Tools require their own errmgr and state components so they can handle any errors that occur in, for example, communication .

Refs trac:3992

This commit was SVN r29972.

The following Trac tickets were found above:
  Ticket 3992 --> https://svn.open-mpi.org/trac/ompi/ticket/3992

orte/mca/errmgr/default_tool/Makefile.am

@@ -0,0 +1,34 @@
+#
+# Copyright (c) 2013      Intel, Inc.  All rights reserved. 
+# $COPYRIGHT$
+# 
+# Additional copyrights may follow
+# 
+# $HEADER$
+#
+
+sources = \
+        errmgr_default_tool.h \
+        errmgr_default_tool_component.c \
+        errmgr_default_tool.c
+
+# Make the output library in this directory, and name it either
+# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
+# (for static builds).
+
+if MCA_BUILD_orte_errmgr_default_tool_DSO
+component_noinst =
+component_install = mca_errmgr_default_tool.la
+else
+component_noinst = libmca_errmgr_default_tool.la
+component_install =
+endif
+
+mcacomponentdir = $(pkglibdir)
+mcacomponent_LTLIBRARIES = $(component_install)
+mca_errmgr_default_tool_la_SOURCES = $(sources)
+mca_errmgr_default_tool_la_LDFLAGS = -module -avoid-version
+
+noinst_LTLIBRARIES = $(component_noinst)
+libmca_errmgr_default_tool_la_SOURCES =$(sources)
+libmca_errmgr_default_tool_la_LDFLAGS = -module -avoid-version

orte/mca/errmgr/default_tool/errmgr_default_tool.c

@@ -0,0 +1,125 @@
+/*
+ * Copyright (c) 2009-2011 The Trustees of Indiana University.
+ *                         All rights reserved.
+ *
+ * Copyright (c) 2010      Cisco Systems, Inc.  All rights reserved. 
+ *
+ * Copyright (c) 2004-2006 The University of Tennessee and The University
+ *                         of Tennessee Research Foundation.  All rights
+ *                         reserved.
+ * Copyright (c) 2011-2013 Los Alamos National Security, LLC.
+ *                         All rights reserved.
+ * Copyright (c) 2013      Intel, Inc. All rights reserved.
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "orte_config.h"
+
+#include <sys/types.h>
+#ifdef HAVE_UNISTD_H
+#include <unistd.h>
+#endif  /* HAVE_UNISTD_H */
+#ifdef HAVE_STRING_H
+#include <string.h>
+#endif
+
+#include "opal/util/output.h"
+#include "opal/dss/dss.h"
+
+#include "orte/util/error_strings.h"
+#include "orte/util/name_fns.h"
+#include "orte/util/show_help.h"
+#include "orte/runtime/orte_globals.h"
+#include "orte/mca/rml/rml.h"
+#include "orte/mca/odls/odls_types.h"
+#include "orte/mca/state/state.h"
+
+#include "orte/mca/errmgr/base/base.h"
+#include "orte/mca/errmgr/base/errmgr_private.h"
+#include "errmgr_default_tool.h"
+
+/*
+ * Module functions: Global
+ */
+static int init(void);
+static int finalize(void);
+
+static int abort_peers(orte_process_name_t *procs,
+                       orte_std_cntr_t num_procs,
+                       int error_code);
+
+/******************
+ * HNP module
+ ******************/
+orte_errmgr_base_module_t orte_errmgr_default_tool_module = {
+    init,
+    finalize,
+    orte_errmgr_base_log,
+    orte_errmgr_base_abort,
+    abort_peers,
+    NULL,
+    NULL,
+    NULL,
+    orte_errmgr_base_register_migration_warning,
+    orte_errmgr_base_register_error_callback,
+    orte_errmgr_base_execute_error_callbacks
+};
+
+static void proc_errors(int fd, short args, void *cbdata);
+
+/************************
+ * API Definitions
+ ************************/
+static int init(void)
+{
+    /* setup state machine to trap proc errors */
+    orte_state.add_proc_state(ORTE_PROC_STATE_ERROR, proc_errors, ORTE_ERROR_PRI);
+
+    return ORTE_SUCCESS;
+}
+
+static int finalize(void)
+{
+    return ORTE_SUCCESS;
+}
+
+static void proc_errors(int fd, short args, void *cbdata)
+{
+    orte_state_caddy_t *caddy = (orte_state_caddy_t*)cbdata;
+
+    OPAL_OUTPUT_VERBOSE((1, orte_errmgr_base_framework.framework_output,
+                         "%s errmgr:default_tool: proc %s state %s",
+                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
+                         ORTE_NAME_PRINT(&caddy->name),
+                         orte_proc_state_to_str(caddy->proc_state)));
+    
+    /*
+     * if orte is trying to shutdown, just let it
+     */
+    if (orte_finalizing) {
+        OBJ_RELEASE(caddy);
+        return;
+    }
+
+    orte_errmgr_base_abort(ORTE_ERROR_DEFAULT_EXIT_CODE, NULL);
+
+    OBJ_RELEASE(caddy);
+}
+
+static int abort_peers(orte_process_name_t *procs,
+                       orte_std_cntr_t num_procs,
+                       int error_code)
+{
+    /* just abort */
+    if (0 < opal_output_get_verbosity(orte_errmgr_base_framework.framework_output)) {
+        orte_errmgr_base_abort(error_code, "%s called abort_peers",
+                               ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
+    } else {
+        orte_errmgr_base_abort(error_code, NULL);
+    }
+    return ORTE_SUCCESS;
+}

