b110a247be
Comm_spawn was sticking during spawn_multiple because of a problem in the dpm - the modex there is asking processes to talk to each other in an allgather_list operation, but the procs don't have the required contact info to do so. The solution here was to ensure that all parent procs have full contact info for procs in the child job. Admittedly, this isn't the long-term answer. We would like to have the contact info given to only the parent procs that were involved in the comm_spawn. There is a way to do that, but this will suffice to keep things working until that can be implemented and tested. This commit was SVN r17772.
307 строки
11 KiB
C
307 строки
11 KiB
C
/* -*- C -*-
|
|
*
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
* University Research and Technology
|
|
* Corporation. All rights reserved.
|
|
* Copyright (c) 2004-2005 The University of Tennessee and The University
|
|
* of Tennessee Research Foundation. All rights
|
|
* reserved.
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
|
* University of Stuttgart. All rights reserved.
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
* All rights reserved.
|
|
* $COPYRIGHT$
|
|
*
|
|
* Additional copyrights may follow
|
|
*
|
|
* $HEADER$
|
|
*/
|
|
/** @file:
|
|
*
|
|
*/
|
|
|
|
/*
|
|
* includes
|
|
*/
|
|
#include "orte_config.h"
|
|
#include "orte/constants.h"
|
|
#include "orte/types.h"
|
|
|
|
#include "opal/class/opal_list.h"
|
|
#include "opal/util/output.h"
|
|
#include "opal/mca/mca.h"
|
|
#include "opal/mca/base/mca_base_param.h"
|
|
|
|
#include "opal/dss/dss.h"
|
|
#include "orte/util/proc_info.h"
|
|
#include "orte/mca/errmgr/errmgr.h"
|
|
#include "orte/mca/rml/rml.h"
|
|
#include "orte/mca/rml/base/rml_contact.h"
|
|
#include "orte/mca/grpcomm/grpcomm.h"
|
|
#include "orte/mca/routed/routed.h"
|
|
#include "orte/util/name_fns.h"
|
|
#include "orte/runtime/orte_globals.h"
|
|
#include "orte/runtime/orte_wakeup.h"
|
|
#include "orte/runtime/orte_wait.h"
|
|
|
|
#include "orte/mca/plm/plm_types.h"
|
|
#include "orte/mca/plm/plm.h"
|
|
#include "orte/mca/plm/base/plm_private.h"
|
|
#include "orte/mca/plm/base/base.h"
|
|
|
|
static bool recv_issued=false;
|
|
|
|
int orte_plm_base_comm_start(void)
|
|
{
|
|
int rc;
|
|
|
|
if (recv_issued) {
|
|
return ORTE_SUCCESS;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
|
|
"%s plm:base:receive start comm",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
|
|
|
|
if (ORTE_SUCCESS != (rc = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD,
|
|
ORTE_RML_TAG_PLM,
|
|
ORTE_RML_NON_PERSISTENT,
|
|
orte_plm_base_recv,
|
|
NULL))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
}
|
|
recv_issued = true;
|
|
|
|
return rc;
|
|
}
|
|
|
|
|
|
int orte_plm_base_comm_stop(void)
|
|
{
|
|
int rc;
|
|
|
|
if (!recv_issued) {
|
|
return ORTE_SUCCESS;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
|
|
"%s plm:base:receive stop comm",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
|
|
|
|
if (ORTE_SUCCESS != (rc = orte_rml.recv_cancel(ORTE_NAME_WILDCARD, ORTE_RML_TAG_PLM))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
}
|
|
recv_issued = false;
|
|
|
|
return rc;
|
|
}
|
|
|
|
|
|
/* process incoming messages in order of receipt */
|
|
void orte_plm_base_receive_process_msg(int fd, short event, void *data)
|
|
{
|
|
orte_message_event_t *mev = (orte_message_event_t*)data;
|
|
