/**
* Function for finding and opening either all MCA components, or the one
* that was specifically requested via a MCA parameter.
*/
int mca_btl_base_open(void)
{
int i;
if( ++mca_btl_base_already_opened > 1 ) return OMPI_SUCCESS;
/* Verbose output */
mca_base_param_reg_int_name("btl",
"base_verbose",
"Verbosity level of the BTL framework",
false, false,
0,
&mca_btl_base_verbose);
mca_btl_base_output = opal_output_open(NULL);
opal_output_set_verbosity(mca_btl_base_output, mca_btl_base_verbose);
/* Override the per-BTL "don't run if THREAD_MULTIPLE selected"
embargo? */
mca_base_param_reg_int_name("btl",
"base_thread_multiple_override",
"Enable BTLs that are not normally enabled when MPI_THREAD_MULTIPLE is enabled (THIS IS FOR DEVELOPERS ONLY! SHOULD NOT BE USED BY END USERS!)",
true, false,
0, &i);
mca_btl_base_thread_multiple_override = OPAL_INT_TO_BOOL(i);
/* Open up all available components */
if (OMPI_SUCCESS !=
mca_base_components_open("btl", mca_btl_base_output, mca_btl_base_static_components,
&mca_btl_base_components_opened, true)) {
return OMPI_ERROR;
}
/* Initialize the list so that in mca_btl_base_close(), we can
iterate over it (even if it's empty, as in the case of
ompi_info) */
OBJ_CONSTRUCT(&mca_btl_base_modules_initialized, opal_list_t);
/* register parameters */
mca_base_param_lookup_string(
mca_base_param_register_string("btl","base","include",NULL,NULL), &mca_btl_base_include);
mca_base_param_lookup_string(
mca_base_param_register_string("btl","base","exclude",NULL,NULL), &mca_btl_base_exclude);
mca_base_param_reg_int_name("btl", "base_warn_component_unused",
"This parameter is used to turn on warning messages when certain NICs are not used",
false, false, 1, &mca_btl_base_warn_component_unused);
/* All done */
return OMPI_SUCCESS;
}
/**
* Removes the bproc directory
* @code /tmp/openmpi-bproc-<user>/ @endcode and all of its contents
* @retval ORTE_SUCCESS
* @retval error
*/
static int
odls_bproc_remove_dir()
{
char *frontend = NULL, *user = NULL, *filename = NULL;
int id;
/* get the username set by the bproc pls. We need to get it from here
* because on many bproc systems the method we use to get the username
* from the system on the backend fails and we only get the uid. */
id = mca_base_param_register_string("pls", "bproc", "username", NULL,
orte_system_info.user);
mca_base_param_lookup_string(id,&user);
asprintf(&filename, "openmpi-bproc-%s", user );
if( NULL == filename ) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERROR;
}
frontend = opal_os_path(false, "tmp", filename, NULL );
free(filename); /* Always free the filename */
if (NULL == frontend) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERROR;
}
/* we do our best to clean up the directory tree, but we ignore errors*/
odls_bproc_delete_dir_tree(frontend);
free(frontend);
return ORTE_SUCCESS;
}
static int slave_set_name(void)
{
char *jobid_str, *procid_str;
int id, rc;
orte_jobid_t jobid;
orte_vpid_t vpid;
id = mca_base_param_register_string("orte", "ess", "jobid", NULL, NULL);
mca_base_param_lookup_string(id, &jobid_str);
if (NULL == jobid_str) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, jobid_str))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
free(jobid_str);
id = mca_base_param_register_string("orte", "ess", "vpid", NULL, NULL);
mca_base_param_lookup_string(id, &procid_str);
if (NULL == procid_str) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&vpid, procid_str))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
free(procid_str);
ORTE_PROC_MY_NAME->jobid = jobid;
ORTE_PROC_MY_NAME->vpid = vpid;
ORTE_EPOCH_SET(ORTE_PROC_MY_NAME->epoch,orte_ess.proc_get_epoch(ORTE_PROC_MY_NAME));
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slave set name to %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* get the non-name common environmental variables */
