static void unbind_procs(orte_job_t *jdata)
{
int j;
orte_proc_t *proc;
for (j=0; j < jdata->procs->size; j++) {
if (NULL == (proc = (orte_proc_t*)opal_pointer_array_get_item(jdata->procs, j))) {
continue;
}
orte_remove_attribute(&proc->attributes, ORTE_PROC_HWLOC_BOUND);
orte_remove_attribute(&proc->attributes, ORTE_PROC_CPU_BITMAP);
}
}
static void orte_job_destruct(orte_job_t* job)
{
orte_proc_t *proc;
orte_app_context_t *app;
orte_job_t *jdata;
int n;
orte_timer_t *evtimer;
if (NULL == job) {
/* probably just a race condition - just return */
return;
}
if (orte_debug_flag) {
opal_output(0, "%s Releasing job data for %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), ORTE_JOBID_PRINT(job->jobid));
}
for (n=0; n < job->apps->size; n++) {
if (NULL == (app = (orte_app_context_t*)opal_pointer_array_get_item(job->apps, n))) {
continue;
}
OBJ_RELEASE(app);
}
OBJ_RELEASE(job->apps);
/* release any pointers in the attributes */
evtimer = NULL;
if (orte_get_attribute(&job->attributes, ORTE_JOB_FAILURE_TIMER_EVENT,
(void**)&evtimer, OPAL_PTR)) {
orte_remove_attribute(&job->attributes, ORTE_JOB_FAILURE_TIMER_EVENT);
/* the timer is a pointer to orte_timer_t */
OBJ_RELEASE(evtimer);
}
proc = NULL;
if (orte_get_attribute(&job->attributes, ORTE_JOB_ABORTED_PROC,
(void**)&proc, OPAL_PTR)) {
orte_remove_attribute(&job->attributes, ORTE_JOB_ABORTED_PROC);
/* points to an orte_proc_t */
OBJ_RELEASE(proc);
}
if (NULL != job->map) {
OBJ_RELEASE(job->map);
job->map = NULL;
}
for (n=0; n < job->procs->size; n++) {
if (NULL == (proc = (orte_proc_t*)opal_pointer_array_get_item(job->procs, n))) {
continue;
}
OBJ_RELEASE(proc);
}
OBJ_RELEASE(job->procs);
/* release the attributes */
OPAL_LIST_DESTRUCT(&job->attributes);
/* find the job in the global array */
if (NULL != orte_job_data && ORTE_JOBID_INVALID != job->jobid) {
for (n=0; n < orte_job_data->size; n++) {
if (NULL == (jdata = (orte_job_t*)opal_pointer_array_get_item(orte_job_data, n))) {
continue;
}
if (jdata->jobid == job->jobid) {
/* set the entry to NULL */
opal_pointer_array_set_item(orte_job_data, n, NULL);
break;
}
}
}
}
/* stuff proc attributes for sending back to a proc */
int orte_pmix_server_register_nspace(orte_job_t *jdata)
{
int rc;
orte_proc_t *pptr;
int i, k, n;
opal_list_t *info, *pmap;
opal_value_t *kv;
orte_node_t *node, *mynode;
opal_vpid_t vpid;
char **list, **procs, **micro, *tmp, *regex, *cpulist, *peerlist;
orte_job_t *dmns;
orte_job_map_t *map;
orte_app_context_t *app;
uid_t uid;
gid_t gid;
opal_list_t *cache;
opal_output_verbose(2, orte_pmix_server_globals.output,
"%s register nspace for %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_JOBID_PRINT(jdata->jobid));
/* setup the info list */
info = OBJ_NEW(opal_list_t);
uid = geteuid();
gid = getegid();
/* jobid */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_JOBID);
kv->data.string = strdup(ORTE_JOBID_PRINT(jdata->jobid));
kv->type = OPAL_STRING;
opal_list_append(info, &kv->super);
/* offset */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NPROC_OFFSET);
kv->data.uint32 = jdata->offset;
kv->type = OPAL_UINT32;
opal_list_append(info, &kv->super);
/* check for cached values to add to the job info */
cache = NULL;
if (orte_get_attribute(&jdata->attributes, ORTE_JOB_INFO_CACHE, (void**)&cache, OPAL_PTR) &&
NULL != cache) {
while (NULL != (kv = (opal_value_t*)opal_list_remove_first(cache))) {
opal_list_append(info, &kv->super);
}
orte_remove_attribute(&jdata->attributes, ORTE_JOB_INFO_CACHE);
OBJ_RELEASE(cache);
}
/* assemble the node and proc map info */
list = NULL;
procs = NULL;
map = jdata->map;
for (i=0; i < map->nodes->size; i++) {
micro = NULL;
if (NULL != (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
opal_argv_append_nosize(&list, node->name);
/* assemble all the ranks for this job that are on this node */
for (k=0; k < node->procs->size; k++) {
if (NULL != (pptr = (orte_proc_t*)opal_pointer_array_get_item(node->procs, k))) {
if (jdata->jobid == pptr->name.jobid) {
opal_argv_append_nosize(µ, ORTE_VPID_PRINT(pptr->name.vpid));
}
}
}
/* assemble the rank/node map */
if (NULL != micro) {
tmp = opal_argv_join(micro, ',');
opal_argv_free(micro);
opal_argv_append_nosize(&procs, tmp);
free(tmp);
}
}
}
/* let the PMIx server generate the nodemap regex */
if (NULL != list) {
tmp = opal_argv_join(list, ',');
opal_argv_free(list);
list = NULL;
if (OPAL_SUCCESS != (rc = opal_pmix.generate_regex(tmp, ®ex))) {
ORTE_ERROR_LOG(rc);
free(tmp);
OPAL_LIST_RELEASE(info);
return rc;
}
free(tmp);
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODE_MAP);
