/**
* 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;
}
int
opal_finalize(void)
{
if( --opal_initialized != 0 ) {
if( opal_initialized < 0 ) {
return OPAL_ERROR;
}
return OPAL_SUCCESS;
}
/* close the checkpoint and restart service */
opal_cr_finalize();
opal_progress_finalize();
opal_event_fini();
/* close high resolution timers */
opal_timer_base_close();
opal_backtrace_base_close();
/* close the memory manager components. Registered hooks can
still be fired any time between now and the call to
opal_mem_free_finalize(), and callbacks from the memory manager
hooks to the bowels of the mem_free code can still occur any
time between now and end of application (even post main()!) */
opal_memory_base_close();
/* finalize the memory manager / tracker */
opal_mem_hooks_finalize();
/* close the carto framework */
opal_carto_base_close();
/* close the processor affinity base */
opal_paffinity_base_close();
/* close the memcpy base */
opal_memcpy_base_close();
/* finalize the mca */
mca_base_close();
/* finalize util code */
opal_finalize_util();
return OPAL_SUCCESS;
}
int opal_crs_blcr_restart(opal_crs_base_snapshot_t *base_snapshot, bool spawn_child, pid_t *child_pid)
{
opal_crs_blcr_snapshot_t *snapshot = OBJ_NEW(opal_crs_blcr_snapshot_t);
char **cr_argv = NULL;
char *cr_cmd = NULL;
int ret;
int exit_status = OPAL_SUCCESS;
int status;
snapshot->super = *base_snapshot;
opal_output_verbose(10, mca_crs_blcr_component.super.output_handle,
"crs:blcr: restart(%s, %d)", snapshot->super.reference_name, spawn_child);
/*
* If we need to reconstruct the snapshot,
*/
if(snapshot->super.cold_start) {
if( OPAL_SUCCESS != (ret = blcr_cold_start(snapshot)) ) {
exit_status = OPAL_ERROR;
opal_output(mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: Unable to reconstruct the snapshot.");
goto cleanup;
}
}
/*
* Get the restart command
*/
if ( OPAL_SUCCESS != (ret = opal_crs_blcr_restart_cmd(snapshot->context_filename, &cr_cmd)) ) {
exit_status = ret;
goto cleanup;
}
if ( NULL == (cr_argv = opal_argv_split(cr_cmd, ' ')) ) {
exit_status = OPAL_ERROR;
goto cleanup;
}
/*
* Restart by replacing this process
*/
/* Need to shutdown the event engine before this.
* for some reason the BLCR checkpointer and our event engine don't get
* along very well.
*/
opal_progress_finalize();
opal_event_fini();
if (!spawn_child) {
opal_output_verbose(10, mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: SELF: exec :(%s, %s):",
strdup(blcr_restart_cmd),
opal_argv_join(cr_argv, ' '));
status = execvp(strdup(blcr_restart_cmd), cr_argv);
if(status < 0) {
opal_output(mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: SELF: Child failed to execute :(%d):", status);
}
opal_output(mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: SELF: execvp returned %d", status);
exit_status = status;
goto cleanup;
}
/*
* Restart by starting a new process
*/
else {
*child_pid = fork();
if( 0 == *child_pid) {
/* Child Process */
opal_output_verbose(10, mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: CHILD: exec :(%s, %s):",
strdup(blcr_restart_cmd),
opal_argv_join(cr_argv, ' '));
status = execvp(strdup(blcr_restart_cmd), cr_argv);
if(status < 0) {
opal_output(mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: CHILD: Child failed to execute :(%d):", status);
}
opal_output(mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: CHILD: execvp returned %d", status);
exit_status = status;
goto cleanup;
}
else if(*child_pid > 0) {
/* Parent is done once it is started. */
;
}
else {
opal_output(mca_crs_blcr_component.super.output_handle,
"crs:blcr: blcr_restart: CHILD: fork failed :(%d):", *child_pid);
}
}
cleanup:
if(NULL != cr_cmd)
free(cr_cmd);
if(NULL != cr_argv)
opal_argv_free(cr_argv);
return exit_status;
}
