int orte_ess_base_app_finalize(void)
{
orte_notifier_base_close();
orte_cr_finalize();
#if OPAL_ENABLE_FT_CR == 1
orte_snapc_base_close();
#endif
orte_filem_base_close();
orte_wait_finalize();
/* close the multicast */
#if ORTE_ENABLE_MULTICAST
orte_rmcast_base_close();
#endif
/* now can close the rml and its friendly group comm */
orte_grpcomm_base_close();
orte_routed_base_close();
orte_rml_base_close();
orte_session_dir_finalize(ORTE_PROC_MY_NAME);
return ORTE_SUCCESS;
}
int orte_ess_base_tool_finalize(void)
{
orte_wait_finalize();
#if OPAL_ENABLE_FT_CR == 1
mca_base_framework_close(&orte_snapc_base_framework);
mca_base_framework_close(&orte_sstore_base_framework);
#endif
orte_rml.close_conduit(orte_mgmt_conduit);
/* if I am a tool, then all I will have done is
* a very small subset of orte_init - ensure that
* I only back those elements out
*/
if (NULL != orte_process_info.my_hnp_uri) {
(void) mca_base_framework_close(&orte_iof_base_framework);
}
(void) mca_base_framework_close(&orte_routed_base_framework);
(void) mca_base_framework_close(&orte_rml_base_framework);
(void) mca_base_framework_close(&orte_errmgr_base_framework);
(void) mca_base_framework_close(&opal_pmix_base_framework);
return ORTE_SUCCESS;
}
int orte_ess_base_tool_finalize(void)
{
orte_wait_finalize();
#if OPAL_ENABLE_FT_CR == 1
mca_base_framework_close(&orte_snapc_base_framework);
mca_base_framework_close(&orte_sstore_base_framework);
#endif
/* if I am a tool, then all I will have done is
* a very small subset of orte_init - ensure that
* I only back those elements out
*/
if (NULL != orte_process_info.my_hnp_uri) {
(void) mca_base_framework_close(&orte_iof_base_framework);
}
(void) mca_base_framework_close(&orte_routed_base_framework);
(void) mca_base_framework_close(&orte_rml_base_framework);
(void) mca_base_framework_close(&orte_schizo_base_framework);
(void) mca_base_framework_close(&orte_errmgr_base_framework);
/* release the event base */
if (progress_thread_running) {
opal_stop_progress_thread("orte", true);
progress_thread_running = false;
}
return ORTE_SUCCESS;
}
static void local_fin(void)
{
orte_wait_finalize();
/* close pnp */
orcm_leader_base_close();
orcm_pnp_base_close();
orte_rmcast_base_close();
orte_routed_base_close();
orte_rml_base_close();
}
/**
* Leave ORTE.
*
* @retval ORTE_SUCCESS Upon success.
* @retval ORTE_ERROR Upon failure.
*
* This function performs
*/
int orte_system_finalize(void)
{
char *contact_path;
/* if I'm the seed, remove the universe contact info file */
if (orte_process_info.seed) {
contact_path = opal_os_path(false, orte_process_info.universe_session_dir,
"universe-setup.txt", NULL);
unlink(contact_path);
free(contact_path);
}
/* rmgr and odls close depend on wait/iof */
orte_rmgr_base_close();
orte_odls_base_close();
orte_wait_finalize();
orte_iof_base_close();
orte_ns_base_close();
orte_gpr_base_close();
orte_schema_base_close();
/* finalize selected modules so they can de-register
* their receives
*/
orte_rds_base_close();
orte_ras_base_close();
orte_rmaps_base_close();
orte_pls_base_close();
/* the errmgr close function retains the base
* module so that error logging can continue
*/
orte_errmgr_base_close();
/* now can close the rml */
orte_rml_base_close();
orte_dss_close();
opal_progress_finalize();
opal_event_fini();
orte_session_dir_finalize(orte_process_info.my_name);
/* clean out the global structures */
orte_sys_info_finalize();
orte_proc_info_finalize();
orte_univ_info_finalize();
return ORTE_SUCCESS;
}
static void local_fin(void)
{
int i;
orte_node_t *node;
orte_job_t *job;
/* stop the local sensors */
