ORTE-level MCA params are defined in several places. Ompi_info cannot call orte_init due to an issue with the memory allocator, thus making it impossible for ompi_info to display all of the ORTE-level MCA params.

By consolidating them all into one function, ompi_info can call that function and register the desired variables. This also requires, however, that ompi_info call orte_output_init to avoid generating tons of error messages, so make that adjustment too. 

Fixes ticket #1314

In addition, orte_output has a race condition issue whereby calls to orte_output/verbose can occur prior to either the RML being defined/setup, or the HNP being defined. This latter occurs during the initialization of the orte_process_info structure. In both cases, there is no way orte_output can send the output to the HNP. Hence, the message must be simply output locally.

Fixes ticket #1315

This commit was SVN r18524.

ompi/tools/ompi_info/components.cc

@@ -114,6 +114,7 @@
 #endif
 #include "orte/mca/filem/filem.h"
 #include "orte/mca/filem/base/base.h"
+#include "orte/util/output.h"
 
 using namespace std;
 using namespace ompi_info;
@@ -189,6 +190,9 @@ void ompi_info::open_components()
 
   orte_register_params();
 
+  // Initialize the orte_output system
+  orte_output_init();
+    
   // Register the MPI layer's MCA parameters
 
   ompi_mpi_register_params();

ompi/tools/ompi_info/ompi_info.cc

@@ -221,7 +221,6 @@ int main(int argc, char *argv[])
   ompi_info::mca_types.push_back("rml");
   ompi_info::mca_types.push_back("routed");
   ompi_info::mca_types.push_back("plm");
-  ompi_info::mca_types.push_back("sds");
 #if OPAL_ENABLE_FT == 1
   ompi_info::mca_types.push_back("snapc");
 #endif

orte/runtime/Makefile.am

@@ -47,4 +47,6 @@ libopen_rte_la_SOURCES += \
         runtime/orte_wakeup.c \
         runtime/orte_locks.c \
         runtime/orte_cr.c \
-        runtime/orte_data_server.c
+        runtime/orte_data_server.c \
+        runtime/orte_mca_params.c
+

orte/runtime/orte_globals.c

@@ -52,6 +52,10 @@ bool orte_do_not_launch = false;
 bool orted_spin_flag = false;
 bool orte_static_ports = false;
 bool orte_keep_fqdn_hostnames = false;
+bool orte_help_want_aggregate = true;
+bool orte_help_show_recursions;
+bool orte_params_set = false;
+int orte_debug_verbosity;
 
 int32_t orte_contiguous_nodes;
 int orte_debug_output = -1;
@@ -81,45 +85,13 @@ opal_pointer_array_t *orte_node_pool;
 bool orte_initialized = false;
 bool orte_finalizing = false;
 
-/* whether we have registered params or not */
-static bool params_set = false;
-
-int orte_register_params(void)
+int orte_dt_init(void)
 {
-    int value;
-    int orte_debug_verbosity;
-    
-    if (params_set) {
-        return ORTE_SUCCESS;
-    }
+    int rc;
+    opal_data_type_t tmp;
 
     /* set default output */
     orte_debug_output = orte_output_open(NULL, "ORTE", "DEBUG", NULL);
-    
-    mca_base_param_reg_int_name("orte", "debug",
-                                "Top-level ORTE debug switch (default verbosity: 1)",
-                                false, false, (int)false, &value);
-    orte_debug_flag = OPAL_INT_TO_BOOL(value);
-    
-    mca_base_param_reg_int_name("orte", "debug_verbose",
-                                "Verbosity level for ORTE debug messages (default: 1)",
-                                false, false, -1, &orte_debug_verbosity);
-    
-   mca_base_param_reg_int_name("orte", "debug_daemons",
-                                "Whether to debug the ORTE daemons or not",
-                                false, false, (int)false, &value);
-    orte_debug_daemons_flag = OPAL_INT_TO_BOOL(value);
-
-    mca_base_param_reg_int_name("orte", "debug_daemons_file",
-                                "Whether want stdout/stderr of daemons to go to a file or not",
-                                false, false, (int)false, &value);
-    orte_debug_daemons_file_flag = OPAL_INT_TO_BOOL(value);
-    /* If --debug-daemons-file was specified, that also implies
-        --debug-daemons */
-    if (orte_debug_daemons_file_flag) {
-        orte_debug_daemons_flag = true;
-    }
-    
     /* open up the verbose output for ORTE debugging */
     if (orte_debug_flag || 0 < orte_debug_verbosity ||
         (orte_debug_daemons_flag && (orte_process_info.daemon || orte_process_info.hnp))) {
@@ -130,71 +102,6 @@ int orte_register_params(void)
         }
     }
         