kv->type = OPAL_STRING;
kv->data.string = regex;
opal_list_append(info, &kv->super);
}
/* let the PMIx server generate the procmap regex */
if (NULL != procs) {
tmp = opal_argv_join(procs, ';');
opal_argv_free(procs);
procs = NULL;
if (OPAL_SUCCESS != (rc = opal_pmix.generate_ppn(tmp, ®ex))) {
ORTE_ERROR_LOG(rc);
free(tmp);
OPAL_LIST_RELEASE(info);
return rc;
}
free(tmp);
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_PROC_MAP);
kv->type = OPAL_STRING;
kv->data.string = regex;
opal_list_append(info, &kv->super);
}
/* get our local node */
if (NULL == (dmns = orte_get_job_data_object(ORTE_PROC_MY_NAME->jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
OPAL_LIST_RELEASE(info);
return ORTE_ERR_NOT_FOUND;
}
if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(dmns->procs, ORTE_PROC_MY_NAME->vpid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
OPAL_LIST_RELEASE(info);
return ORTE_ERR_NOT_FOUND;
}
mynode = pptr->node;
if (NULL == mynode) {
/* cannot happen */
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
OPAL_LIST_RELEASE(info);
return ORTE_ERR_NOT_FOUND;
}
/* pass our node ID */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODEID);
kv->type = OPAL_UINT32;
kv->data.uint32 = mynode->index;
opal_list_append(info, &kv->super);
/* pass our node size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODE_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = mynode->num_procs;
opal_list_append(info, &kv->super);
/* pass the number of nodes in the job */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NUM_NODES);
kv->type = OPAL_UINT32;
kv->data.uint32 = map->num_nodes;
opal_list_append(info, &kv->super);
/* univ size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_UNIV_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->total_slots_alloc;
opal_list_append(info, &kv->super);
/* job size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_JOB_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->num_procs;
opal_list_append(info, &kv->super);
/* number of apps in this job */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_JOB_NUM_APPS);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->num_apps;
opal_list_append(info, &kv->super);
/* local size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->num_local_procs;
opal_list_append(info, &kv->super);
/* max procs */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_MAX_PROCS);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->total_slots_alloc;
opal_list_append(info, &kv->super);
/* register any local clients */
vpid = ORTE_VPID_MAX;
for (i=0; i < mynode->procs->size; i++) {
if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(mynode->procs, i))) {
continue;
}
if (pptr->name.jobid == jdata->jobid) {
if (pptr->name.vpid < vpid) {
vpid = pptr->name.vpid;
}
/* go ahead and register this client */
if (OPAL_SUCCESS != (rc = opal_pmix.server_register_client(&pptr->name, uid, gid,
(void*)pptr, NULL, NULL))) {
ORTE_ERROR_LOG(rc);
}
}
}
/* pass the local ldr */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCALLDR);
kv->type = OPAL_VPID;
kv->data.name.vpid = vpid;
opal_list_append(info, &kv->super);
/* for each proc in this job, create an object that
* includes the info describing the proc so the recipient has a complete
* picture. This allows procs to connect to each other without
* any further info exchange, assuming the underlying transports
* support it. We also pass all the proc-specific data here so
* that each proc can lookup info about every other proc in the job */
for (n=0; n < map->nodes->size; n++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, n))) {
continue;
}
/* construct the list of local peers, while adding
* each proc's locality info */
list = NULL;
procs = NULL;
cpulist = NULL;
peerlist = NULL;
vpid = ORTE_VPID_MAX;
for (i=0; i < node->procs->size; i++) {
if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(node->procs, i))) {
continue;
}
if (pptr->name.jobid == jdata->jobid) {
opal_argv_append_nosize(&list, ORTE_VPID_PRINT(pptr->name.vpid));
if (pptr->name.vpid < vpid) {
vpid = pptr->name.vpid;
}
/* note that we have to pass the cpuset for each local
* peer so locality can be computed */
tmp = NULL;
if (orte_get_attribute(&pptr->attributes, ORTE_PROC_CPU_BITMAP, (void**)&tmp, OPAL_STRING)) {
if (NULL != tmp) {
opal_argv_append_nosize(&procs, tmp);
free(tmp);
} else {
opal_argv_append_nosize(&procs, "UNBOUND");
}
} else {
opal_argv_append_nosize(&procs, "UNBOUND");
}
}
}
/* construct the list of peers for transmission */