orte_plm_cmd_flag_t command;
|
|
orte_rml_cmd_flag_t cmd;
|
|
orte_std_cntr_t count;
|
|
orte_jobid_t job;
|
|
orte_job_t *jdata;
|
|
opal_buffer_t answer, xchg;
|
|
orte_vpid_t vpid;
|
|
orte_proc_t **procs;
|
|
orte_proc_state_t state;
|
|
orte_exit_code_t exit_code;
|
|
int rc, ret;
|
|
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (rc = opal_dss.unpack(mev->buffer, &command, &count, ORTE_PLM_CMD))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
return;
|
|
}
|
|
|
|
switch (command) {
|
|
case ORTE_PLM_LAUNCH_JOB_CMD:
|
|
OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
|
|
"%s plm:base:receive job launch command",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
|
|
/* setup a default response */
|
|
OBJ_CONSTRUCT(&answer, opal_buffer_t);
|
|
job = ORTE_JOBID_INVALID;
|
|
|
|
/* get the job object */
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (rc = opal_dss.unpack(mev->buffer, &jdata, &count, ORTE_JOB))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
goto ANSWER_LAUNCH;
|
|
}
|
|
|
|
/* launch it */
|
|
if (ORTE_SUCCESS != (rc = orte_plm.spawn(jdata))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
goto ANSWER_LAUNCH;
|
|
}
|
|
job = jdata->jobid;
|
|
|
|
/* if the child is an ORTE job, wait for the procs to report they are alive */
|
|
if (!(jdata->controls & ORTE_JOB_CONTROL_NON_ORTE_JOB)) {
|
|
ORTE_PROGRESSED_WAIT(false, jdata->num_reported, jdata->num_procs);
|
|
/* pack the update command */
|
|
OBJ_CONSTRUCT(&xchg, opal_buffer_t);
|
|
cmd = ORTE_RML_UPDATE_CMD;
|
|
if (ORTE_SUCCESS != (rc = opal_dss.pack(&xchg, &cmd, 1, ORTE_RML_CMD))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
goto ANSWER_LAUNCH;
|
|
}
|
|
/* get the contact data of the child job */
|
|
if (ORTE_SUCCESS != (rc = orte_rml_base_get_contact_info(job, &xchg))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
OBJ_DESTRUCT(&xchg);
|
|
goto ANSWER_LAUNCH;
|
|
}
|
|
/* send it to the parents */
|
|
if (ORTE_SUCCESS != (ret = orte_grpcomm.xcast(mev->sender.jobid, &xchg, ORTE_RML_TAG_RML_INFO_UPDATE))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
}
|
|
OBJ_DESTRUCT(&xchg);
|
|
}
|
|
|
|
ANSWER_LAUNCH:
|
|
OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
|
|
"%s plm:base:receive job %s launched",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
|
|
ORTE_JOBID_PRINT(job)));
|
|
|
|
/* pack the jobid to be returned */
|
|
if (ORTE_SUCCESS != (ret = opal_dss.pack(&answer, &job, 1, ORTE_JOBID))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
}
|
|
|
|
/* send the response back to the sender */
|
|
if (0 > (ret = orte_rml.send_buffer(&mev->sender, &answer, ORTE_RML_TAG_PLM_PROXY, 0))) {
|
|
ORTE_ERROR_LOG(ret);
|
|
}
|
|
OBJ_DESTRUCT(&answer);
|
|
break;
|
|
|
|
case ORTE_PLM_UPDATE_PROC_STATE:
|
|
count = 1;
|
|
jdata = NULL;
|
|
while (ORTE_SUCCESS == (rc = opal_dss.unpack(mev->buffer, &job, &count, ORTE_JOBID))) {
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
|
|
"%s plm:base:receive got update_proc_state for job %s",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
|
|
ORTE_JOBID_PRINT(job)));
|
|
|
|
/* lookup the job object */
|
|
if (NULL == (jdata = orte_get_job_data_object(job))) {
|
|
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
|
|
return;
|
|
}
|
|
procs = (orte_proc_t**)jdata->procs->addr;
|
|
count = 1;
|
|
while (ORTE_SUCCESS == (rc = opal_dss.unpack(mev->buffer, &vpid, &count, ORTE_VPID))) {