-    mca_base_param_reg_int_name("orte", "do_not_launch",
-                                "Perform all necessary operations to prepare to launch the application, but do not actually launch it",
-                                false, false, (int)false, &value);
-    orte_do_not_launch = OPAL_INT_TO_BOOL(value);
-    
-    mca_base_param_reg_int_name("orted", "spin",
-                                "Have any orteds spin until we can connect a debugger to them",
-                                false, false, (int)false, &value);
-    orted_spin_flag = OPAL_INT_TO_BOOL(value);
-
-    /* check for timing requests */
-    mca_base_param_reg_int_name("orte", "timing",
-                                "Request that critical timing loops be measured",
-                                false, false, (int)false, &value);
-    orte_timing = OPAL_INT_TO_BOOL(value);
-    
-    /* User-level debugger info string */
-
-    mca_base_param_reg_string_name("orte", "base_user_debugger",
-                                   "Sequence of user-level debuggers to search for in orterun",
-                                   false, false, "totalview @mpirun@ -a @mpirun_args@ : ddt -n @np@ -start @executable@ @executable_argv@ @single_app@ : fxp @mpirun@ -a @mpirun_args@", NULL);
-
-
-    mca_base_param_reg_int_name("orte", "abort_timeout",
-                                "Max time to wait [in secs] before aborting an ORTE operation (default: 1sec)",
-                                false, false, 1, &value);
-    orte_max_timeout = 1000000.0 * value;  /* convert to usec */
-
-    mca_base_param_reg_int_name("orte", "timeout_step",
-                                "Time to wait [in usecs/proc] before aborting an ORTE operation (default: 100 usec/proc)",
-                                false, false, 100, &orte_timeout_usec_per_proc);
-    
-    /* default hostfile */
-    mca_base_param_reg_string_name("orte", "default_hostfile",
-                                   "Name of the default hostfile (relative or absolute path)",
-                                   false, false, NULL, &orte_default_hostfile);
-    
-    
-    /* whether or not to keep FQDN hostnames */
-    mca_base_param_reg_int_name("orte", "keep_fqdn_hostnames",
-                                "Whether or not to keep FQDN hostnames [default: no]",
-                                false, false, (int)false, &value);
-    orte_keep_fqdn_hostnames = OPAL_INT_TO_BOOL(value);
-    
-    /* whether or not static ports exist */
-    mca_base_param_reg_int_name("orte", "static_ports",
-                                "Whether or not static ports are in use [default: no]",
-                                false, false, (int)false, &value);
-    orte_static_ports = OPAL_INT_TO_BOOL(value);
-
-    /* whether or not contiguous nodenames are in use */
-    mca_base_param_reg_int_name("orte", "contiguous_nodes",
-                                "Number of nodes after which contiguous nodenames will be used [default: INT_MAX]",
-                                false, false, INT32_MAX, &orte_contiguous_nodes);
-    
-    /* All done */
-    params_set = true;
-    return ORTE_SUCCESS;
-}
-
-int orte_dt_init(void)
-{
-    int rc;
-    opal_data_type_t tmp;
-    
     /** register the base system types with the DSS */
     tmp = ORTE_STD_CNTR;
     if (ORTE_SUCCESS != (rc = opal_dss.register_type(orte_dt_pack_std_cntr,

orte/runtime/orte_globals.h

@@ -319,6 +319,10 @@ ORTE_DECLSPEC extern bool orte_static_ports;
 ORTE_DECLSPEC extern int32_t orte_contiguous_nodes;
 ORTE_DECLSPEC extern int orte_debug_output;
 ORTE_DECLSPEC extern bool orte_keep_fqdn_hostnames;
+ORTE_DECLSPEC extern bool orte_help_want_aggregate;
+ORTE_DECLSPEC extern bool orte_help_show_recursions;
+ORTE_DECLSPEC extern bool orte_params_set;
+ORTE_DECLSPEC extern int orte_debug_verbosity;
 
 ORTE_DECLSPEC extern char **orte_launch_environ;
 ORTE_DECLSPEC extern opal_pointer_array_t orte_daemonmap;

orte/runtime/orte_init.c

@@ -69,6 +69,15 @@ int orte_init(char flags)
         return ret;
     }
     
+    /* ensure we know the tool setting for when we finalize */
+    if ((flags & ORTE_TOOL) || (flags & ORTE_TOOL_WITH_NAME)) {
+        orte_process_info.tool = true;
+    }
+    
+    if (orte_process_info.hnp) {
+        orte_process_info.daemon = false;
+    }
+    
     /* setup the orte_output system */
     if (ORTE_SUCCESS != (ret = orte_output_init())) {
         ORTE_ERROR_LOG(ret);
@@ -91,29 +100,23 @@ int orte_init(char flags)
         goto error;
     }
 