if (NULL != list) {
peerlist = opal_argv_join(list, ',');
opal_argv_free(list);
list = NULL;
}
/* construct the list of cpusets for transmission */
if (NULL != procs) {
cpulist = opal_argv_join(procs, ':');
opal_argv_free(procs);
procs = NULL;
}
/* if this is me, then pass the peers and cpusets to myself
* in order to maintain backward compatibility for the non-pmix
* components in OPAL/pmix */
if (node == mynode) {
/* pass the list of peers */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_PEERS);
kv->type = OPAL_STRING;
kv->data.string = strdup(peerlist);
opal_list_append(info, &kv->super);
/* pass the list of cpusets */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_CPUSETS);
kv->type = OPAL_STRING;
kv->data.string = strdup(cpulist);
opal_list_append(info, &kv->super);
}
/* now cycle across each proc on this node, passing all data that
* varies by proc */
for (i=0; i < node->procs->size; i++) {
if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(node->procs, i))) {
continue;
}
/* only consider procs from this job */
if (pptr->name.jobid != jdata->jobid) {
continue;
}
/* setup the proc map object */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_PROC_DATA);
kv->type = OPAL_PTR;
kv->data.ptr = OBJ_NEW(opal_list_t);
opal_list_append(info, &kv->super);
pmap = kv->data.ptr;
/* must start with rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_RANK);
kv->type = OPAL_VPID;
kv->data.name.vpid = pptr->name.vpid;
opal_list_append(pmap, &kv->super);
/* pass the list of peers */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_PEERS);
kv->type = OPAL_STRING;
kv->data.string = strdup(peerlist);
opal_list_append(pmap, &kv->super);
/* pass the list of cpusets */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_CPUSETS);
kv->type = OPAL_STRING;
kv->data.string = strdup(cpulist);
opal_list_append(pmap, &kv->super);
/* appnum */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APPNUM);
kv->type = OPAL_UINT32;
kv->data.uint32 = pptr->app_idx;
opal_list_append(pmap, &kv->super);
/* app ldr */
app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, pptr->app_idx);
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APPLDR);
kv->type = OPAL_VPID;
kv->data.name.vpid = app->first_rank;
opal_list_append(pmap, &kv->super);
/* global/univ rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_GLOBAL_RANK);
kv->type = OPAL_VPID;
kv->data.name.vpid = pptr->name.vpid + jdata->offset;
opal_list_append(pmap, &kv->super);
/* app rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APP_RANK);
kv->type = OPAL_VPID;
kv->data.name.vpid = pptr->app_rank;
opal_list_append(pmap, &kv->super);
/* app size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APP_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = app->num_procs;
opal_list_append(info, &kv->super);
/* local rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_RANK);
kv->type = OPAL_UINT16;
kv->data.uint16 = pptr->local_rank;
opal_list_append(pmap, &kv->super);
/* node rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODE_RANK);
kv->type = OPAL_UINT16;
kv->data.uint32 = pptr->node_rank;
opal_list_append(pmap, &kv->super);
/* hostname */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_HOSTNAME);
kv->type = OPAL_STRING;
kv->data.string = strdup(pptr->node->name);
opal_list_append(pmap, &kv->super);
/* node ID */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODEID);
kv->type = OPAL_UINT32;
kv->data.uint32 = pptr->node->index;
opal_list_append(pmap, &kv->super);
}
/* cleanup */
if (NULL != cpulist) {
free(cpulist);
}
if (NULL != peerlist) {
free(peerlist);
}
}
/* mark the job as registered */
orte_set_attribute(&jdata->attributes, ORTE_JOB_NSPACE_REGISTERED, ORTE_ATTR_LOCAL, NULL, OPAL_BOOL);
/* pass it down */
/* we are in an event, so no need to callback */
rc = opal_pmix.server_register_nspace(jdata->jobid,
jdata->num_local_procs,
info, NULL, NULL);
OPAL_LIST_RELEASE(info);
return rc;
}
static void recv_data(int fd, short args, void *cbdata)
{
bool found;
int i, rc;
orte_node_t *nd, *nd2;
opal_list_t nds, ndtmp;
opal_list_item_t *item, *itm;
char recv_msg[8192];
int nbytes, idx, sjob;
char **alloc, *nodelist, *tpn;
local_jobtracker_t *ptr, *jtrk;
local_apptracker_t *aptrk;
orte_app_context_t *app;
orte_jobid_t jobid;
orte_job_t *jdata;
char **dash_host = NULL;
opal_output_verbose(2, orte_ras_base_framework.framework_output,
"%s ras:slurm: dynamic allocation - data recvd",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
/* read the data from the socket and put it in the