orte_sensor.stop(ORTE_PROC_MY_NAME->jobid);
orte_sensor_base_close();
orte_odls_base_close();
orte_wait_finalize();
/* finalize selected modules */
orte_errmgr_base_close();
/* close the orcm-related frameworks */
orcm_leader_base_close();
orcm_pnp_base_close();
/* close the multicast */
orte_rmcast_base_close();
orte_routed_base_close();
orte_rml_base_close();
/* cleanup the job and node info arrays */
if (NULL != orte_node_pool) {
for (i=0; i < orte_node_pool->size; i++) {
if (NULL != (node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool,i))) {
OBJ_RELEASE(node);
}
}
OBJ_RELEASE(orte_node_pool);
}
if (NULL != orte_job_data) {
for (i=0; i < orte_job_data->size; i++) {
if (NULL != (job = (orte_job_t*)opal_pointer_array_get_item(orte_job_data,i))) {
OBJ_RELEASE(job);
}
}
OBJ_RELEASE(orte_job_data);
}
/* handle the orcm-specific OPAL stuff */
opal_sysinfo_base_close();
opal_pstat_base_close();
}
int orte_ess_base_tool_finalize(void)
{
orte_wait_finalize();
/* if I am a tool, then all I will have done is
* a very small subset of orte_init - ensure that
* I only back those elements out
*/
if (NULL != orte_process_info.my_hnp_uri) {
(void) mca_base_framework_close(&orte_iof_base_framework);
}
(void) mca_base_framework_close(&orte_routed_base_framework);
(void) mca_base_framework_close(&orte_rml_base_framework);
(void) mca_base_framework_close(&orte_schizo_base_framework);
(void) mca_base_framework_close(&orte_errmgr_base_framework);
return ORTE_SUCCESS;
}
static void tool_finalize(void)
{
if (signals_set) {
/* Release all local signal handlers */
opal_event_del(&epipe_handler);
opal_event_del(&term_handler);
opal_event_del(&int_handler);
opal_event_signal_del(&sigusr1_handler);
opal_event_signal_del(&sigusr2_handler);
}
(void) mca_base_framework_close(&orte_errmgr_base_framework);
(void) mca_base_framework_close(&orte_routed_base_framework);
orte_wait_finalize();
if (progress_thread_running) {
/* we had to leave the progress thread running until
* we closed the routed framework as that closure
* sends a "sync" message to the local daemon. it
* is now safe to stop the progress thread
*/
orte_event_base_active = false;
/* break the event loop */
opal_event_base_loopbreak(orte_event_base);
/* wait for thread to exit */
opal_thread_join(&progress_thread, NULL);
OBJ_DESTRUCT(&progress_thread);
progress_thread_running = false;
}
(void) mca_base_framework_close(&orte_rml_base_framework);
(void) mca_base_framework_close(&orte_oob_base_framework);
(void) mca_base_framework_close(&orte_state_base_framework);
(void) mca_base_framework_close(&orcm_db_base_framework);
(void) mca_base_framework_close(&opal_dstore_base_framework);
}
int orte_ess_base_app_finalize(void)
{
orte_cr_finalize();
#if OPAL_ENABLE_FT_CR == 1
orte_snapc_base_close();
#endif
orte_filem_base_close();
orte_wait_finalize();
orte_errmgr_base_close();
/* now can close the rml and its friendly group comm */
orte_grpcomm_base_close();
orte_db_base_close();
orte_routed_base_close();
orte_rml_base_close();
orte_session_dir_finalize(ORTE_PROC_MY_NAME);
return ORTE_SUCCESS;
}
int orte_restart(orte_process_name_t *name, const char* uri)
{
int rc;
orte_process_name_t* old_name;
orte_process_name_t* new_name;
if (ORTE_SUCCESS != (rc = orte_dss.copy((void**)&old_name, orte_process_info.my_name, ORTE_NAME))) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_dss.copy((void**)&new_name, name, ORTE_NAME))) {
ORTE_ERROR_LOG(rc);
return rc;
}
/*
* Restart event library
*/
if (ORTE_SUCCESS != (rc = opal_event_restart())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/*
* Close selected components.
