Add the initial_errhandler info key to MPI_INFO_ENV and populate the

value from prun populated paremeters

Signed-off-by: Aurélien Bouteiller <bouteill@icl.utk.edu>

Allow errhandlers to invoke the initial error handler before MPI_INIT

Signed-off-by: Aurelien Bouteiller <bouteill@icl.utk.edu>

Indentation

Signed-off-by: Aurelien Bouteiller <bouteill@icl.utk.edu>

ompi/errhandler/errhandler.c

@@ -78,6 +78,55 @@ ompi_predefined_errhandler_t ompi_mpi_errors_throw_exceptions = {{{0}}};
 ompi_predefined_errhandler_t *ompi_mpi_errors_throw_exceptions_addr =
     &ompi_mpi_errors_throw_exceptions;
 
+static opal_mutex_t errhandler_init_lock = OPAL_MUTEX_STATIC_INIT;
+ompi_errhandler_t* ompi_initial_error_handler_eh = NULL;
+void (*ompi_initial_error_handler)(struct ompi_communicator_t **comm, int *error_code, ...) = NULL;
+
+/*
+ * Initialize the initial errhandler infrastructure only.
+ * This does not allocate any memory and does not require a corresponding fini.
+ */
+int ompi_initial_errhandler_init(void) {
+    opal_mutex_lock(&errhandler_init_lock);
+    if ( NULL != ompi_initial_error_handler ) {
+        /* Already initialized (presumably by an API call before MPI_init) */
+        opal_mutex_unlock(&errhandler_init_lock);
+        return OMPI_SUCCESS;
+    }
+
+    /* If it has been requested from the launch keys, set the initial
+     * error handler that will be attached by default with predefined
+     * communicators. We use an env because that can be obtained before
+     * OPAL and PMIx initialization.
+     */
+    char *env = getenv("OMPI_MCA_mpi_initial_errhandler");
+    if( NULL != env ) {
+        if( 0 == strcasecmp(env, "mpi_errors_are_fatal") ) {
+            ompi_initial_error_handler = &ompi_mpi_errors_are_fatal_comm_handler;
+            ompi_initial_error_handler_eh = &ompi_mpi_errors_are_fatal.eh;
+        }
+        else if( 0 == strcasecmp(env, "mpi_errors_abort") ) {
+            ompi_initial_error_handler = &ompi_mpi_errors_abort_comm_handler;
+            ompi_initial_error_handler_eh = &ompi_mpi_errors_abort.eh;
+        }
+        else if( 0 == strcasecmp(env, "mpi_errors_return") ) {
+            ompi_initial_error_handler = &ompi_mpi_errors_return_comm_handler;
+            ompi_initial_error_handler_eh = &ompi_mpi_errors_return.eh;
+        }
+        else {
+            /* invalid entry detected, ignore it, set fatal by default */
+            opal_output(0, "WARNING: invalid value for launch key 'mpi_initial_errhandler'; defaulting to 'mpi_errors_are_fatal'.");
+            ompi_initial_error_handler = &ompi_mpi_errors_are_fatal_comm_handler;
+            ompi_initial_error_handler_eh = &ompi_mpi_errors_are_fatal.eh;
+        }
+    }
+    else {
+        ompi_initial_error_handler = &ompi_mpi_errors_are_fatal_comm_handler;
+        ompi_initial_error_handler_eh = &ompi_mpi_errors_are_fatal.eh;
+    }
+    opal_mutex_unlock(&errhandler_init_lock);
+    return OMPI_SUCCESS;
+}
 
 /*
  * Initialize OMPI errhandler infrastructure
@@ -163,9 +212,12 @@ int ompi_errhandler_init(void)
                    "MPI_ERRORS_THROW_EXCEPTIONS",
                    sizeof(ompi_mpi_errors_throw_exceptions.eh.eh_name));
 
-  /* All done */
-
-  return OMPI_SUCCESS;
+  /* Lets initialize the initial error handler if not already done */
+  char *env = getenv("OMPI_MCA_mpi_initial_errhandler");
+  if( NULL != env ) {
+    ompi_process_info.initial_errhandler = strndup(env, MPI_MAX_INFO_VAL);
+  }
+  return ompi_initial_errhandler_init();
 }
 
