to create the envars needed by the singleton */
        if (OPAL_SUCCESS != (ret = opal_pmix.server_setup_fork(&proc->name, &singenv))) {
            ORTE_ERROR_LOG(ret);
            goto DONE;
        }

        /* append the transport key to the envars needed by the singleton */
        if (!orte_get_attribute(&jdata->attributes, ORTE_JOB_TRANSPORT_KEY, (void**)&string_key, OPAL_STRING) || NULL == string_key) {
            ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
            goto DONE;
        }
        asprintf(&env_str, OPAL_MCA_PREFIX"orte_precondition_transports=%s", string_key);
        opal_argv_append_nosize(&singenv, env_str);
        free(env_str);

        nptr = opal_argv_join(singenv, '*');
        opal_argv_free(singenv);

        /* create a string that contains our uri + sysinfo + PMIx server URI envars */
        orte_util_convert_sysinfo_to_string(&sysinfo, orte_local_cpu_type, orte_local_cpu_model);
        asprintf(&tmp, "%s[%s]%s", orte_process_info.my_daemon_uri, sysinfo, nptr);
        free(sysinfo);
        free(nptr);

        /* pass that info to the singleton */
        if (OPAL_SUCCESS != (ret = opal_fd_write(orted_globals.uri_pipe, strlen(tmp)+1, tmp))) { ; /* need to add 1 to get the NULL */
            ORTE_ERROR_LOG(ret);
            goto DONE;
        }

        /* cleanup */
        free(tmp);
        close(orted_globals.uri_pipe);

        /* since a singleton spawned us, we need to harvest
         * any MCA params from the local environment so
         * we can pass them along to any subsequent daemons
         * we may start as the result of a comm_spawn
         */
        for (i=0; NULL != environ[i]; i++) {
            if (0 == strncmp(environ[i], OPAL_MCA_PREFIX, 9)) {
                /* make a copy to manipulate */
                tmp = strdup(environ[i]);
                /* find the equal sign */
                nptr = strchr(tmp, '=');
                *nptr = '\0';
                nptr++;
                /* add the mca param to the orted cmd line */
                opal_argv_append_nosize(&orted_cmd_line, "-"OPAL_MCA_CMD_LINE_ID);
                opal_argv_append_nosize(&orted_cmd_line, &tmp[9]);
                opal_argv_append_nosize(&orted_cmd_line, nptr);
                free(tmp);
            }
        }
    }

    /* if we were given a pipe to monitor for singleton termination, set that up */
    if (orted_globals.singleton_died_pipe > 0) {
        /* register shutdown handler */
        pipe_handler = (opal_event_t*)malloc(sizeof(opal_event_t));
        opal_event_set(orte_event_base, pipe_handler,
                       orted_globals.singleton_died_pipe,
                       OPAL_EV_READ,
                       pipe_closed,
                       pipe_handler);
        opal_event_add(pipe_handler, NULL);
    }

    /* If I have a parent, then save his contact info so
     * any messages we send can flow thru him.
     */
    orte_parent_uri = NULL;
    (void) mca_base_var_register ("orte", "orte", NULL, "parent_uri",
                                  "URI for the parent if tree launch is enabled.",
                                  MCA_BASE_VAR_TYPE_STRING, NULL, 0,
                                  MCA_BASE_VAR_FLAG_INTERNAL,
                                  OPAL_INFO_LVL_9,
                                  MCA_BASE_VAR_SCOPE_CONSTANT,
                                  &orte_parent_uri);
    if (NULL != orte_parent_uri) {
        orte_process_name_t parent;
        opal_value_t val;

        /* set the contact info into our local database */
        ret = orte_rml_base_parse_uris(orte_parent_uri, &parent, NULL);
        if (ORTE_SUCCESS != ret) {
            ORTE_ERROR_LOG(ret);
            free (orte_parent_uri);
            orte_parent_uri = NULL;
            goto DONE;
        }
        OBJ_CONSTRUCT(&val, opal_value_t);
        val.key = OPAL_PMIX_PROC_URI;
        val.type = OPAL_STRING;
        val.data.string = orte_parent_uri;
        if (OPAL_SUCCESS != (ret = opal_pmix.store_local(&parent, &val))) {
            ORTE_ERROR_LOG(ret);
            OBJ_DESTRUCT(&val);
            goto DONE;
        }
        val.key = NULL;
        val.data.string = NULL;
        OBJ_DESTRUCT(&val);