*/
orte_iof_base.iof_flush = false;
if (ORTE_SUCCESS != (rc = orte_iof_base_close())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_smr_base_close())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_gpr_base_close())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_ns_base_close())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_rml_base_close())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_wait_finalize())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/*
* setup new global state
*/
orte_process_info.seed = false;
/* if NULL, set ns_replica to old_name and set the corresponding uri parameter */
if (NULL == orte_process_info.ns_replica) {
orte_process_info.ns_replica = old_name;
orte_process_info.ns_replica_uri = strdup(uri);
}
/* if NULL, set gpr_replica to old_name and set the corresponding uri parameter */
if (NULL == orte_process_info.gpr_replica) {
orte_process_info.gpr_replica = old_name;
orte_process_info.gpr_replica_uri = strdup(uri);
}
/* ensure my_name is set to the new_name */
if (NULL != orte_process_info.my_name) {
free(orte_process_info.my_name);
}
orte_process_info.my_name = new_name;
#if 0
/* close the proc_info structure so it can be reinitialized */
if (ORTE_SUCCESS != (rc = orte_proc_info_finalize())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/* set seed flag to false */
id = mca_base_param_register_int("seed", NULL, NULL, NULL, (int)false);
if (ORTE_SUCCESS != (rc = mca_base_param_set_int(id, (int)false))) {
ORTE_ERROR_LOG(rc);
return rc;
}
/* call proc_info to reset the structure */
if (ORTE_SUCCESS != (rc = orte_proc_info())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/* finalize the sys_info structure so it can be reinitialized */
if (ORTE_SUCCESS != (rc = orte_sys_info_finalize())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/* call the sys_info function to load structure with any new info */
orte_system_info.init = false;
if (ORTE_SUCCESS != (rc = orte_sys_info())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/* establish the session directory structure for this process */
if (ORTE_SUCCESS != (rc = orte_ns.get_jobid_string(&jobid_str, orte_process_info.my_name))) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_ns.get_vpid_string(&procid_str, orte_process_info.my_name))) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (orte_debug_flag) {
opal_output(0, "[%lu,%lu,%lu] setting up session dir with",
ORTE_NAME_ARGS(orte_process_info.my_name));
if (NULL != orte_process_info.tmpdir_base) {
opal_output(0, "\ttmpdir %s", orte_process_info.tmpdir_base);
}
opal_output(0, "\tuniverse %s", orte_universe_info.name);
opal_output(0, "\tuser %s", orte_system_info.user);
opal_output(0, "\thost %s", orte_system_info.nodename);
opal_output(0, "\tjobid %s", jobid_str);
opal_output(0, "\tprocid %s", procid_str);
}
if (ORTE_SUCCESS != (rc = orte_session_dir(true,
orte_process_info.tmpdir_base,
orte_system_info.user,
orte_system_info.nodename, NULL,
orte_universe_info.name,
jobid_str, procid_str))) {
ORTE_ERROR_LOG(rc);
if (jobid_str != NULL) free(jobid_str);
if (procid_str != NULL) free(procid_str);
return rc;
}
if (NULL != jobid_str) {
free(jobid_str);
}
if (NULL != procid_str) {
free(procid_str);
}
#endif
/*
* Re-open components.
*/
if (ORTE_SUCCESS != (rc = orte_wait_init())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_ns_base_open())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_rml_base_open())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_gpr_base_open())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_smr_base_open())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/*
* Select new modules.
*/
if (ORTE_SUCCESS != (rc = orte_rml_base_select())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_ns_base_select())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_gpr_base_select())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_smr_base_select())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/*
* Set contact info for the replicas
*/
if (ORTE_SUCCESS != (rc = orte_rml.set_uri(orte_process_info.ns_replica_uri))) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_rml.set_uri(orte_process_info.gpr_replica_uri))) {
ORTE_ERROR_LOG(rc);
return rc;
}
/*
* Re-init selected modules.