 

ompi/errhandler/errhandler.h

@@ -185,6 +185,26 @@ OMPI_DECLSPEC extern ompi_predefined_errhandler_t ompi_mpi_errors_throw_exceptio
  */
 OMPI_DECLSPEC extern opal_pointer_array_t ompi_errhandler_f_to_c_table;
 
+/**
+ * This function selects the initial error handler.
+ * It may be called during MPI_INIT, or during the first MPI call
+ * that raises an error. This function does not allocate memory,
+ * and will only populate the ompi_initial_error_handler_eh and
+ * ompi_initial_error_handler pointers with predefined error handler
+ * and error handler functions aliases.
+ */
+OMPI_DECLSPEC int ompi_initial_errhandler_init(void);
+/**
+ * The initial error handler pointer. Will be set to alias one of the
+ * predefined error handlers through launch keys during the first MPI call,
+ * and will then be attached to predefined communicators.
+ */
+OMPI_DECLSPEC extern ompi_errhandler_t* ompi_initial_error_handler_eh;
+/**
+ * The initial error handler function pointer. Will be called when an error
+ * is raised before MPI_INIT or after MPI_FINALIZE.
+ */
+OMPI_DECLSPEC extern void (*ompi_initial_error_handler)(struct ompi_communicator_t **comm, int *error_code, ...);
 
 /**
  * Forward declaration so that we don't have to include

ompi/info/info.c

@@ -131,6 +131,12 @@ int ompi_mpiinfo_init(void)
     opal_info_set(&ompi_mpi_info_env.info.super, "soft", cptr);
     free(cptr);
 
+    /* the initial error handler, set it as requested (nothing if not
+     * requested) */
+    if (NULL != ompi_process_info.initial_errhandler) {
+        opal_info_set(&ompi_mpi_info_env.info.super, "mpi_initial_errhandler", ompi_process_info.initial_errhandler);
+    }
+
     /* local host name */
     opal_info_set(&ompi_mpi_info_env.info.super, "host", ompi_process_info.nodename);
 

ompi/runtime/ompi_mpi_init.c

@@ -751,12 +751,6 @@ int ompi_mpi_init(int argc, char **argv, int requested, int *provided,
         goto error;
     }
 
-    /* initialize info */
-    if (OMPI_SUCCESS != (ret = ompi_mpiinfo_init())) {
-        error = "ompi_info_init() failed";
-        goto error;
-    }
-
     /* initialize error handlers */
     if (OMPI_SUCCESS != (ret = ompi_errhandler_init())) {
         error = "ompi_errhandler_init() failed";
@@ -775,6 +769,12 @@ int ompi_mpi_init(int argc, char **argv, int requested, int *provided,
         goto error;
     }
 
+    /* initialize info */
+    if (OMPI_SUCCESS != (ret = ompi_mpiinfo_init())) {
+        error = "ompi_info_init() failed";
+        goto error;
+    }
+
     /* initialize groups  */
     if (OMPI_SUCCESS != (ret = ompi_group_init())) {
         error = "ompi_group_init() failed";

ompi/runtime/ompi_rte.c

@@ -935,6 +935,11 @@ int ompi_rte_finalize(void)
         opal_process_info.initial_wdir = NULL;
     }
 
+    if (NULL != opal_process_info.initial_errhandler) {
+        free(opal_process_info.initial_errhandler);
+        opal_process_info.initial_errhandler = NULL;
+    }
+
     /* cleanup our internal nspace hack */
     opal_pmix_finalize_nspace_tracker();
 

opal/util/proc.c

@@ -51,7 +51,8 @@ opal_process_info_t opal_process_info = {
     .num_apps = 0,
     .initial_wdir = NULL,
     .reincarnation = 0,
-    .proc_is_bound = false
+    .proc_is_bound = false,
+    .initial_errhandler = NULL,
 };
 
 static opal_proc_t opal_local_proc = {

opal/util/proc.h

@@ -126,6 +126,7 @@ typedef struct opal_process_info_t {
     char *initial_wdir;
     uint32_t reincarnation;
     bool proc_is_bound;
+    char *initial_errhandler;
 } opal_process_info_t;
 OPAL_DECLSPEC extern opal_process_info_t opal_process_info;