        /* don't need this value anymore */
        free(orte_parent_uri);
        orte_parent_uri = NULL;

        /* tell the routed module that we have a path
         * back to the HNP
         */
        if (ORTE_SUCCESS != (ret = orte_routed.update_route(NULL, ORTE_PROC_MY_HNP, &parent))) {
            ORTE_ERROR_LOG(ret);
            goto DONE;
        }
        /* set the lifeline to point to our parent so that we
         * can handle the situation if that lifeline goes away
         */
        if (ORTE_SUCCESS != (ret = orte_routed.set_lifeline(NULL, &parent))) {
            ORTE_ERROR_LOG(ret);
            goto DONE;
        }
    }

    /* if we are not the HNP...the only time we will be an HNP
     * is if we are launched by a singleton to provide support
     * for it
     */
    if (!ORTE_PROC_IS_HNP) {
        orte_process_name_t target;
        target.jobid = ORTE_PROC_MY_NAME->jobid;

        if (orte_fwd_mpirun_port || orte_static_ports) {
            /* setup the rollup callback */
            orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_ORTED_CALLBACK,
                                    ORTE_RML_PERSISTENT, rollup, NULL);
            target.vpid = ORTE_PROC_MY_NAME->vpid;
            /* since we will be waiting for any children to send us
             * their rollup info before sending to our parent, save
             * a little time in the launch phase by "warming up" the
             * connection to our parent while we wait for our children */
            buffer = OBJ_NEW(opal_buffer_t);  // zero-byte message
            if (0 > (ret = orte_rml.send_buffer_nb(orte_mgmt_conduit,
                                                   ORTE_PROC_MY_PARENT, buffer,
                                                   ORTE_RML_TAG_WARMUP_CONNECTION,
                                                   orte_rml_send_callback, NULL))) {
                ORTE_ERROR_LOG(ret);
                OBJ_RELEASE(buffer);
                goto DONE;
            }
        } else {
            target.vpid = 0;
        }

        /* send the information to the orted report-back point - this function
         * will process the data, but also counts the number of
         * orteds that reported back so the launch procedure can continue.
         * We need to do this at the last possible second as the HNP
         * can turn right around and begin issuing orders to us
         */

        buffer = OBJ_NEW(opal_buffer_t);
        /* insert our name for rollup purposes */
        if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, ORTE_PROC_MY_NAME, 1, ORTE_NAME))) {
            ORTE_ERROR_LOG(ret);
            OBJ_RELEASE(buffer);
            goto DONE;
        }

        /* get any connection info we may have pushed */
        {
            opal_value_t *val = NULL, *kv;
            opal_list_t *modex;
            int32_t flag;

            if (OPAL_SUCCESS != (ret = opal_pmix.get(ORTE_PROC_MY_NAME, NULL, NULL, &val)) || NULL == val) {
                /* just pack a marker indicating we don't have any to share */
                flag = 0;
                if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &flag, 1, OPAL_INT32))) {
                    ORTE_ERROR_LOG(ret);
                    OBJ_RELEASE(buffer);
                    goto DONE;
                }
            } else {
                /* the data is returned as a list of key-value pairs in the opal_value_t */
                if (OPAL_PTR == val->type) {
                    modex = (opal_list_t*)val->data.ptr;
                    flag = (int32_t)opal_list_get_size(modex);
                    if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &flag, 1, OPAL_INT32))) {
                        ORTE_ERROR_LOG(ret);
                        OBJ_RELEASE(buffer);
                        goto DONE;
                    }
                    OPAL_LIST_FOREACH(kv, modex, opal_value_t) {
                        if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &kv, 1, OPAL_VALUE))) {
                            ORTE_ERROR_LOG(ret);
                            OBJ_RELEASE(buffer);
                            goto DONE;
                        }
                    }
                    OPAL_LIST_RELEASE(modex);
                } else {
                    opal_output(0, "VAL KEY: %s", (NULL == val->key) ? "NULL" : val->key);
                    /* single value */
                    flag = 1;
                    if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &flag, 1, OPAL_INT32))) {
                        ORTE_ERROR_LOG(ret);
                        OBJ_RELEASE(buffer);
                        goto DONE;
                    }
                    if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &val, 1, OPAL_VALUE))) {
                        ORTE_ERROR_LOG(ret);
                        OBJ_RELEASE(buffer);
                        goto DONE;
                    }
                }
                OBJ_RELEASE(val);
            }
Example #9
0
void output_console(hwloc_topology_t topology, const char *filename, int logical, int legend __hwloc_attribute_unused, int verbose_mode)
{
  unsigned topodepth;
  FILE *output;

  if (!filename || !strcmp(filename, "-"))
    output = stdout;
  else {
    output = open_file(filename, "w"); 
    if (!output) {
      fprintf(stderr, "Failed to open %s for writing (%s)\n", filename, strerror(errno));
      return;
    }
  }

  topodepth = hwloc_topology_get_depth(topology);

