Update mapping system

Ralph Castain · Ralph Castain · commit 45f23ca5c926 · 2018-09-26T10:00:09.000-07:00
Correctly transfer job-level mapping directives for dynamically spawned
jobs to the mapping system.

Signed-off-by: Ralph Castain &lt;rhc@open-mpi.org&gt;
diff --git a/orte/mca/rmaps/base/base.h b/orte/mca/rmaps/base/base.h
@@ -123,7 +123,8 @@ ORTE_DECLSPEC int orte_rmaps_base_filter_nodes(orte_app_context_t *app,
                                                opal_list_t *nodes,
                                                bool remove);
 
-ORTE_DECLSPEC int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,
+ORTE_DECLSPEC int orte_rmaps_base_set_mapping_policy(orte_job_t *jdata,
+                                                     orte_mapping_policy_t *policy,
                                                      char **device, char *spec);
 ORTE_DECLSPEC int orte_rmaps_base_set_ranking_policy(orte_ranking_policy_t *policy,
                                                      orte_mapping_policy_t mapping,
diff --git a/orte/mca/rmaps/base/rmaps_base_frame.c b/orte/mca/rmaps/base/rmaps_base_frame.c
@@ -296,7 +296,7 @@ static int orte_rmaps_base_open(mca_base_open_flag_t flags)
                        "rmaps_base_cpus_per_proc", "rmaps_base_mapping_policy=<obj>:PE=N, default <obj>=NUMA");
     }
 
-    if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(&orte_rmaps_base.mapping,
+    if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(NULL, &orte_rmaps_base.mapping,
                                                                  &orte_rmaps_base.device,
                                                                  rmaps_base_mapping_policy))) {
         return rc;
@@ -593,7 +593,8 @@ static int check_modifiers(char *ck, orte_mapping_policy_t *tmp)
     return ORTE_ERR_TAKE_NEXT_OPTION;
 }
 
-int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,
+int orte_rmaps_base_set_mapping_policy(orte_job_t *jdata,
+                                       orte_mapping_policy_t *policy,
                                        char **device, char *inspec)
 {
     char *ck;
@@ -681,7 +682,11 @@ int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,
                     }
                 }
                 /* now save the pattern */
-                orte_rmaps_base.ppr = strdup(ck);
+                if (NULL == jdata || NULL == jdata->map) {
+                    orte_rmaps_base.ppr = strdup(ck);
+                } else {
+                    jdata->map->ppr = strdup(ck);
+                }
                 ORTE_SET_MAPPING_POLICY(tmp, ORTE_MAPPING_PPR);
                 ORTE_SET_MAPPING_DIRECTIVE(tmp, ORTE_MAPPING_GIVEN);
                 free(spec);
@@ -747,7 +752,11 @@ int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,
     }
 
  setpolicy:
-    *policy = tmp;
+    if (NULL == jdata || NULL == jdata->map) {
+        *policy = tmp;
+    } else {
+        jdata->map->mapping = tmp;
+    }
 
     return ORTE_SUCCESS;
 }
diff --git a/orte/orted/orted_submit.c b/orte/orted/orted_submit.c
@@ -876,7 +876,7 @@ int orte_submit_job(char *argv[], int *index,
     jdata->map = OBJ_NEW(orte_job_map_t);
 