* nodes field of op
*/
memset(recv_msg, 0, sizeof(recv_msg));
nbytes = read(fd, recv_msg, sizeof(recv_msg) - 1);
opal_output_verbose(2, orte_ras_base_framework.framework_output,
"%s ras:slurm: dynamic allocation msg: %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), recv_msg);
/* check if we got something */
if (0 == nbytes || 0 == strlen(recv_msg) || strstr(recv_msg, "failure") != NULL) {
/* show an error here - basically, a "nothing was available"
* message
*/
orte_show_help("help-ras-slurm.txt", "slurm-dyn-alloc-failed", true,
(0 == strlen(recv_msg)) ? "NO MSG" : recv_msg);
ORTE_ACTIVATE_JOB_STATE(NULL, ORTE_JOB_STATE_ALLOC_FAILED);
return;
}
/* break the message into its component parts, separated by colons */
alloc = opal_argv_split(recv_msg, ':');
/* the first section contains the ORTE jobid for this allocation */
tpn = strchr(alloc[0], '=');
orte_util_convert_string_to_jobid(&jobid, tpn+1);
/* get the corresponding job object */
jdata = orte_get_job_data_object(jobid);
jtrk = NULL;
/* find the associated tracking object */
for (item = opal_list_get_first(&jobs);
item != opal_list_get_end(&jobs);
item = opal_list_get_next(item)) {
ptr = (local_jobtracker_t*)item;
if (ptr->jobid == jobid) {
jtrk = ptr;
break;
}
}
if (NULL == jtrk) {
orte_show_help("help-ras-slurm.txt", "slurm-dyn-alloc-failed", true, "NO JOB TRACKER");
ORTE_ACTIVATE_JOB_STATE(NULL, ORTE_JOB_STATE_ALLOC_FAILED);
opal_argv_free(alloc);
return;
}
/* stop the timeout event */
opal_event_del(&jtrk->timeout_ev);
/* cycle across all the remaining parts - each is the allocation for
* an app in this job
*/
OBJ_CONSTRUCT(&nds, opal_list_t);
OBJ_CONSTRUCT(&ndtmp, opal_list_t);
idx = -1;
sjob = -1;
nodelist = NULL;
tpn = NULL;
for (i=1; NULL != alloc[i]; i++) {
if (ORTE_SUCCESS != parse_alloc_msg(alloc[i], &idx, &sjob, &nodelist, &tpn)) {
orte_show_help("help-ras-slurm.txt", "slurm-dyn-alloc-failed", true, jtrk->cmd);
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_ALLOC_FAILED);
opal_argv_free(alloc);
if (NULL != nodelist) {
free(nodelist);
}
if (NULL != tpn) {
free(tpn);
}
return;
}
if (idx < 0) {
orte_show_help("help-ras-slurm.txt", "slurm-dyn-alloc-failed", true, jtrk->cmd);
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_ALLOC_FAILED);
opal_argv_free(alloc);
free(nodelist);
free(tpn);
return;
}
if (NULL == (app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, idx))) {
orte_show_help("help-ras-slurm.txt", "slurm-dyn-alloc-failed", true, jtrk->cmd);
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_ALLOC_FAILED);
opal_argv_free(alloc);
free(nodelist);
free(tpn);
return;
}
/* release the current dash_host as that contained the *desired* allocation */
orte_remove_attribute(&app->attributes, ORTE_APP_DASH_HOST);
/* track the Slurm jobid */
if (NULL == (aptrk = (local_apptracker_t*)opal_pointer_array_get_item(&jtrk->apps, idx))) {
aptrk = OBJ_NEW(local_apptracker_t);
opal_pointer_array_set_item(&jtrk->apps, idx, aptrk);
}
aptrk->sjob = sjob;
/* since the nodelist/tpn may contain regular expressions, parse them */
if (ORTE_SUCCESS != (rc = orte_ras_slurm_discover(nodelist, tpn, &ndtmp))) {
ORTE_ERROR_LOG(rc);
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_ALLOC_FAILED);
opal_argv_free(alloc);
free(nodelist);
free(tpn);
return;
}
/* transfer the discovered nodes to our node list, and construct
* the new dash_host entry to match what was allocated
*/
while (NULL != (item = opal_list_remove_first(&ndtmp))) {
nd = (orte_node_t*)item;
opal_argv_append_nosize(&dash_host, nd->name);
/* check for duplicates */
found = false;
for (itm = opal_list_get_first(&nds);
itm != opal_list_get_end(&nds);
itm = opal_list_get_next(itm)) {
nd2 = (orte_node_t*)itm;
if (0 == strcmp(nd->name, nd2->name)) {
found = true;
nd2->slots += nd->slots;
OBJ_RELEASE(item);
break;
}
}
if (!found) {
/* append the new node to our list */
opal_list_append(&nds, item);
}
}
/* cleanup */
free(nodelist);
free(tpn);
}
/* cleanup */
opal_argv_free(alloc);
OBJ_DESTRUCT(&ndtmp);
if (NULL != dash_host) {
tpn = opal_argv_join(dash_host, ',');
orte_set_attribute(&app->attributes, ORTE_APP_DASH_HOST, ORTE_ATTR_LOCAL, (void*)tpn, OPAL_STRING);
opal_argv_free(dash_host);
free(tpn);
}
if (opal_list_is_empty(&nds)) {
/* if we get here, then we were able to contact slurm,
* which means we are in an actively managed cluster.
* However, slurm indicated that nothing is currently
* available that meets our requirements. This is a fatal
* situation - we do NOT have the option of running on
* user-specified hosts as the cluster is managed.