  /*
   * if verbose_mode == 0, only print the summary.
   * if verbose_mode == 1, only print the topology tree.
   * if verbose_mode > 1, print both.
   */

  if (lstopo_show_only != (hwloc_obj_type_t)-1) {
    if (verbose_mode > 1)
      fprintf(output, "Only showing %s objects\n", hwloc_obj_type_string(lstopo_show_only));
    output_only (topology, hwloc_get_root_obj(topology), output, logical, verbose_mode);
  } else if (verbose_mode >= 1) {
    output_topology (topology, hwloc_get_root_obj(topology), NULL, output, 0, logical, verbose_mode);
    fprintf(output, "\n");
  }

  if ((verbose_mode > 1 || !verbose_mode) && lstopo_show_only == (hwloc_obj_type_t)-1) {
    hwloc_lstopo_show_summary(output, topology);
 }

  if (verbose_mode > 1 && lstopo_show_only == (hwloc_obj_type_t)-1) {
    const struct hwloc_distances_s * distances;
    unsigned depth;

    for (depth = 0; depth < topodepth; depth++) {
      distances = hwloc_get_whole_distance_matrix_by_depth(topology, depth);
      if (!distances || !distances->latency)
        continue;
      printf("latency matrix between %ss (depth %u) by %s indexes:\n",
	     hwloc_obj_type_string(hwloc_get_depth_type(topology, depth)),
	     depth,
	     logical ? "logical" : "physical");
      hwloc_utils_print_distance_matrix(topology, hwloc_get_root_obj(topology), distances->nbobjs, depth, distances->latency, logical);
    }
  }

  if (verbose_mode > 1 && lstopo_show_only == (hwloc_obj_type_t)-1) {
    hwloc_const_bitmap_t complete = hwloc_topology_get_complete_cpuset(topology);
    hwloc_const_bitmap_t topo = hwloc_topology_get_topology_cpuset(topology);
    hwloc_const_bitmap_t online = hwloc_topology_get_online_cpuset(topology);
    hwloc_const_bitmap_t allowed = hwloc_topology_get_allowed_cpuset(topology);

    if (complete && !hwloc_bitmap_isequal(topo, complete)) {
      hwloc_bitmap_t unknown = hwloc_bitmap_alloc();
      char *unknownstr;
      hwloc_bitmap_copy(unknown, complete);
      hwloc_bitmap_andnot(unknown, unknown, topo);
      hwloc_bitmap_asprintf(&unknownstr, unknown);
      fprintf (output, "%d processors not represented in topology: %s\n", hwloc_bitmap_weight(unknown), unknownstr);
      free(unknownstr);
      hwloc_bitmap_free(unknown);
    }
    if (complete && !hwloc_bitmap_isequal(online, complete)) {
      hwloc_bitmap_t offline = hwloc_bitmap_alloc();
      char *offlinestr;
      hwloc_bitmap_copy(offline, complete);
      hwloc_bitmap_andnot(offline, offline, online);
      hwloc_bitmap_asprintf(&offlinestr, offline);
      fprintf (output, "%d processors offline: %s\n", hwloc_bitmap_weight(offline), offlinestr);
      free(offlinestr);
      hwloc_bitmap_free(offline);
    }
    if (complete && !hwloc_bitmap_isequal(allowed, online)) {
      if (!hwloc_bitmap_isincluded(online, allowed)) {
        hwloc_bitmap_t forbidden = hwloc_bitmap_alloc();
        char *forbiddenstr;
        hwloc_bitmap_copy(forbidden, online);
        hwloc_bitmap_andnot(forbidden, forbidden, allowed);
        hwloc_bitmap_asprintf(&forbiddenstr, forbidden);
        fprintf(output, "%d processors online but not allowed: %s\n", hwloc_bitmap_weight(forbidden), forbiddenstr);
        free(forbiddenstr);
        hwloc_bitmap_free(forbidden);
      }
      if (!hwloc_bitmap_isincluded(allowed, online)) {
        hwloc_bitmap_t potential = hwloc_bitmap_alloc();
        char *potentialstr;
        hwloc_bitmap_copy(potential, allowed);
        hwloc_bitmap_andnot(potential, potential, online);
        hwloc_bitmap_asprintf(&potentialstr, potential);
        fprintf(output, "%d processors allowed but not online: %s\n", hwloc_bitmap_weight(potential), potentialstr);
        free(potentialstr);
        hwloc_bitmap_free(potential);
      }
    }
    if (!hwloc_topology_is_thissystem(topology))
      fprintf (output, "Topology not from this system\n");
  }