orte/mca/errmgr/default_tool/errmgr_default_tool.h

@@ -0,0 +1,35 @@
+/*
+ * Copyright (c) 2013      Intel, Inc. All rights reserved.
+ *
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+/**
+ * @file
+ * 
+ */
+
+#ifndef MCA_ERRMGR_default_tool_EXPORT_H
+#define MCA_ERRMGR_default_tool_EXPORT_H
+
+#include "orte_config.h"
+
+#include "orte/mca/errmgr/errmgr.h"
+
+BEGIN_C_DECLS
+
+/*
+ * Local Component structures
+ */
+
+ORTE_MODULE_DECLSPEC extern orte_errmgr_base_component_t mca_errmgr_default_tool_component;
+
+ORTE_DECLSPEC extern orte_errmgr_base_module_t orte_errmgr_default_tool_module;
+
+END_C_DECLS
+
+#endif /* MCA_ERRMGR_tool_EXPORT_H */

orte/mca/errmgr/default_tool/errmgr_default_tool_component.c

@@ -0,0 +1,98 @@
+/*
+ * Copyright (c) 2013      Intel, Inc. All rights reserved.
+ *
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "orte_config.h"
+#include "opal/util/output.h"
+
+#include "orte/mca/errmgr/errmgr.h"
+#include "orte/mca/errmgr/base/base.h"
+#include "errmgr_default_tool.h"
+
+/*
+ * Public string for version number
+ */
+const char *orte_errmgr_default_tool_component_version_string = 
+    "ORTE ERRMGR default_tool MCA component version " ORTE_VERSION;
+
+/*
+ * Local functionality
+ */
+static int errmgr_default_tool_register(void);
+static int errmgr_default_tool_open(void);
+static int errmgr_default_tool_close(void);
+static int errmgr_default_tool_component_query(mca_base_module_t **module, int *priority);
+
+/*
+ * Instantiate the public struct with all of our public information
+ * and pointer to our public functions in it
+ */
+orte_errmgr_base_component_t mca_errmgr_default_tool_component =
+{
+    /* Handle the general mca_component_t struct containing 
+     *  meta information about the component
+     */
+    {
+        ORTE_ERRMGR_BASE_VERSION_3_0_0,
+        /* Component name and version */
+        "default_tool",
+        ORTE_MAJOR_VERSION,
+        ORTE_MINOR_VERSION,
+        ORTE_RELEASE_VERSION,
+        
+        /* Component open and close functions */
+        errmgr_default_tool_open,
+        errmgr_default_tool_close,
+        errmgr_default_tool_component_query,
+        errmgr_default_tool_register
+    },
+    {
+        /* The component is checkpoint ready */
+        MCA_BASE_METADATA_PARAM_CHECKPOINT
+    },
+};
+
+static int my_priority;
+
+static int errmgr_default_tool_register(void)
+{
+    mca_base_component_t *c = &mca_errmgr_default_tool_component.base_version;
+
+    my_priority = 1000;
+    (void) mca_base_component_var_register(c, "priority",
+                                           "Priority of the default_tool errmgr component",
+                                           MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
+                                           OPAL_INFO_LVL_9,
+                                           MCA_BASE_VAR_SCOPE_READONLY, &my_priority);
+    return ORTE_SUCCESS;
+}
+
+static int errmgr_default_tool_open(void) 
+{
+    return ORTE_SUCCESS;
+}
+
+static int errmgr_default_tool_close(void)
+{
+    return ORTE_SUCCESS;
+}
+
+static int errmgr_default_tool_component_query(mca_base_module_t **module, int *priority)
+{
+    if (ORTE_PROC_IS_TOOL) {
+        /* set our priority high as we are the default for tools */
+        *priority = my_priority;
+        *module = (mca_base_module_t *)&orte_errmgr_default_tool_module;
+        return ORTE_SUCCESS;        
+    }
+    
+    *priority = -1;
+    *module = NULL;
+    return ORTE_ERROR;
+}