     if (NULL != orte_cmd_options.mapping_policy) {
-        if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(&jdata->map->mapping, NULL, orte_cmd_options.mapping_policy))) {
+        if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(jdata, &jdata->map->mapping, NULL, orte_cmd_options.mapping_policy))) {
             ORTE_ERROR_LOG(rc);
             return rc;
         }
diff --git a/orte/orted/pmix/pmix_server_dyn.c b/orte/orted/pmix/pmix_server_dyn.c
@@ -308,7 +308,7 @@ int pmix_server_spawn_fn(opal_process_name_t *requestor,
                                orte_rmaps_base_print_mapping(orte_rmaps_base.mapping));
                 return ORTE_ERR_BAD_PARAM;
             }
-            rc = orte_rmaps_base_set_mapping_policy(&jdata->map->mapping,
+            rc = orte_rmaps_base_set_mapping_policy(jdata, &jdata->map->mapping,
                                                     NULL, info->data.string);
             if (ORTE_SUCCESS != rc) {
                 return rc;
diff --git a/orte/orted/pmix/pmix_server_gen.c b/orte/orted/pmix/pmix_server_gen.c
@@ -489,7 +489,7 @@ static void _query(int sd, short args, void *cbdata)
     orte_job_t *jdata;
     orte_proc_t *proct;
     orte_app_context_t *app;
-    int rc, i, k, num_replies;
+    int rc = ORTE_SUCCESS, i, k, num_replies;
     opal_list_t *results, targets, *array;
     size_t n;
     uint32_t key;
@@ -716,7 +716,7 @@ static void _query(int sd, short args, void *cbdata)
                     }
                 }
                 if (ORTE_JOBID_INVALID == jobid) {
-                    rc = ORTE_ERR_BAD_PARAM;
+                    rc = ORTE_ERR_NOT_FOUND;
                     goto done;
                 }
                 /* construct a list of values with opal_proc_info_t
@@ -810,12 +810,12 @@ static void _query(int sd, short args, void *cbdata)
     }
 
   done:
-    if (0 == opal_list_get_size(results)) {
-        rc = ORTE_ERR_NOT_FOUND;
-    } else if (opal_list_get_size(results) < opal_list_get_size(cd->info)) {
-        rc = ORTE_ERR_PARTIAL_SUCCESS;
-    } else {
-        rc = ORTE_SUCCESS;
+    if (ORTE_SUCCESS == rc) {
+        if (0 == opal_list_get_size(results)) {
+            rc = ORTE_ERR_NOT_FOUND;
+        } else if (opal_list_get_size(results) < opal_list_get_size(cd->info)) {
+            rc = ORTE_ERR_PARTIAL_SUCCESS;
+        }
     }
     cd->infocbfunc(rc, results, cd->cbdata, qrel, results);
 }