  if (output != stdout)
    fclose(output);
}
Example #10
0
static int orcmd_init(void)
{
    int ret = ORTE_ERROR;
    char *error = NULL;
    opal_buffer_t buf, *clusterbuf, *uribuf;
    orte_job_t *jdata;
    orte_node_t *node;
    orte_proc_t *proc;
    opal_list_t config;
    orcm_scheduler_t *scheduler;
    orcm_node_t *mynode=NULL;
    int32_t n;

    if (initialized) {
        return ORCM_SUCCESS;
    }
    initialized = true;

    /* Initialize the ORTE data type support */
    if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
        error = "orte_std_prolog";
        goto error;
    }

    /* setup the global job and node arrays */
    orte_job_data = OBJ_NEW(opal_pointer_array_t);
    if (ORTE_SUCCESS != (ret = opal_pointer_array_init(orte_job_data,
                                                       1,
                                                       ORTE_GLOBAL_ARRAY_MAX_SIZE,
                                                       1))) {
        ORTE_ERROR_LOG(ret);
        error = "setup job array";
        goto error;
    }

    orte_node_pool = OBJ_NEW(opal_pointer_array_t);
    if (ORTE_SUCCESS != (ret = opal_pointer_array_init(orte_node_pool,
                                                       ORTE_GLOBAL_ARRAY_BLOCK_SIZE,
                                                       ORTE_GLOBAL_ARRAY_MAX_SIZE,
                                                       ORTE_GLOBAL_ARRAY_BLOCK_SIZE))) {
        ORTE_ERROR_LOG(ret);
        error = "setup node array";
        goto error;
    }
    orte_node_topologies = OBJ_NEW(opal_pointer_array_t);
    if (ORTE_SUCCESS != (ret = opal_pointer_array_init(orte_node_topologies,
                                                       ORTE_GLOBAL_ARRAY_BLOCK_SIZE,
                                                       ORTE_GLOBAL_ARRAY_MAX_SIZE,
                                                       ORTE_GLOBAL_ARRAY_BLOCK_SIZE))) {
        ORTE_ERROR_LOG(ret);
        error = "setup node topologies array";
        goto error;
    }

    /* create a job tracker for the daemons */
    jdata = OBJ_NEW(orte_job_t);
    jdata->jobid = 0;
    ORTE_PROC_MY_NAME->jobid = 0;
    opal_pointer_array_set_item(orte_job_data, 0, jdata);

    /* read the site configuration */
    OBJ_CONSTRUCT(&config, opal_list_t);
    if (ORCM_SUCCESS != (ret = orcm_cfgi.read_config(&config))) {
        error = "getting config";
        goto error;
    }

    /* define the cluster and collect contact info for all
     * aggregators - we'll need to know how to talk to any
     * of them in case of failures
     */
    OBJ_CONSTRUCT(&buf, opal_buffer_t);
    if (ORCM_SUCCESS != (ret = orcm_cfgi.define_system(&config,
                                                       &mynode,
                                                       &orte_process_info.num_procs,
                                                       &buf))) {
        OBJ_DESTRUCT(&buf);
        error = "define system";
        goto error;
    }

    /* if my name didn't get set, then we didn't find our node
     * in the config - report it and die
     */
    if (NULL == mynode) {
        orte_show_help("help-ess-orcm.txt", "node-not-found", true,
                       orcm_cfgi_base.config_file,
                       orte_process_info.nodename);
        OBJ_DESTRUCT(&buf);
        return ORTE_ERR_SILENT;
    }