-   /* Ensure the process info structure is instantiated and initialized */
-    if (ORTE_SUCCESS != (ret = orte_proc_info())) {
-        error = "orte_proc_info";
-        goto error;
-    }
-
-    /* ensure we know the tool setting for when we finalize */
-    if ((flags & ORTE_TOOL) || (flags & ORTE_TOOL_WITH_NAME)) {
-        orte_process_info.tool = true;
-    }
-    
     /* Initialize the ORTE data type support */
     if (ORTE_SUCCESS != (ret = orte_dt_init())) {
         error = "orte_dt_init";
         goto error;
     }
     
+    /* Ensure the rest of the process info structure is initialized */
+    if (ORTE_SUCCESS != (ret = orte_proc_info())) {
+        error = "orte_proc_info";
+        goto error;
+    }
+
     /* if I'm the HNP, make sure that the daemon flag is NOT set so that
      * components unique to non-HNP orteds can be selected and init
      * my basic storage elements
      */
     if (orte_process_info.hnp) {
-        orte_process_info.daemon = false;
         if (ORTE_SUCCESS != (ret = orte_hnp_globals_init())) {
             error = "orte_hnp_globals_init";
             goto error;

orte/runtime/orte_mca_params.c

@@ -0,0 +1,139 @@
+/*
+ * Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
+ *                         University Research and Technology
+ *                         Corporation.  All rights reserved.
+ * Copyright (c) 2004-2008 The University of Tennessee and The University
+ *                         of Tennessee Research Foundation.  All rights
+ *                         reserved.
+ * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart, 
+ *                         University of Stuttgart.  All rights reserved.
+ * Copyright (c) 2004-2005 The Regents of the University of California.
+ *                         All rights reserved.
+ * Copyright (c) 2007      Cisco, Inc.  All rights reserved.
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "orte_config.h"
+#include "orte/constants.h"
+#include "orte/types.h"
+
+#ifdef HAVE_SYS_TIME_H
+#include <sys/time.h>
+#endif
+
+#include "opal/mca/base/mca_base_param.h"
+
+#include "orte/util/proc_info.h"
+#include "orte/util/output.h"
+
+#include "orte/runtime/runtime.h"
+#include "orte/runtime/orte_globals.h"
+
+int orte_register_params(void)
+{
+    int value;
+    
+    if (orte_params_set) {
+        return ORTE_SUCCESS;
+    }
+    
+    mca_base_param_reg_int_name("orte", "debug",
+                                "Top-level ORTE debug switch (default verbosity: 1)",
+                                false, false, (int)false, &value);
+    orte_debug_flag = OPAL_INT_TO_BOOL(value);
+    
+    mca_base_param_reg_int_name("orte", "debug_verbose",
+                                "Verbosity level for ORTE debug messages (default: 1)",
+                                false, false, -1, &orte_debug_verbosity);
+    
+   mca_base_param_reg_int_name("orte", "debug_daemons",
+                                "Whether to debug the ORTE daemons or not",
+                                false, false, (int)false, &value);
+    orte_debug_daemons_flag = OPAL_INT_TO_BOOL(value);
+
+    mca_base_param_reg_int_name("orte", "debug_daemons_file",
+                                "Whether want stdout/stderr of daemons to go to a file or not",
+                                false, false, (int)false, &value);
+    orte_debug_daemons_file_flag = OPAL_INT_TO_BOOL(value);
+    /* If --debug-daemons-file was specified, that also implies
+        --debug-daemons */
+    if (orte_debug_daemons_file_flag) {
+        orte_debug_daemons_flag = true;
+    }
+    
+    mca_base_param_reg_int_name("orte", "do_not_launch",
+                                "Perform all necessary operations to prepare to launch the application, but do not actually launch it",
+                                false, false, (int)false, &value);
+    orte_do_not_launch = OPAL_INT_TO_BOOL(value);
+    
+    mca_base_param_reg_int_name("orte", "daemon_spin",
+                                "Have any orteds spin until we can connect a debugger to them",
+                                false, false, (int)false, &value);
+    orted_spin_flag = OPAL_INT_TO_BOOL(value);
+
+    /* check for timing requests */
+    mca_base_param_reg_int_name("orte", "timing",
+                                "Request that critical timing loops be measured",
+                                false, false, (int)false, &value);
+    orte_timing = OPAL_INT_TO_BOOL(value);
+    
+    /* User-level debugger info string */
+
+    mca_base_param_reg_string_name("orte", "base_user_debugger",