if (ORTE_SUCCESS != (rc = orte_ess_env_get())) {
ORTE_ERROR_LOG(rc);
return rc;
}
return ORTE_SUCCESS;
}
static inline char *mca_pml_v_param_register_string( const char* param_name,
char *default_value )
{
int id = mca_base_param_register_string("pml", "v", param_name, NULL, default_value);
char *param_value = default_value;
mca_base_param_lookup_string(id, ¶m_value);
return param_value;
}
static inline char* mca_btl_template_param_register_string(
const char* param_name,
const char* default_value)
{
char *param_value;
int id = mca_base_param_register_string("btl","template",param_name,NULL,default_value);
mca_base_param_lookup_string(id, ¶m_value);
return param_value;
}
int orte_proc_info(void)
{
int id, tmp;
/* all other params are set elsewhere */
id = mca_base_param_register_int("seed", NULL, NULL, NULL, orte_process_info.seed);
mca_base_param_lookup_int(id, &tmp);
orte_process_info.seed = OPAL_INT_TO_BOOL(tmp);
/* if we are a seed, then make sure the daemon flag is NOT set so that
* framework components are properly selected
*/
if (orte_process_info.seed) {
orte_process_info.daemon = false;
}
id = mca_base_param_register_int("orte", "app", "num", NULL, -1);
mca_base_param_lookup_int(id, &tmp);
orte_process_info.app_num = tmp;
id = mca_base_param_register_string("gpr", "replica", "uri", NULL, orte_process_info.gpr_replica_uri);
mca_base_param_lookup_string(id, &(orte_process_info.gpr_replica_uri));
mca_base_param_set_internal(id, true);
id = mca_base_param_register_string("ns", "replica", "uri", NULL, orte_process_info.ns_replica_uri);
mca_base_param_lookup_string(id, &(orte_process_info.ns_replica_uri));
mca_base_param_set_internal(id, true);
id = mca_base_param_register_string("tmpdir", "base", NULL, NULL, orte_process_info.tmpdir_base);
mca_base_param_lookup_string(id, &(orte_process_info.tmpdir_base));
/* get the process id */
orte_process_info.pid = getpid();
return ORTE_SUCCESS;
}
/**
* Returns a path of the form:
* @code
* /tmp/openmpi-bproc-<user>/<universe>/<jobid>-<app_context>/<proc_rank>/
* @endcode
* which is used to put links to the pty/pipes in
* @param proc_rank the process's rank on the node
* @param jobid the jobid the proc belongs to
* @param app_context the application context number within the job
* @retval path
*/
static char *
odls_bproc_get_base_dir_name(int proc_rank, orte_jobid_t jobid,
orte_std_cntr_t app_context)
{
char *path = NULL, *user = NULL, *job = NULL;
int rc;
/* ensure that system info is set */
orte_sys_info();
if (NULL == orte_universe_info.name) { /* error condition */
ORTE_ERROR_LOG(ORTE_ERROR);
return NULL;
}
rc = orte_ns.convert_jobid_to_string(&job, jobid);
if(ORTE_SUCCESS != rc) {
ORTE_ERROR_LOG(rc);
return NULL;
}
/* get the username set by the bproc pls. We need to get it from here
* because on many bproc systems the method we use to get the username
* from the system on the backend fails and we only get the uid. */
rc = mca_base_param_register_string("pls", "bproc", "username", NULL,
orte_system_info.user);
mca_base_param_lookup_string(rc,&user);
if (0 > asprintf(&path, OPAL_PATH_SEP"tmp"OPAL_PATH_SEP"openmpi-bproc-%s"OPAL_PATH_SEP"%s"OPAL_PATH_SEP"%s-%d"OPAL_PATH_SEP"%d",
user, orte_universe_info.name,
job, (int) app_context, proc_rank)) {
ORTE_ERROR_LOG(ORTE_ERROR);
path = NULL;
}
if(0 < mca_odls_bproc_component.debug) {
opal_output(0, "odls bproc io setup. Path: %s\n", path);
}
free(user);
free(job);
return path;
}
int
orte_sds_slurm_set_name(void)
{
int rc;
int id;
int vpid_start;
int num_procs;
char* name_string = NULL;
int slurm_nodeid;
/* start by getting our cellid, jobid, and vpid (which is the
starting vpid for the list of daemons) */
id = mca_base_param_register_string("ns", "nds", "name", NULL, NULL);
mca_base_param_lookup_string(id, &name_string);
if(name_string != NULL) {