    /* define a node and proc object for ourselves as some parts
     * of ORTE and ORCM require it */
    if (NULL == (node = OBJ_NEW(orte_node_t))) {
        ret = ORTE_ERR_OUT_OF_RESOURCE;
        error = "out of memory";
        goto error;
    }
    node->name = strdup(orte_process_info.nodename);
    opal_pointer_array_set_item(orte_node_pool, ORTE_PROC_MY_NAME->vpid, node);
    if (NULL == (proc = OBJ_NEW(orte_proc_t))) {
        ret = ORTE_ERR_OUT_OF_RESOURCE;
        error = "out of memory";
        goto error;
    }
    proc->name.jobid = ORTE_PROC_MY_NAME->jobid;
    proc->name.vpid = ORTE_PROC_MY_NAME->vpid;
    OBJ_RETAIN(proc);
    node->daemon = proc;
    OBJ_RETAIN(node);
    proc->node = node;
    opal_pointer_array_set_item(jdata->procs, ORTE_PROC_MY_NAME->vpid, proc);

    /* For now, we only support a single scheduler daemon in the system.
     * This *may* change someday in the future */
    scheduler = (orcm_scheduler_t*)opal_list_get_first(orcm_schedulers);

    /* If we are in test mode, then we don't *require* that a scheduler
     * be defined in the system - otherwise, we do */
    if (NULL == scheduler) {
        if (mca_sst_orcmd_component.scheduler_reqd) {
            error = "no scheduler found";
            ret = ORTE_ERR_NOT_FOUND;
            goto error;
        }
    } else {
        ORTE_PROC_MY_SCHEDULER->jobid = scheduler->controller.daemon.jobid;
        ORTE_PROC_MY_SCHEDULER->vpid = scheduler->controller.daemon.vpid;
    }

    /* register the ORTE-level params at this time now that the
     * config has had a chance to push things into the environ
     */
    if (ORTE_SUCCESS != (ret = orte_register_params())) {
        OBJ_DESTRUCT(&buf);
        error = "orte_register_params";
        goto error;
    }

    /* setup callback for SIGPIPE */
    setup_sighandler(SIGPIPE, &epipe_handler, epipe_signal_callback);
    /* Set signal handlers to catch kill signals so we can properly clean up
     * after ourselves.
     */
    setup_sighandler(SIGTERM, &term_handler, shutdown_signal);
    setup_sighandler(SIGINT, &int_handler, shutdown_signal);

    /** setup callbacks for signals we should ignore */
    setup_sighandler(SIGUSR1, &sigusr1_handler, signal_callback);
    setup_sighandler(SIGUSR2, &sigusr2_handler, signal_callback);
    signals_set = true;

#if OPAL_HAVE_HWLOC
    {
        hwloc_obj_t obj;
        unsigned i, j;

        /* get the local topology */
        if (NULL == opal_hwloc_topology) {
            if (OPAL_SUCCESS != opal_hwloc_base_get_topology()) {
                OBJ_DESTRUCT(&buf);
                error = "topology discovery";
                goto error;
            }
        }

        /* remove the hostname from the topology. Unfortunately, hwloc
         * decided to add the source hostname to the "topology", thus
         * rendering it unusable as a pure topological description. So
         * we remove that information here.
         */
        obj = hwloc_get_root_obj(opal_hwloc_topology);
        for (i=0; i < obj->infos_count; i++) {
            if (NULL == obj->infos[i].name ||
                NULL == obj->infos[i].value) {
                continue;
            }
            if (0 == strncmp(obj->infos[i].name, "HostName", strlen("HostName"))) {
                free(obj->infos[i].name);
                free(obj->infos[i].value);
                /* left justify the array */
                for (j=i; j < obj->infos_count-1; j++) {
                    obj->infos[j] = obj->infos[j+1];
                }
                obj->infos[obj->infos_count-1].name = NULL;
                obj->infos[obj->infos_count-1].value = NULL;
                obj->infos_count--;
                break;
            }
        }

        if (15 < opal_output_get_verbosity(orcm_sst_base_framework.framework_output)) {
            opal_output(0, "%s Topology Info:", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
            opal_dss.dump(0, opal_hwloc_topology, OPAL_HWLOC_TOPO);
        }

        /* if we were asked to bind to specific core(s), do so now */
        if (NULL != orte_daemon_cores) {
            char **cores=NULL, tmp[128];
            hwloc_obj_t pu;
            hwloc_cpuset_t ours, pucpus, res;
            int core;