+                                   "Sequence of user-level debuggers to search for in orterun",
+                                   false, false, "totalview @mpirun@ -a @mpirun_args@ : ddt -n @np@ -start @executable@ @executable_argv@ @single_app@ : fxp @mpirun@ -a @mpirun_args@", NULL);
+
+
+    mca_base_param_reg_int_name("orte", "abort_timeout",
+                                "Max time to wait [in secs] before aborting an ORTE operation (default: 1sec)",
+                                false, false, 1, &value);
+    orte_max_timeout = 1000000.0 * value;  /* convert to usec */
+
+    mca_base_param_reg_int_name("orte", "timeout_step",
+                                "Time to wait [in usecs/proc] before aborting an ORTE operation (default: 100 usec/proc)",
+                                false, false, 100, &orte_timeout_usec_per_proc);
+    
+    /* default hostfile */
+    mca_base_param_reg_string_name("orte", "default_hostfile",
+                                   "Name of the default hostfile (relative or absolute path)",
+                                   false, false, NULL, &orte_default_hostfile);
+    
+    
+    /* whether or not to keep FQDN hostnames */
+    mca_base_param_reg_int_name("orte", "keep_fqdn_hostnames",
+                                "Whether or not to keep FQDN hostnames [default: no]",
+                                false, false, (int)false, &value);
+    orte_keep_fqdn_hostnames = OPAL_INT_TO_BOOL(value);
+    
+    /* whether or not contiguous nodenames are in use */
+    mca_base_param_reg_int_name("orte", "contiguous_nodes",
+                                "Number of nodes after which contiguous nodename encoding will automatically be used [default: INT_MAX]",
+                                false, false, INT32_MAX, &orte_contiguous_nodes);
+    
+    mca_base_param_reg_int_name("orte", "base_help_aggregate",
+                                "If orte_base_help_aggregate is true, duplicate help messages will be aggregated rather than displayed individually.  This can be helpful for parallel jobs that experience multiple identical failures; rather than print out the same help/failure message N times, display it once with a count of how many processes sent the same message.",
+                                false, false,
+                                (int) orte_help_want_aggregate, &value);
+    orte_help_want_aggregate = OPAL_INT_TO_BOOL(value);
+    
+    mca_base_param_reg_int_name("orte", "base_show_output_recursions",
+                                "If orte_base_show_output_recursion is true, recursive calls to orte_output will be reported to stderr",
+                                false, false,
+                                (int) false, &value);
+    orte_help_show_recursions = OPAL_INT_TO_BOOL(value);
+    
+    /* some params that are accessed elsewhere, but simply registered here so they will
+     * be visible to ompi_info
+     */
+    mca_base_param_reg_string_name("orte", "tmpdir_base",
+                                   "Base of the session directory tree",
+                                   false, false, NULL,  &(orte_process_info.tmpdir_base));
+    
+    /* All done */
+    orte_params_set = true;
+    return ORTE_SUCCESS;
+}

orte/util/output.c

@@ -80,31 +80,6 @@ int orte_output_get_verbosity(int output_id)
     return opal_output_get_verbosity(output_id);
 }
 
-/* Whether we aggregate show_help() messages or not */
-static bool want_aggregate = true;
-
-/* Whether to report recursions or not */
-static bool show_recursions;
-
-static void register_mca(void)
-{
-    int tmp;
-
-    mca_base_param_reg_int_name("orte", "base_help_aggregate",
-                                "If orte_base_help_aggregate is true, duplicate help messages will be aggregated rather than displayed individually.  This can be helpful for parallel jobs that experience multiple identical failures; rather than print out the same help/failure message N times, display it once with a count of how many processes sent the same message.",
-                                false, false,
-                                (int) want_aggregate, &tmp);
-    want_aggregate = OPAL_INT_TO_BOOL(tmp);
-
-    mca_base_param_reg_int_name("orte", "base_show_output_recursions",
-                                "If orte_base_show_output_recursion is true, recursive calls to orte_output will be reported to stderr",
-                                false, false,
-                                (int) false, &tmp);
-    show_recursions = OPAL_INT_TO_BOOL(tmp);
-    
-}
-
-
 