orte/mca/ess/base/ess_base_std_tool.c

@@ -40,6 +40,7 @@
 #include "orte/mca/routed/base/base.h"
 #include "orte/mca/errmgr/errmgr.h"
 #include "orte/mca/iof/base/base.h"
+#include "orte/mca/state/base/base.h"
 #if OPAL_ENABLE_FT_CR == 1
 #include "orte/mca/snapc/base/base.h"
 #endif
@@ -67,6 +68,18 @@ int orte_ess_base_tool_setup(void)
         orte_process_info.proc_type |= ORTE_PROC_NON_MPI;
     }
     
+    /* open and setup the state machine */
+    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_state_base_framework, 0))) {
+        ORTE_ERROR_LOG(ret);
+        error = "orte_state_base_open";
+        goto error;
+    }
+    if (ORTE_SUCCESS != (ret = orte_state_base_select())) {
+        ORTE_ERROR_LOG(ret);
+        error = "orte_state_base_select";
+        goto error;
+    }
+
     /* Setup the communication infrastructure */
     if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_oob_base_framework, 0))) {
         ORTE_ERROR_LOG(ret);

orte/mca/state/tool/Makefile.am

@@ -0,0 +1,34 @@
+#
+# Copyright (c) 2013      Intel, Inc. All rights reserved
+# $COPYRIGHT$
+# 
+# Additional copyrights may follow
+# 
+# $HEADER$
+#
+
+sources = \
+        state_tool.h \
+        state_tool_component.c \
+        state_tool.c
+
+# Make the output library in this directory, and name it either
+# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
+# (for static builds).
+
+if MCA_BUILD_orte_state_tool_DSO
+component_noinst =
+component_install = mca_state_tool.la
+else
+component_noinst = libmca_state_tool.la
+component_install =
+endif
+
+mcacomponentdir = $(pkglibdir)
+mcacomponent_LTLIBRARIES = $(component_install)
+mca_state_tool_la_SOURCES = $(sources)
+mca_state_tool_la_LDFLAGS = -module -avoid-version
+
+noinst_LTLIBRARIES = $(component_noinst)
+libmca_state_tool_la_SOURCES =$(sources)
+libmca_state_tool_la_LDFLAGS = -module -avoid-version

orte/mca/state/tool/state_tool.c

@@ -0,0 +1,102 @@
+/*
+ * Copyright (c) 2011      Los Alamos National Security, LLC.
+ *                         All rights reserved.
+ * Copyright (c) 2013      Intel, Inc. All rights reserved.
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "orte_config.h"
+
+#include <sys/types.h>
+#ifdef HAVE_UNISTD_H
+#include <unistd.h>
+#endif  /* HAVE_UNISTD_H */
+#ifdef HAVE_STRING_H
+#include <string.h>
+#endif
+
+#include "opal/util/output.h"
+
+#include "orte/runtime/orte_quit.h"
+#include "orte/mca/errmgr/errmgr.h"
+
+#include "orte/mca/state/state.h"
+#include "orte/mca/state/base/state_private.h"
+#include "state_tool.h"
+
+/*
+ * Module functions: Global
+ */
+static int init(void);
+static int finalize(void);
+
+/******************
+ * TOOL module - just uses base functions after
+ * initializing the proc state machine. Job state
+ * machine is unused by tools at this
+ * time.
+ ******************/
+orte_state_base_module_t orte_state_tool_module = {
+    init,
+    finalize,
+    orte_state_base_activate_job_state,
+    orte_state_base_add_job_state,
+    orte_state_base_set_job_state_callback,
+    orte_state_base_set_job_state_priority,
+    orte_state_base_remove_job_state,
+    orte_state_base_activate_proc_state,
+    orte_state_base_add_proc_state,
+    orte_state_base_set_proc_state_callback,
+    orte_state_base_set_proc_state_priority,
+    orte_state_base_remove_proc_state
+};
+
+static void force_quit(int fd, short args, void *cbdata)
+{
+    /* dont attempt to finalize as it could throw
+     * us into an infinite loop on errors
+     */
+    exit(orte_exit_status);
+}
+
+/************************
+ * API Definitions
+ ************************/
+static int init(void)
+{
+    int rc;
+
+    OBJ_CONSTRUCT(&orte_job_states, opal_list_t);
+    OBJ_CONSTRUCT(&orte_proc_states, opal_list_t);
+
+    /* add a default error response */
+    if (ORTE_SUCCESS != (rc = orte_state.add_job_state(ORTE_JOB_STATE_FORCED_EXIT,
+                                                       force_quit, ORTE_ERROR_PRI))) {
+        ORTE_ERROR_LOG(rc);
+        return rc;
+    }
+
+    return ORTE_SUCCESS;
+}
+
+static int finalize(void)
+{
+    opal_list_item_t *item;
+
+    /* cleanup the state machines */
+    while (NULL != (item = opal_list_remove_first(&orte_proc_states))) {
+        OBJ_RELEASE(item);
+    }
+    OBJ_DESTRUCT(&orte_proc_states);
+
+    while (NULL != (item = opal_list_remove_first(&orte_job_states))) {
+        OBJ_RELEASE(item);
+    }
+    OBJ_DESTRUCT(&orte_job_states);
+
+    return ORTE_SUCCESS;
+}