            /* could be a collection of comma-delimited ranges, so
             * use our handy utility to parse it
             */
            orte_util_parse_range_options(orte_daemon_cores, &cores);
            if (NULL != cores) {
                ours = hwloc_bitmap_alloc();
                hwloc_bitmap_zero(ours);
                pucpus = hwloc_bitmap_alloc();
                res = hwloc_bitmap_alloc();
                for (i=0; NULL != cores[i]; i++) {
                    core = strtoul(cores[i], NULL, 10);
                    if (NULL == (pu = opal_hwloc_base_get_pu(opal_hwloc_topology, core, OPAL_HWLOC_LOGICAL))) {
                        orte_show_help("help-orted.txt", "orted:cannot-bind",
                                       true, orte_process_info.nodename,
                                       orte_daemon_cores);
                        ret = ORTE_ERR_NOT_SUPPORTED;
                        OBJ_DESTRUCT(&buf);
                        error = "cannot bind";
                        goto error;
                    }
                    hwloc_bitmap_and(pucpus, pu->online_cpuset, pu->allowed_cpuset);
                    hwloc_bitmap_or(res, ours, pucpus);
                    hwloc_bitmap_copy(ours, res);
                }
                /* if the result is all zeros, then don't bind */
                if (!hwloc_bitmap_iszero(ours)) {
                    (void)hwloc_set_cpubind(opal_hwloc_topology, ours, 0);
                    if (opal_hwloc_report_bindings) {
                        opal_hwloc_base_cset2mapstr(tmp, sizeof(tmp), opal_hwloc_topology, ours);
                        opal_output(0, "Daemon %s is bound to cores %s",
                                    ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), tmp);
                    }
                }
                /* cleanup */
                hwloc_bitmap_free(ours);
                hwloc_bitmap_free(pucpus);
                hwloc_bitmap_free(res);
                opal_argv_free(cores);
            }
        }
    }
#endif

    /* open and select the pstat framework */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&opal_pstat_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "opal_pstat_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = opal_pstat_base_select())) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "opal_pstat_base_select";
        goto error;
    }

    /* open and setup the state machine */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_state_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_state_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_state_base_select())) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_state_base_select";
        goto error;
    }

    /* open the notifier */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_notifier_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_notifier_base_open";
        goto error;
    }

    /* open the errmgr */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_errmgr_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_errmgr_base_open";
        goto error;
    }

    /* Setup the communication infrastructure */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_oob_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_oob_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_oob_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_oob_base_select";
        goto error;
    }
    if (!opal_list_get_size(&orte_oob_base.actives)) {
        ret = ORTE_ERROR;
        error = "orte_oob: Found 0 active transports";
        goto error;
    }

    /* Runtime Messaging Layer */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_rml_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_rml_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_rml_base_select())) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_rml_base_select";
        goto error;
    }

    /* select the notifier*/
    if (ORTE_SUCCESS != (ret = orte_notifier_base_select())) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_notifier_base_select";
        goto error;
    }

    /* select the errmgr */
    if (ORTE_SUCCESS != (ret = orte_errmgr_base_select())) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_errmgr_base_select";
        goto error;
    }

    /* Routed system */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_routed_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_rml_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_routed_base_select())) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orte_routed_base_select";
        goto error;
    }

    /* database */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orcm_db_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orcm_db_base_open";
        goto error;
    }
    /* always restrict daemons to local database components */
    if (ORTE_SUCCESS != (ret = orcm_db_base_select())) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "orcm_db_base_select";
        goto error;
    }

    /* datastore - ensure we don't pickup the pmi component, but
     * don't override anything set by user
     */
    if (NULL == getenv(OPAL_MCA_PREFIX"dstore")) {
        putenv(OPAL_MCA_PREFIX"dstore=^pmi");
    }
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&opal_dstore_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "opal_dstore_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = opal_dstore_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "opal_dstore_base_select";
        goto error;
    }
    /* create the handle */
    if (0 > (opal_dstore_internal = opal_dstore.open("INTERNAL", NULL, NULL))) {
        error = "opal dstore internal";
        ret = ORTE_ERR_FATAL;
        goto error;
    }

    /* extract the cluster description and setup the routed info - the orcm routed component
     * will know what to do. */
    n = 1;
    if (OPAL_SUCCESS != (ret = opal_dss.unpack(&buf, &clusterbuf, &n, OPAL_BUFFER))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "extract cluster buf";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_routed.init_routes(ORTE_PROC_MY_NAME->jobid, clusterbuf))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        OBJ_RELEASE(clusterbuf);
        error = "orte_routed.init_routes";
        goto error;
    }
    OBJ_RELEASE(clusterbuf);