if (ORTE_SUCCESS !=
(rc = orte_ns.convert_string_to_process_name(&(orte_process_info.my_name),
name_string))) {
ORTE_ERROR_LOG(rc);
free(name_string);
return rc;
}
free(name_string);
} else {
orte_cellid_t cellid;
orte_jobid_t jobid;
orte_vpid_t vpid;
char* cellid_string;
char* jobid_string;
char* vpid_string;
id = mca_base_param_register_string("ns", "nds", "cellid", NULL, NULL);
mca_base_param_lookup_string(id, &cellid_string);
if (NULL == cellid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_ns.convert_string_to_cellid(&cellid, cellid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
id = mca_base_param_register_string("ns", "nds", "jobid", NULL, NULL);
mca_base_param_lookup_string(id, &jobid_string);
if (NULL == jobid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_ns.convert_string_to_jobid(&jobid, jobid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
id = mca_base_param_register_string("ns", "nds", "vpid", NULL, NULL);
mca_base_param_lookup_string(id, &vpid_string);
if (NULL == vpid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_ns.convert_string_to_vpid(&vpid, vpid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
if (ORTE_SUCCESS != (rc = orte_ns.create_process_name(&(orte_process_info.my_name),
cellid,
jobid,
vpid))) {
ORTE_ERROR_LOG(rc);
return rc;
}
}
/* fix up the base name and make it the "real" name */
slurm_nodeid = atoi(getenv("SLURM_NODEID"));
orte_process_info.my_name->vpid += slurm_nodeid;
/* fix up the system info nodename to match exactly what slurm returned */
if (NULL != orte_system_info.nodename) {
free(orte_system_info.nodename);
}
orte_system_info.nodename = get_slurm_nodename(slurm_nodeid);
id = mca_base_param_register_int("ns", "nds", "vpid_start", NULL, -1);
mca_base_param_lookup_int(id, &vpid_start);
if (vpid_start < 0) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
id = mca_base_param_register_int("ns", "nds", "num_procs", NULL, -1);
mca_base_param_lookup_int(id, &num_procs);
if (num_procs < 0) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
orte_process_info.vpid_start = (orte_vpid_t)vpid_start;
orte_process_info.num_procs = (size_t)num_procs;
return ORTE_SUCCESS;
}
static int slurm_set_name(void)
{
int slurm_nodeid;
int rc;
int id;
orte_jobid_t jobid;
orte_vpid_t vpid;
char* jobid_string;
char* vpid_string;
char *nodeid;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm setting name"));
id = mca_base_param_register_string("orte", "ess", "jobid", NULL, NULL);
mca_base_param_lookup_string(id, &jobid_string);
if (NULL == jobid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, jobid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
id = mca_base_param_register_string("orte", "ess", "vpid", NULL, NULL);
mca_base_param_lookup_string(id, &vpid_string);
if (NULL == vpid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&vpid, vpid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
ORTE_PROC_MY_NAME->jobid = jobid;
/* fix up the vpid and make it the "real" vpid */
if (NULL == (nodeid = getenv("SLURM_NODEID"))) {
opal_output(0, "SLURM_NODEID not found - cannot define name");
return ORTE_ERR_NOT_FOUND;
}
slurm_nodeid = atoi(nodeid);
ORTE_PROC_MY_NAME->vpid = vpid + slurm_nodeid;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm set name to %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* fix up the system info nodename to match exactly what slurm returned */
if (NULL != orte_process_info.nodename) {
free(orte_process_info.nodename);
}
orte_process_info.nodename = get_slurm_nodename(slurm_nodeid);
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm set nodename to %s",
orte_process_info.nodename));
/* get the non-name common environmental variables */
if (ORTE_SUCCESS != (rc = orte_ess_env_get())) {
ORTE_ERROR_LOG(rc);
return rc;
}
return ORTE_SUCCESS;
}