*/
OBJ_DESTRUCT(&nds);
orte_show_help("help-ras-base.txt", "ras-base:no-allocation", true);
ORTE_FORCED_TERMINATE(ORTE_ERROR_DEFAULT_EXIT_CODE);
}
/* store the found nodes */
if (ORTE_SUCCESS != (rc = orte_ras_base_node_insert(&nds, jdata))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&nds);
ORTE_FORCED_TERMINATE(ORTE_ERROR_DEFAULT_EXIT_CODE);
return;
}
OBJ_DESTRUCT(&nds);
/* default to no-oversubscribe-allowed for managed systems */
if (!(ORTE_MAPPING_SUBSCRIBE_GIVEN & ORTE_GET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping))) {
ORTE_SET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping, ORTE_MAPPING_NO_OVERSUBSCRIBE);
}
/* flag that the allocation is managed */
orte_managed_allocation = true;
/* move the job along */
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_ALLOCATION_COMPLETE);
/* all done */
return;
}
/* stuff proc attributes for sending back to a proc */
int orte_pmix_server_register_nspace(orte_job_t *jdata, bool force)
{
int rc;
orte_proc_t *pptr;
int i, k, n;
opal_list_t *info, *pmap;
opal_value_t *kv;
orte_node_t *node, *mynode;
opal_vpid_t vpid;
char **list, **procs, **micro, *tmp, *regex;
orte_job_t *dmns;
orte_job_map_t *map;
orte_app_context_t *app;
uid_t uid;
gid_t gid;
opal_list_t *cache;
hwloc_obj_t machine;
opal_output_verbose(2, orte_pmix_server_globals.output,
"%s register nspace for %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_JOBID_PRINT(jdata->jobid));
/* setup the info list */
info = OBJ_NEW(opal_list_t);
uid = geteuid();
gid = getegid();
/* pass our nspace/rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_SERVER_NSPACE);
kv->data.string = strdup(ORTE_JOBID_PRINT(ORTE_PROC_MY_NAME->jobid));
kv->type = OPAL_STRING;
opal_list_append(info, &kv->super);
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_SERVER_RANK);
kv->data.uint32 = ORTE_PROC_MY_NAME->vpid;
kv->type = OPAL_UINT32;
opal_list_append(info, &kv->super);
/* jobid */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_JOBID);
kv->data.string = strdup(ORTE_JOBID_PRINT(jdata->jobid));
kv->type = OPAL_STRING;
opal_list_append(info, &kv->super);
/* offset */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NPROC_OFFSET);
kv->data.uint32 = jdata->offset;
kv->type = OPAL_UINT32;
opal_list_append(info, &kv->super);
/* check for cached values to add to the job info */
cache = NULL;
if (orte_get_attribute(&jdata->attributes, ORTE_JOB_INFO_CACHE, (void**)&cache, OPAL_PTR) &&
NULL != cache) {
while (NULL != (kv = (opal_value_t*)opal_list_remove_first(cache))) {
opal_list_append(info, &kv->super);
}
orte_remove_attribute(&jdata->attributes, ORTE_JOB_INFO_CACHE);
OBJ_RELEASE(cache);
}
/* assemble the node and proc map info */
list = NULL;
procs = NULL;
map = jdata->map;
for (i=0; i < map->nodes->size; i++) {
micro = NULL;
if (NULL != (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
opal_argv_append_nosize(&list, node->name);
/* assemble all the ranks for this job that are on this node */
for (k=0; k < node->procs->size; k++) {
if (NULL != (pptr = (orte_proc_t*)opal_pointer_array_get_item(node->procs, k))) {
if (jdata->jobid == pptr->name.jobid) {
opal_argv_append_nosize(µ, ORTE_VPID_PRINT(pptr->name.vpid));
}
}
}
/* assemble the rank/node map */
if (NULL != micro) {
tmp = opal_argv_join(micro, ',');
opal_argv_free(micro);
opal_argv_append_nosize(&procs, tmp);
free(tmp);
}
}
}
/* let the PMIx server generate the nodemap regex */
if (NULL != list) {
tmp = opal_argv_join(list, ',');
opal_argv_free(list);
list = NULL;
if (OPAL_SUCCESS != (rc = opal_pmix.generate_regex(tmp, ®ex))) {
ORTE_ERROR_LOG(rc);
free(tmp);
OPAL_LIST_RELEASE(info);
return rc;
}
free(tmp);
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODE_MAP);
kv->type = OPAL_STRING;
kv->data.string = regex;
opal_list_append(info, &kv->super);
}
/* let the PMIx server generate the procmap regex */
if (NULL != procs) {
tmp = opal_argv_join(procs, ';');
opal_argv_free(procs);
procs = NULL;
if (OPAL_SUCCESS != (rc = opal_pmix.generate_ppn(tmp, ®ex))) {
ORTE_ERROR_LOG(rc);
free(tmp);
OPAL_LIST_RELEASE(info);
return rc;
}
free(tmp);
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_PROC_MAP);
kv->type = OPAL_STRING;
kv->data.string = regex;
opal_list_append(info, &kv->super);
}
/* get our local node */
if (NULL == (dmns = orte_get_job_data_object(ORTE_PROC_MY_NAME->jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
OPAL_LIST_RELEASE(info);
return ORTE_ERR_NOT_FOUND;
}
if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(dmns->procs, ORTE_PROC_MY_NAME->vpid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
OPAL_LIST_RELEASE(info);
return ORTE_ERR_NOT_FOUND;
}
mynode = pptr->node;
if (NULL == mynode) {
/* cannot happen */
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
OPAL_LIST_RELEASE(info);
return ORTE_ERR_NOT_FOUND;
}
/* pass our node ID */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODEID);
kv->type = OPAL_UINT32;