*/
if (ORTE_SUCCESS != (rc = orte_rml.init())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_ns.init())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_gpr.init())) {
ORTE_ERROR_LOG(rc);
return rc;
}
/*
* Complete restart
*/
if (ORTE_SUCCESS != (rc = orte_iof_base_open())) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_iof_base_select())) {
ORTE_ERROR_LOG(rc);
return rc;
}
return ORTE_SUCCESS;
}
static int rte_finalize(void)
{
char *contact_path;
opal_list_item_t *item;
/* remove my contact info file */
contact_path = opal_os_path(false, orte_process_info.top_session_dir,
"contact.txt", NULL);
unlink(contact_path);
free(contact_path);
orte_notifier_base_close();
orte_cr_finalize();
#if OPAL_ENABLE_FT == 1
orte_snapc_base_close();
#endif
orte_filem_base_close();
orte_odls_base_close();
orte_wait_finalize();
orte_iof_base_close();
/* finalize selected modules so they can de-register
* any receives
*/
orte_ras_base_close();
orte_rmaps_base_close();
orte_plm_base_close();
orte_errmgr_base_close();
/* now can close the rml and its friendly group comm */
orte_grpcomm_base_close();
orte_routed_base_close();
orte_rml_base_close();
/* cleanup the global list of local children and job data */
while (NULL != (item = opal_list_remove_first(&orte_local_children))) {
OBJ_RELEASE(item);
}
OBJ_DESTRUCT(&orte_local_children);
while (NULL != (item = opal_list_remove_first(&orte_local_jobdata))) {
OBJ_RELEASE(item);
}
OBJ_DESTRUCT(&orte_local_jobdata);
/* finalize the session directory tree */
orte_session_dir_finalize(ORTE_PROC_MY_NAME);
/* clean out the global structures */
orte_proc_info_finalize();
if (NULL != orte_job_ident) {
free(orte_job_ident);
}
/* close the xml output file, if open */
if (orte_xml_output) {
fprintf(orte_xml_fp, "</%s>\n", orte_cmd_basename);
fflush(orte_xml_fp);
if (stdout != orte_xml_fp) {
fclose(orte_xml_fp);
}
}
return ORTE_SUCCESS;
}
static int rte_finalize(void)
{
char *contact_path;
opal_list_item_t *item;
orte_node_t *node;
orte_job_t *job;
int i;
/* remove my contact info file */
contact_path = opal_os_path(false, orte_process_info.top_session_dir,
"contact.txt", NULL);
unlink(contact_path);
free(contact_path);
orte_notifier_base_close();
orte_cr_finalize();
#if OPAL_ENABLE_FT_CR == 1
orte_snapc_base_close();
#endif
orte_filem_base_close();
orte_odls_base_close();
orte_wait_finalize();
orte_iof_base_close();
/* finalize selected modules so they can de-register
* any receives
*/
orte_ras_base_close();
orte_rmaps_base_close();
orte_plm_base_close();
orte_errmgr_base_close();
/* close the multicast */
#if ORTE_ENABLE_MULTICAST
orte_rmcast_base_close();
#endif
/* now can close the rml and its friendly group comm */
orte_grpcomm_base_close();
orte_routed_base_close();
orte_rml_base_close();
/* if we were doing timing studies, close the timing file */
if (orte_timing) {
if (stdout != orte_timing_output &&
stderr != orte_timing_output) {
fclose(orte_timing_output);
}
}
/* cleanup the global list of local children and job data */
while (NULL != (item = opal_list_remove_first(&orte_local_children))) {
OBJ_RELEASE(item);
}
OBJ_DESTRUCT(&orte_local_children);
while (NULL != (item = opal_list_remove_first(&orte_local_jobdata))) {
OBJ_RELEASE(item);
}
OBJ_DESTRUCT(&orte_local_jobdata);
/* cleanup the job and node info arrays */
if (NULL != orte_node_pool) {
for (i=0; i < orte_node_pool->size; i++) {
if (NULL != (node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool,i))) {
OBJ_RELEASE(node);
}
}
OBJ_RELEASE(orte_node_pool);
}
if (NULL != orte_job_data) {
for (i=0; i < orte_job_data->size; i++) {
if (NULL != (job = (orte_job_t*)opal_pointer_array_get_item(orte_job_data,i))) {
OBJ_RELEASE(job);
}
}
OBJ_RELEASE(orte_job_data);
}
/* finalize the session directory tree */
orte_session_dir_finalize(ORTE_PROC_MY_NAME);
/* clean out the global structures */
orte_proc_info_finalize();
if (NULL != orte_job_ident) {
free(orte_job_ident);
}
/* close the xml output file, if open */
if (orte_xml_output) {
fprintf(orte_xml_fp, "</mpirun>\n");
fflush(orte_xml_fp);
if (stdout != orte_xml_fp) {
fclose(orte_xml_fp);
}
}
/* handle the orted-specific OPAL stuff */
opal_sysinfo_base_close();
opal_pstat_base_close();
return ORTE_SUCCESS;
}