|
|
if (ORTE_VPID_INVALID == vpid) {
|
|
/* flag indicates that this job is complete - move on */
|
|
break;
|
|
}
|
|
/* unpack the state */
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (rc = opal_dss.unpack(mev->buffer, &state, &count, ORTE_PROC_STATE))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
return;
|
|
}
|
|
/* unpack the exit code */
|
|
count = 1;
|
|
if (ORTE_SUCCESS != (rc = opal_dss.unpack(mev->buffer, &exit_code, &count, ORTE_EXIT_CODE))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
return;
|
|
}
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
|
|
"%s plm:base:receive got update_proc_state for vpid %lu state %lu exit_code %d",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
|
|
(unsigned long)vpid, (unsigned long)state, (int)exit_code));
|
|
|
|
/* update the termination counter IFF the state is changing to something
|
|
* indicating terminated
|
|
*/
|
|
if (ORTE_PROC_STATE_UNTERMINATED < state &&
|
|
ORTE_PROC_STATE_UNTERMINATED > procs[vpid]->state) {
|
|
++jdata->num_terminated;
|
|
}
|
|
/* update the data */
|
|
procs[vpid]->state = state;
|
|
procs[vpid]->exit_code = exit_code;
|
|
}
|
|
count = 1;
|
|
}
|
|
if (ORTE_ERR_UNPACK_READ_PAST_END_OF_BUFFER != rc) {
|
|
ORTE_ERROR_LOG(rc);
|
|
} else {
|
|
rc = ORTE_SUCCESS;
|
|
}
|
|
/* NOTE: jdata CAN BE NULL. This is caused by an orted
|
|
* being ordered to kill all its procs, but there are no
|
|
* procs left alive on that node. This can happen, for example,
|
|
* when a proc aborts somewhere, but the procs on this node
|
|
* have completed.
|
|
* So check job has to know how to handle a NULL pointer
|
|
*/
|
|
orte_plm_base_check_job_completed(jdata);
|
|
break;
|
|
|
|
default:
|
|
ORTE_ERROR_LOG(ORTE_ERR_VALUE_OUT_OF_BOUNDS);
|
|
return;
|
|
}
|
|
|
|
/* release the message */
|
|
OBJ_RELEASE(mev);
|
|
|
|
/* see if an error occurred - if so, wakeup so we can exit */
|
|
if (ORTE_SUCCESS != rc) {
|
|
orte_wakeup(1);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* handle message from proxies
|
|
* NOTE: The incoming buffer "buffer" is OBJ_RELEASED by the calling program.
|
|
* DO NOT RELEASE THIS BUFFER IN THIS CODE
|
|
*/
|
|
|
|
void orte_plm_base_recv(int status, orte_process_name_t* sender,
|
|
opal_buffer_t* buffer, orte_rml_tag_t tag,
|
|
void* cbdata)
|
|
{
|
|
int rc;
|
|
|
|
OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
|
|
"%s plm:base:receive got message from %s",
|
|
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
|
|
ORTE_NAME_PRINT(sender)));
|
|
|
|
/* don't process this right away - we need to get out of the recv before
|
|
* we process the message as it may ask us to do something that involves
|
|
* more messaging! Instead, setup an event so that the message gets processed
|
|
* as soon as we leave the recv.
|
|
*
|
|
* The macro makes a copy of the buffer, which we release above - the incoming
|
|
* buffer, however, is NOT released here, although its payload IS transferred
|
|
* to the message buffer for later processing
|
|
*/
|
|
ORTE_MESSAGE_EVENT(sender, buffer, tag, orte_plm_base_receive_process_msg);
|
|
|
|
/* reissue the recv */
|
|
if (ORTE_SUCCESS != (rc = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD,
|
|
ORTE_RML_TAG_PLM,
|
|
ORTE_RML_NON_PERSISTENT,
|
|
orte_plm_base_recv,
|
|
NULL))) {
|
|
ORTE_ERROR_LOG(rc);
|
|
}
|
|
return;
|
|
}
|
|
|