kv->data.uint32 = mynode->index;
opal_list_append(info, &kv->super);
/* pass our node size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODE_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = mynode->num_procs;
opal_list_append(info, &kv->super);
/* pass the number of nodes in the job */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NUM_NODES);
kv->type = OPAL_UINT32;
kv->data.uint32 = map->num_nodes;
opal_list_append(info, &kv->super);
/* univ size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_UNIV_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->total_slots_alloc;
opal_list_append(info, &kv->super);
/* job size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_JOB_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->num_procs;
opal_list_append(info, &kv->super);
/* number of apps in this job */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_JOB_NUM_APPS);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->num_apps;
opal_list_append(info, &kv->super);
/* local size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->num_local_procs;
opal_list_append(info, &kv->super);
/* max procs */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_MAX_PROCS);
kv->type = OPAL_UINT32;
kv->data.uint32 = jdata->total_slots_alloc;
opal_list_append(info, &kv->super);
/* topology signature */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_TOPOLOGY_SIGNATURE);
kv->type = OPAL_STRING;
kv->data.string = strdup(orte_topo_signature);
opal_list_append(info, &kv->super);
/* total available physical memory */
machine = hwloc_get_next_obj_by_type (opal_hwloc_topology, HWLOC_OBJ_MACHINE, NULL);
if (NULL != machine) {
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_AVAIL_PHYS_MEMORY);
kv->type = OPAL_UINT64;
#if HWLOC_API_VERSION < 0x20000
kv->data.uint64 = machine->memory.total_memory;
#else
kv->data.uint64 = machine->total_memory;
#endif
opal_list_append(info, &kv->super);
}
/* pass the mapping policy used for this job */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_MAPBY);
kv->type = OPAL_STRING;
kv->data.string = strdup(orte_rmaps_base_print_mapping(jdata->map->mapping));
opal_list_append(info, &kv->super);
/* pass the ranking policy used for this job */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_RANKBY);
kv->type = OPAL_STRING;
kv->data.string = strdup(orte_rmaps_base_print_ranking(jdata->map->ranking));
opal_list_append(info, &kv->super);
/* pass the binding policy used for this job */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_BINDTO);
kv->type = OPAL_STRING;
kv->data.string = strdup(opal_hwloc_base_print_binding(jdata->map->binding));
opal_list_append(info, &kv->super);
/* register any local clients */
vpid = ORTE_VPID_MAX;
micro = NULL;
for (i=0; i < mynode->procs->size; i++) {
if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(mynode->procs, i))) {
continue;
}
if (pptr->name.jobid == jdata->jobid) {
opal_argv_append_nosize(µ, ORTE_VPID_PRINT(pptr->name.vpid));
if (pptr->name.vpid < vpid) {
vpid = pptr->name.vpid;
}
/* go ahead and register this client */
if (OPAL_SUCCESS != (rc = opal_pmix.server_register_client(&pptr->name, uid, gid,
(void*)pptr, NULL, NULL))) {
ORTE_ERROR_LOG(rc);
}
}
}
if (NULL != micro) {
/* pass the local peers */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_PEERS);
kv->type = OPAL_STRING;
kv->data.string = opal_argv_join(micro, ',');
opal_argv_free(micro);
opal_list_append(info, &kv->super);
}
/* pass the local ldr */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCALLDR);
kv->type = OPAL_VPID;
kv->data.name.vpid = vpid;
opal_list_append(info, &kv->super);
/* for each proc in this job, create an object that
* includes the info describing the proc so the recipient has a complete
* picture. This allows procs to connect to each other without
* any further info exchange, assuming the underlying transports
* support it. We also pass all the proc-specific data here so
* that each proc can lookup info about every other proc in the job */
for (n=0; n < map->nodes->size; n++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, n))) {
continue;
}
/* cycle across each proc on this node, passing all data that
* varies by proc */
for (i=0; i < node->procs->size; i++) {
if (NULL == (pptr = (orte_proc_t*)opal_pointer_array_get_item(node->procs, i))) {
continue;
}
/* only consider procs from this job */
if (pptr->name.jobid != jdata->jobid) {
continue;
}
/* setup the proc map object */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_PROC_DATA);
kv->type = OPAL_PTR;
kv->data.ptr = OBJ_NEW(opal_list_t);
opal_list_append(info, &kv->super);
pmap = kv->data.ptr;
/* must start with rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_RANK);
kv->type = OPAL_VPID;
kv->data.name.vpid = pptr->name.vpid;
opal_list_append(pmap, &kv->super);
/* location, for local procs */
if (node == mynode) {
tmp = NULL;