 /************************************************************************/
 
@@ -118,7 +93,7 @@ int orte_output_init(void)
 {
     stderr_stream = opal_output_open(NULL);
     regiester_mca();
-    if (0 == ORTE_PROC_MY_NAME->vpid && want_aggregate) {
+    if (0 == ORTE_PROC_MY_NAME->vpid && orte_help_want_aggregate) {
         orte_output(stderr_stream, "WARNING: orte_base_help_aggregate was set to true, but this system does not support help message aggregation");
     }
     return ORTE_SUCCESS;
@@ -429,7 +404,7 @@ static void output_vverbose(int verbose_level, int output_id,
        HNP).
     */
     if (am_inside) {
-        if (show_recursions) {
+        if (orte_help_want_aggregate) {
             opal_output(0, "%s orte_output recursion detected!", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
         }
         opal_output(output_id, filtered);
@@ -474,9 +449,12 @@ static void output_vverbose(int verbose_level, int output_id,
                              ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                              filtered, output_id));
 
-        /* If RML is not yet setup, then just output this locally.
-           What else can we do? */
-        if (NULL == orte_rml.send_buffer) {
+        /* If RML is not yet setup, or we haven't yet defined the HNP,
+         * then just output this locally.
+         * What else can we do?
+         */
+        if (NULL == orte_rml.send_buffer ||
+            ORTE_PROC_MY_HNP->vpid == ORTE_VPID_INVALID) {
             opal_output(0, filtered);
         } else {
             /* setup a buffer to send to the HNP */
@@ -541,7 +519,7 @@ static int show_help(const char *filename, const char *topic,
 
     /* If we're aggregating, check for duplicates.  Otherwise, don't
        track duplicates at all and always display the message. */
-    if (orte_output_ready && want_aggregate) {
+    if (orte_output_ready && orte_help_want_aggregate) {
         rc = get_tli(filename, topic, &tli);
     } else {
         rc = ORTE_ERR_NOT_FOUND;
@@ -598,7 +576,7 @@ static int show_help(const char *filename, const char *topic,
     }
 
     /* If we're aggregating, add this process name to the list */
-    if (orte_output_ready && want_aggregate) {
+    if (orte_output_ready && orte_help_want_aggregate) {
         pnli = OBJ_NEW(process_name_list_item_t);
         if (NULL == pnli) {
             rc = ORTE_ERR_OUT_OF_RESOURCE;
@@ -711,8 +689,6 @@ int orte_output_init(void)
 {
     OPAL_OUTPUT_VERBOSE((5, orte_debug_output, "orte_output init"));
 
-    register_mca();
-
     /* define the default stream that has everything off */
     OBJ_CONSTRUCT(&orte_output_default, opal_output_stream_t);
 
@@ -1031,11 +1007,17 @@ int orte_show_help(const char *filename, const char *topic,
         return ORTE_SUCCESS;
     }
 
-    if (orte_process_info.hnp) {
+    /* if we are the HNP, or the RML has not yet been setup,
+     * or we don't yet know our HNP, then all we can do
+     * is process this locally
+     */
+    if (orte_process_info.hnp ||
+        NULL == orte_rml.send_buffer ||
+        ORTE_PROC_MY_HNP->vpid == ORTE_VPID_INVALID) {
         rc = show_help(filename, topic, output, ORTE_PROC_MY_NAME);
     }
     
-    /* if we are not the HNP, then we must relay the output message to
+    /* otherwise, we relay the output message to
      * the HNP for processing
      */
     else {

orte/util/proc_info.c

@@ -42,7 +42,7 @@ ORTE_DECLSPEC orte_proc_info_t orte_process_info = {
     /*  .my_name =              */   {ORTE_JOBID_INVALID, ORTE_VPID_INVALID},
     /*  .my_daemon =            */   {ORTE_JOBID_INVALID, ORTE_VPID_INVALID},
     /*  .my_daemon_uri =        */   NULL,
-    /*  .my_hnp =               */   {0, 0},
+    /*  .my_hnp =               */   {ORTE_JOBID_INVALID, ORTE_VPID_INVALID},
     /*  .my_hnp_uri =           */   NULL,
     /*  .hnp_pid =              */    0,
     /*  ,app_num =              */   -1,
@@ -134,10 +134,6 @@ int orte_proc_info(void)
                                 true, false, -1, &tmp);
     orte_process_info.universe_size = tmp;
 
-    mca_base_param_reg_string_name("orte", "tmpdir_base",
-                                   "Base of the session directory tree",
-                                   false, false, NULL,  &(orte_process_info.tmpdir_base));
-
     /* get the process id */
     orte_process_info.pid = getpid();