    /* extract the uri buffer and load the hash tables */
    n = 1;
    if (OPAL_SUCCESS != (ret = opal_dss.unpack(&buf, &uribuf, &n, OPAL_BUFFER))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        error = "extract uri buffer";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_rml_base_update_contact_info(uribuf))) {
        ORTE_ERROR_LOG(ret);
        OBJ_DESTRUCT(&buf);
        OBJ_RELEASE(uribuf);
        error = "load hash tables";
        goto error;
    }
    OBJ_DESTRUCT(&buf);
    OBJ_RELEASE(uribuf);

    /*
     * Group communications
     */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_grpcomm_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orte_grpcomm_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_grpcomm_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_grpcomm_base_select";
        goto error;
    }

    /* Open/select the odls */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_odls_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orte_odls_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_odls_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_odls_base_select";
        goto error;
    }

    /* enable communication with the rml */
    if (ORTE_SUCCESS != (ret = orte_rml.enable_comm())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_rml.enable_comm";
        goto error;
    }

    /* setup the FileM */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_filem_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orte_filem_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_filem_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_filem_base_select";
        goto error;
    }

    /*
     * Initalize the CR setup
     * Note: Always do this, even in non-FT builds.
     * If we don't some user level tools may hang.
     */
    opal_cr_set_enabled(false);
    if (ORTE_SUCCESS != (ret = orte_cr_init())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_cr_init";
        goto error;
    }

    /* setup the ANALYTICS framework */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orcm_analytics_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_analytics_base_open";
        goto error;
    }

    /* setup the EVGEN framework */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orcm_evgen_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_evgen_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orcm_evgen_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_evgen_select";
        goto error;
    }

    /* setup the SENSOR framework */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orcm_sensor_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_sensor_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orcm_sensor_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_sensor_select";
        goto error;
    }
    /* start the local sensors */
    orcm_sensor.start(ORTE_PROC_MY_NAME->jobid);

    /* setup the PWRMGMT framework */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orcm_pwrmgmt_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_pwrmgmt_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orcm_pwrmgmt_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_pwrmgmt_select";
        goto error;
    }

    /* setup the DFS framework */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_dfs_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orte_dfs_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_dfs_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_dfs_select";
        goto error;
    }

    /* open and setup the DIAG framework */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orcm_diag_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_diag_base_open";
        goto error;
    }
    if (ORCM_SUCCESS != (ret = orcm_diag_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orcm_diag_select";
        goto error;
    }

    return ORTE_SUCCESS;
    
 error:
    orte_show_help("help-orcm-runtime.txt",
                   "orcm_init:startup:internal-failure",
                   true, error, ORTE_ERROR_NAME(ret), ret);
    
    return ORTE_ERR_SILENT;
}
int main(void)
{
  const struct hwloc_topology_support *support;
  char *buffer;
  hwloc_topology_t topology;
  hwloc_bitmap_t set = hwloc_bitmap_alloc();
  hwloc_bitmap_t total = hwloc_bitmap_alloc();
  hwloc_obj_t node;
  char *s;
  int err;

  err = hwloc_topology_init(&topology);
  assert(!err);
  err = hwloc_topology_load(topology);
  assert(!err);

  support = hwloc_topology_get_support(topology);
  if (!support->membind->get_area_memlocation)
    goto out;

  buffer = hwloc_alloc(topology, LEN);
  assert(buffer);
  printf("buffer %p length %u\n", buffer, LEN);

  err = hwloc_get_area_memlocation(topology, buffer, LEN, set, HWLOC_MEMBIND_BYNODESET);
  if (err < 0 && errno == ENOSYS) {
    fprintf(stderr, "hwloc_get_area_memlocation() failed with ENOSYS, aborting\n");
    goto out_with_buffer;
  }
  assert(!err);
  hwloc_bitmap_asprintf(&s, set);
  printf("address %p length %u allocated in nodeset %s\n", buffer, LEN, s);
  free(s);
  hwloc_bitmap_copy(total, set);