if (orte_get_attribute(&pptr->attributes, ORTE_PROC_CPU_BITMAP, (void**)&tmp, OPAL_STRING) &&
NULL != tmp) {
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCALITY_STRING);
kv->type = OPAL_STRING;
kv->data.string = opal_hwloc_base_get_locality_string(opal_hwloc_topology, tmp);
opal_list_append(pmap, &kv->super);
free(tmp);
} else {
/* the proc is not bound */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCALITY_STRING);
kv->type = OPAL_STRING;
kv->data.string = NULL;
opal_list_append(pmap, &kv->super);
}
}
/* global/univ rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_GLOBAL_RANK);
kv->type = OPAL_VPID;
kv->data.name.vpid = pptr->name.vpid + jdata->offset;
opal_list_append(pmap, &kv->super);
if (1 < jdata->num_apps) {
/* appnum */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APPNUM);
kv->type = OPAL_UINT32;
kv->data.uint32 = pptr->app_idx;
opal_list_append(pmap, &kv->super);
/* app ldr */
app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, pptr->app_idx);
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APPLDR);
kv->type = OPAL_VPID;
kv->data.name.vpid = app->first_rank;
opal_list_append(pmap, &kv->super);
/* app rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APP_RANK);
kv->type = OPAL_VPID;
kv->data.name.vpid = pptr->app_rank;
opal_list_append(pmap, &kv->super);
/* app size */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_APP_SIZE);
kv->type = OPAL_UINT32;
kv->data.uint32 = app->num_procs;
opal_list_append(info, &kv->super);
}
/* local rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_LOCAL_RANK);
kv->type = OPAL_UINT16;
kv->data.uint16 = pptr->local_rank;
opal_list_append(pmap, &kv->super);
/* node rank */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODE_RANK);
kv->type = OPAL_UINT16;
kv->data.uint32 = pptr->node_rank;
opal_list_append(pmap, &kv->super);
/* node ID */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_NODEID);
kv->type = OPAL_UINT32;
kv->data.uint32 = pptr->node->index;
opal_list_append(pmap, &kv->super);
if (map->num_nodes < orte_hostname_cutoff) {
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_HOSTNAME);
kv->type = OPAL_STRING;
kv->data.string = strdup(pptr->node->name);
opal_list_append(pmap, &kv->super);
}
}
}
/* mark the job as registered */
orte_set_attribute(&jdata->attributes, ORTE_JOB_NSPACE_REGISTERED, ORTE_ATTR_LOCAL, NULL, OPAL_BOOL);
/* pass it down */
/* we are in an event, so no need to callback */
rc = opal_pmix.server_register_nspace(jdata->jobid,
jdata->num_local_procs,
info, NULL, NULL);
OPAL_LIST_RELEASE(info);
/* if the user has connected us to an external server, then we must
* assume there is going to be some cross-mpirun exchange, and so
* we protect against that situation by publishing the job info
* for this job - this allows any subsequent "connect" to retrieve
* the job info */
if (NULL != orte_data_server_uri) {
opal_buffer_t buf;
OBJ_CONSTRUCT(&buf, opal_buffer_t);
if (OPAL_SUCCESS != (rc = opal_dss.pack(&buf, &jdata, 1, ORTE_JOB))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&buf);
return rc;
}
info = OBJ_NEW(opal_list_t);
/* create a key-value with the key being the string jobid
* and the value being the byte object */
kv = OBJ_NEW(opal_value_t);
orte_util_convert_jobid_to_string(&kv->key, jdata->jobid);
kv->type = OPAL_BYTE_OBJECT;
opal_dss.unload(&buf, (void**)&kv->data.bo.bytes, &kv->data.bo.size);
OBJ_DESTRUCT(&buf);
opal_list_append(info, &kv->super);
/* set the range to be session */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_RANGE);
kv->type = OPAL_UINT;
kv->data.uint = OPAL_PMIX_RANGE_SESSION;
opal_list_append(info, &kv->super);
/* set the persistence to be app */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_PERSISTENCE);
kv->type = OPAL_INT;
kv->data.integer = OPAL_PMIX_PERSIST_APP;
opal_list_append(info, &kv->super);
/* add our effective userid to the directives */
kv = OBJ_NEW(opal_value_t);
kv->key = strdup(OPAL_PMIX_USERID);
kv->type = OPAL_UINT32;
kv->data.uint32 = geteuid();
opal_list_append(info, &kv->super);
/* now publish it */
if (ORTE_SUCCESS != (rc = pmix_server_publish_fn(ORTE_PROC_MY_NAME,
info, mycbfunc, info))) {
ORTE_ERROR_LOG(rc);
}
}
return rc;
}
int orte_errmgr_base_update_app_context_for_cr_recovery(orte_job_t *jobdata,
orte_proc_t *proc,
opal_list_t *local_snapshots)
{
int exit_status = ORTE_SUCCESS;
opal_list_item_t *item = NULL;
orte_std_cntr_t i_app;
int argc = 0;
orte_app_context_t *cur_app_context = NULL;
orte_app_context_t *new_app_context = NULL;
orte_sstore_base_local_snapshot_info_t *vpid_snapshot = NULL;
char *reference_fmt_str = NULL;
char *location_str = NULL;
char *cache_location_str = NULL;
char *ref_location_fmt_str = NULL;
char *tmp_str = NULL;
char *global_snapshot_ref = NULL;
char *global_snapshot_seq = NULL;
char *sload;
/*
* Get the snapshot restart command for this process
* JJH CLEANUP: Pass in the vpid_snapshot, so we don't have to look it up every time?