Original file line number	Diff line number	Diff line change
`@@ -296,7 +296,7 @@ static int orte_rmaps_base_open(mca_base_open_flag_t flags)`
`296`	`296`	`"rmaps_base_cpus_per_proc", "rmaps_base_mapping_policy=<obj>:PE=N, default <obj>=NUMA");`
`297`	`297`	`}`
`298`	`298`
`299`		`- if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(&orte_rmaps_base.mapping,`
	`299`	`+ if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(NULL, &orte_rmaps_base.mapping,`
`300`	`300`	`&orte_rmaps_base.device,`
`301`	`301`	`rmaps_base_mapping_policy))) {`
`302`	`302`	`return rc;`
`@@ -593,7 +593,8 @@ static int check_modifiers(char ck, orte_mapping_policy_t tmp)`
`593`	`593`	`return ORTE_ERR_TAKE_NEXT_OPTION;`
`594`	`594`	`}`
`595`	`595`
`596`		`-int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,`
	`596`	`+int orte_rmaps_base_set_mapping_policy(orte_job_t *jdata,`
	`597`	`+ orte_mapping_policy_t *policy,`
`597`	`598`	`char *device, char inspec)`
`598`	`599`	`{`
`599`	`600`	`char *ck;`
`@@ -681,7 +682,11 @@ int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,`
`681`	`682`	`}`
`682`	`683`	`}`
`683`	`684`	`/* now save the pattern */`
`684`		`- orte_rmaps_base.ppr = strdup(ck);`
	`685`	`+ if (NULL == jdata \|\| NULL == jdata->map) {`
	`686`	`+ orte_rmaps_base.ppr = strdup(ck);`
	`687`	`+ } else {`
	`688`	`+ jdata->map->ppr = strdup(ck);`
	`689`	`+ }`
`685`	`690`	`ORTE_SET_MAPPING_POLICY(tmp, ORTE_MAPPING_PPR);`
`686`	`691`	`ORTE_SET_MAPPING_DIRECTIVE(tmp, ORTE_MAPPING_GIVEN);`
`687`	`692`	`free(spec);`
`@@ -747,7 +752,11 @@ int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,`
`747`	`752`	`}`
`748`	`753`
`749`	`754`	`setpolicy:`
`750`		`- *policy = tmp;`
	`755`	`+ if (NULL == jdata \|\| NULL == jdata->map) {`
	`756`	`+ *policy = tmp;`
	`757`	`+ } else {`
	`758`	`+ jdata->map->mapping = tmp;`
	`759`	`+ }`
`751`	`760`
`752`	`761`	`return ORTE_SUCCESS;`
`753`	`762`	`}`
Original file line number	Diff line number	Diff line change
`@@ -876,7 +876,7 @@ int orte_submit_job(char argv[], int index,`
`876`	`876`	`jdata->map = OBJ_NEW(orte_job_map_t);`
`877`	`877`
`878`	`878`	`if (NULL != orte_cmd_options.mapping_policy) {`
`879`		`- if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(&jdata->map->mapping, NULL, orte_cmd_options.mapping_policy))) {`
	`879`	`+ if (ORTE_SUCCESS != (rc = orte_rmaps_base_set_mapping_policy(jdata, &jdata->map->mapping, NULL, orte_cmd_options.mapping_policy))) {`
`880`	`880`	`ORTE_ERROR_LOG(rc);`
`881`	`881`	`return rc;`
`882`	`882`	`}`
Original file line number	Diff line number	Diff line change
`@@ -308,7 +308,7 @@ int pmix_server_spawn_fn(opal_process_name_t *requestor,`
`308`	`308`	`orte_rmaps_base_print_mapping(orte_rmaps_base.mapping));`
`309`	`309`	`return ORTE_ERR_BAD_PARAM;`
`310`	`310`	`}`
`311`		`- rc = orte_rmaps_base_set_mapping_policy(&jdata->map->mapping,`
	`311`	`+ rc = orte_rmaps_base_set_mapping_policy(jdata, &jdata->map->mapping,`
`312`	`312`	`NULL, info->data.string);`
`313`	`313`	`if (ORTE_SUCCESS != rc) {`
`314`	`314`	`return rc;`
Original file line number	Diff line number	Diff line change
`@@ -489,7 +489,7 @@ static void _query(int sd, short args, void *cbdata)`
`489`	`489`	`orte_job_t *jdata;`
`490`	`490`	`orte_proc_t *proct;`
`491`	`491`	`orte_app_context_t *app;`
`492`		`- int rc, i, k, num_replies;`
	`492`	`+ int rc = ORTE_SUCCESS, i, k, num_replies;`
`493`	`493`	`opal_list_t results, targets, array;`
`494`	`494`	`size_t n;`
`495`	`495`	`uint32_t key;`
`@@ -716,7 +716,7 @@ static void _query(int sd, short args, void *cbdata)`
`716`	`716`	`}`
`717`	`717`	`}`
`718`	`718`	`if (ORTE_JOBID_INVALID == jobid) {`
`719`		`- rc = ORTE_ERR_BAD_PARAM;`
	`719`	`+ rc = ORTE_ERR_NOT_FOUND;`
`720`	`720`	`goto done;`
`721`	`721`	`}`
`722`	`722`	`/* construct a list of values with opal_proc_info_t`
`@@ -810,12 +810,12 @@ static void _query(int sd, short args, void *cbdata)`
`810`	`810`	`}`
`811`	`811`
`812`	`812`	`done:`
`813`		`- if (0 == opal_list_get_size(results)) {`
`814`		`- rc = ORTE_ERR_NOT_FOUND;`
`815`		`- } else if (opal_list_get_size(results) < opal_list_get_size(cd->info)) {`
`816`		`- rc = ORTE_ERR_PARTIAL_SUCCESS;`
`817`		`- } else {`
`818`		`- rc = ORTE_SUCCESS;`
	`813`	`+ if (ORTE_SUCCESS == rc) {`
	`814`	`+ if (0 == opal_list_get_size(results)) {`
	`815`	`+ rc = ORTE_ERR_NOT_FOUND;`
	`816`	`+ } else if (opal_list_get_size(results) < opal_list_get_size(cd->info)) {`
	`817`	`+ rc = ORTE_ERR_PARTIAL_SUCCESS;`
	`818`	`+ }`
`819`	`819`	`}`
`820`	`820`	`cd->infocbfunc(rc, results, cd->cbdata, qrel, results);`
`821`	`821`	`}`