  node = NULL;
 next1:
  node = hwloc_get_next_obj_by_type(topology, HWLOC_OBJ_NUMANODE, node);
  if (!node)
    goto out_with_buffer;
  if (!node->memory.local_memory)
    goto next1;
  printf("binding to 1st node and touching 1st quarter\n");
  err = hwloc_set_area_membind(topology, buffer, LEN, node->nodeset, HWLOC_MEMBIND_BIND, HWLOC_MEMBIND_BYNODESET);
  if (err < 0 && errno == ENOSYS) {
    fprintf(stderr, "hwloc_set_area_membind() failed with ENOSYS, aborting\n");
    goto out_with_buffer;
  }
  assert(!err);

  memset(buffer, 0, LEN/4);
  err = hwloc_get_area_memlocation(topology, buffer, 1, set, HWLOC_MEMBIND_BYNODESET);
  assert(!err);
  hwloc_bitmap_asprintf(&s, set);
  printf("address %p length %u allocated in nodeset %s\n", buffer, LEN/4, s);
  free(s);
  hwloc_bitmap_or(total, total, set);

 next2:
  node = hwloc_get_next_obj_by_type(topology, HWLOC_OBJ_NUMANODE, node);
  if (!node)
    goto out_with_nomorenodes;
  if (!node->memory.local_memory)
    goto next2;
  printf("binding to 2nd node and touching 2nd quarter\n");
  err = hwloc_set_area_membind(topology, buffer, LEN, node->nodeset, HWLOC_MEMBIND_BIND, HWLOC_MEMBIND_BYNODESET);
  assert(!err);

  memset(buffer+LEN/4, 0, LEN/4);
  err = hwloc_get_area_memlocation(topology, buffer+LEN/4, LEN/4, set, HWLOC_MEMBIND_BYNODESET);
  assert(!err);
  hwloc_bitmap_asprintf(&s, set);
  printf("address %p length %u allocated in nodeset %s\n", buffer+LEN/4, LEN/4, s);
  free(s);
  hwloc_bitmap_or(total, total, set);

 next3:
  node = hwloc_get_next_obj_by_type(topology, HWLOC_OBJ_NUMANODE, node);
  if (!node)
    goto out_with_nomorenodes;
  if (!node->memory.local_memory)
    goto next3;
  printf("binding to 3rd node and touching 3rd quarter\n");
  err = hwloc_set_area_membind(topology, buffer, LEN, node->nodeset, HWLOC_MEMBIND_BIND, HWLOC_MEMBIND_BYNODESET);
  assert(!err);

  memset(buffer+LEN/2, 0, LEN/4);
  err = hwloc_get_area_memlocation(topology, buffer+LEN/2, LEN/4, set, HWLOC_MEMBIND_BYNODESET);
  assert(!err);
  hwloc_bitmap_asprintf(&s, set);
  printf("address %p length %u allocated in nodeset %s\n", buffer+LEN/2, LEN/4, s);
  free(s);
  hwloc_bitmap_or(total, total, set);

 next4:
  node = hwloc_get_next_obj_by_type(topology, HWLOC_OBJ_NUMANODE, node);
  if (!node)
    goto out_with_nomorenodes;
  if (!node->memory.local_memory)
    goto next4;
  printf("binding to 4th node and touching 4th quarter\n");
  err = hwloc_set_area_membind(topology, buffer, LEN, node->nodeset, HWLOC_MEMBIND_BIND, HWLOC_MEMBIND_BYNODESET);
  assert(!err);

  memset(buffer+3*LEN/4, 0, LEN/4);
  err = hwloc_get_area_memlocation(topology, buffer+3*LEN/4, LEN/4, set, HWLOC_MEMBIND_BYNODESET);
  assert(!err);
  hwloc_bitmap_asprintf(&s, set);
  printf("address %p length %u allocated in nodeset %s\n", buffer+3*LEN/4, LEN/4, s);
  free(s);
  hwloc_bitmap_or(total, total, set);

 out_with_nomorenodes:
  err = hwloc_get_area_memlocation(topology, buffer, LEN, set, HWLOC_MEMBIND_BYNODESET);
  assert(!err);
  hwloc_bitmap_asprintf(&s, set);
  printf("address %p length %u located on %s\n", buffer, LEN, s);
  free(s);
  assert(hwloc_bitmap_isincluded(total, set));

 out_with_buffer:
  hwloc_free(topology, buffer, LEN);

 out:
  hwloc_topology_destroy(topology);
  hwloc_bitmap_free(set);
  hwloc_bitmap_free(total);
  return 0;
}