orte/mca/state/tool/state_tool.h

@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2011      Los Alamos National Security, LLC.
+ *                         All rights reserved.
+ * Copyright (c) 2013      Intel, Inc. All rights reserved
+ *
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+/**
+ * @file
+ * 
+ */
+
+#ifndef MCA_STATE_TOOL_EXPORT_H
+#define MCA_STATE_TOOL_EXPORT_H
+
+#include "orte_config.h"
+
+#include "orte/mca/state/state.h"
+
+BEGIN_C_DECLS
+
+/*
+ * Local Component structures
+ */
+
+ORTE_MODULE_DECLSPEC extern orte_state_base_component_t mca_state_tool_component;
+
+ORTE_DECLSPEC extern orte_state_base_module_t orte_state_tool_module;
+
+END_C_DECLS
+
+#endif /* MCA_STATE_TOOL_EXPORT_H */

orte/mca/state/tool/state_tool_component.c

@@ -0,0 +1,85 @@
+/*
+ * Copyright (c) 2011      Los Alamos National Security, LLC.
+ *                         All rights reserved.
+ * Copyright (c) 2013      Intel, Inc. All rights reserved.
+ *
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "orte_config.h"
+#include "opal/util/output.h"
+
+#include "orte/mca/state/state.h"
+#include "orte/mca/state/base/base.h"
+#include "state_tool.h"
+
+/*
+ * Public string for version number
+ */
+const char *orte_state_tool_component_version_string = 
+    "ORTE STATE tool MCA component version " ORTE_VERSION;
+
+/*
+ * Local functionality
+ */
+static int state_tool_open(void);
+static int state_tool_close(void);
+static int state_tool_component_query(mca_base_module_t **module, int *priority);
+
+/*
+ * Instantiate the public struct with all of our public information
+ * and pointer to our public functions in it
+ */
+orte_state_base_component_t mca_state_tool_component =
+{
+    /* Handle the general mca_component_t struct containing 
+     *  meta information about the component
+     */
+    {
+        ORTE_STATE_BASE_VERSION_1_0_0,
+        /* Component name and version */
+        "tool",
+        ORTE_MAJOR_VERSION,
+        ORTE_MINOR_VERSION,
+        ORTE_RELEASE_VERSION,
+        
+        /* Component open and close functions */
+        state_tool_open,
+        state_tool_close,
+        state_tool_component_query
+    },
+    {
+        /* The component is checkpoint ready */
+        MCA_BASE_METADATA_PARAM_CHECKPOINT
+    },
+};
+
+static int my_priority=1000;
+
+static int state_tool_open(void) 
+{
+    return ORTE_SUCCESS;
+}
+
+static int state_tool_close(void)
+{
+    return ORTE_SUCCESS;
+}
+
+static int state_tool_component_query(mca_base_module_t **module, int *priority)
+{
+    if (ORTE_PROC_IS_TOOL) {
+        /* set our priority high as we are the default for tools */
+        *priority = my_priority;
+        *module = (mca_base_module_t *)&orte_state_tool_module;
+        return ORTE_SUCCESS;        
+    }
+    
+    *priority = -1;
+    *module = NULL;
+    return ORTE_ERROR;
+}