*/
for(item = opal_list_get_first(local_snapshots);
item != opal_list_get_end(local_snapshots);
item = opal_list_get_next(item) ) {
vpid_snapshot = (orte_sstore_base_local_snapshot_info_t*)item;
if(OPAL_EQUAL == orte_util_compare_name_fields(ORTE_NS_CMP_ALL,
&vpid_snapshot->process_name,
&proc->name) ) {
break;
}
else {
vpid_snapshot = NULL;
}
}
if( NULL == vpid_snapshot ) {
ORTE_ERROR_LOG(ORTE_ERROR);
exit_status = ORTE_ERROR;
goto cleanup;
}
orte_sstore.get_attr(vpid_snapshot->ss_handle,
SSTORE_METADATA_LOCAL_SNAP_REF_FMT,
&reference_fmt_str);
orte_sstore.get_attr(vpid_snapshot->ss_handle,
SSTORE_METADATA_LOCAL_SNAP_LOC,
&location_str);
orte_sstore.get_attr(vpid_snapshot->ss_handle,
SSTORE_METADATA_LOCAL_SNAP_REF_LOC_FMT,
&ref_location_fmt_str);
orte_sstore.get_attr(vpid_snapshot->ss_handle,
SSTORE_METADATA_GLOBAL_SNAP_REF,
&global_snapshot_ref);
orte_sstore.get_attr(vpid_snapshot->ss_handle,
SSTORE_METADATA_GLOBAL_SNAP_SEQ,
&global_snapshot_seq);
/*
* Find current app_context
*/
cur_app_context = NULL;
for(i_app = 0; i_app < opal_pointer_array_get_size(jobdata->apps); ++i_app) {
cur_app_context = (orte_app_context_t *)opal_pointer_array_get_item(jobdata->apps,
i_app);
if( NULL == cur_app_context ) {
continue;
}
if(proc->app_idx == cur_app_context->idx) {
break;
}
}
if( NULL == cur_app_context ) {
ORTE_ERROR_LOG(ORTE_ERROR);
exit_status = ORTE_ERROR;
goto cleanup;
}
/*
* if > 1 processes in this app context
* Create a new app_context
* Copy over attributes
* Add it to the job_t data structure
* Associate it with this process in the job
* else
* Reuse this app_context
*/
if( cur_app_context->num_procs > 1 ) {
/* Create a new app_context */
opal_dss.copy((void**)&new_app_context, cur_app_context, ORTE_APP_CONTEXT);
/* clear unused attributes */
new_app_context->idx = cur_app_context->idx;
free(new_app_context->app);
new_app_context->app = NULL;
new_app_context->num_procs = 1;
opal_argv_free(new_app_context->argv);
new_app_context->argv = NULL;
orte_remove_attribute(&new_app_context->attributes, ORTE_APP_PRELOAD_BIN);
asprintf(&tmp_str, reference_fmt_str, vpid_snapshot->process_name.vpid);
asprintf(&sload,
"%s:%s:%s:%s:%s:%s",
location_str,
global_snapshot_ref,
tmp_str,
(vpid_snapshot->compress_comp == NULL ? "" : vpid_snapshot->compress_comp),
(vpid_snapshot->compress_postfix == NULL ? "" : vpid_snapshot->compress_postfix),
global_snapshot_seq);
orte_set_attribute(&new_app_context->attributes, ORTE_APP_SSTORE_LOAD, ORTE_ATTR_LOCAL, sload, OPAL_STRING);
free(sload);
/* Add it to the job_t data structure */
/*current_global_jobdata->num_apps++; */
new_app_context->idx = (jobdata->num_apps);
proc->app_idx = new_app_context->idx;
opal_pointer_array_add(jobdata->apps, new_app_context);
++(jobdata->num_apps);
/* Remove association with the old app_context */
--(cur_app_context->num_procs);
}
else {
new_app_context = cur_app_context;
/* Cleanout old stuff */
free(new_app_context->app);
new_app_context->app = NULL;
opal_argv_free(new_app_context->argv);
new_app_context->argv = NULL;
asprintf(&tmp_str, reference_fmt_str, vpid_snapshot->process_name.vpid);
asprintf(&sload,
"%s:%s:%s:%s:%s:%s",
location_str,
global_snapshot_ref,
tmp_str,
(vpid_snapshot->compress_comp == NULL ? "" : vpid_snapshot->compress_comp),
(vpid_snapshot->compress_postfix == NULL ? "" : vpid_snapshot->compress_postfix),
global_snapshot_seq);
orte_set_attribute(&new_app_context->attributes, ORTE_APP_SSTORE_LOAD, ORTE_ATTR_LOCAL, sload, OPAL_STRING);
free(sload);
}
/*
* Update the app_context with the restart informaiton
*/
new_app_context->app = strdup("opal-restart");
opal_argv_append(&argc, &(new_app_context->argv), new_app_context->app);
opal_argv_append(&argc, &(new_app_context->argv), "-l");
opal_argv_append(&argc, &(new_app_context->argv), location_str);
opal_argv_append(&argc, &(new_app_context->argv), "-m");
opal_argv_append(&argc, &(new_app_context->argv), orte_sstore_base_local_metadata_filename);
opal_argv_append(&argc, &(new_app_context->argv), "-r");
if( NULL != tmp_str ) {
free(tmp_str);
tmp_str = NULL;
}
asprintf(&tmp_str, reference_fmt_str, vpid_snapshot->process_name.vpid);
opal_argv_append(&argc, &(new_app_context->argv), tmp_str);
cleanup:
if( NULL != tmp_str) {
free(tmp_str);
tmp_str = NULL;
}
if( NULL != location_str ) {
free(location_str);
location_str = NULL;
}
if( NULL != cache_location_str ) {
free(cache_location_str);
cache_location_str = NULL;
}
if( NULL != reference_fmt_str ) {
free(reference_fmt_str);
reference_fmt_str = NULL;
}
if( NULL != ref_location_fmt_str ) {
free(ref_location_fmt_str);
ref_location_fmt_str = NULL;
}
return exit_status;
}