ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность

Remove two stale modules

Просмотр исходного кода 
This commit was SVN r24794.
Этот коммит содержится в:
Ralph Castain 2011-06-21 05:57:39 +00:00
родитель b95ede99d5
Коммит 9491fbb60c
14 изменённых файлов: 0 добавлений и 1966 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

46
orte/mca/plm/rshd/Makefile.am
Просмотреть файл

@ -1,46 +0,0 @@
#
# 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 (c) 2010 Cisco Systems, Inc. All rights reserved.
# $COPYRIGHT$
#
# Additional copyrights may follow
#
# $HEADER$
#
dist_pkgdata_DATA = help-plm-rshd.txt
sources = \
plm_rshd.h \
plm_rshd_component.c \
plm_rshd_module.c
# Make the output library in this directory, and name it either
# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
# (for static builds).
if MCA_BUILD_orte_plm_rshd_DSO
component_noinst =
component_install = mca_plm_rshd.la
else
component_noinst = libmca_plm_rshd.la
component_install =
endif
mcacomponentdir = $(pkglibdir)
mcacomponent_LTLIBRARIES = $(component_install)
mca_plm_rshd_la_SOURCES = $(sources)
mca_plm_rshd_la_LDFLAGS = -module -avoid-version
noinst_LTLIBRARIES = $(component_noinst)
libmca_plm_rshd_la_SOURCES =$(sources)
libmca_plm_rshd_la_LDFLAGS = -module -avoid-version

27
orte/mca/plm/rshd/configure.m4
Просмотреть файл

@ -1,27 +0,0 @@
# -*- shell-script -*-
#
# 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.
dnl Copyright (c) 2010 Cisco Systems, Inc. All rights reserved.
# $COPYRIGHT$
#
# Additional copyrights may follow
#
# $HEADER$
#
# MCA_plm_rshd_CONFIG([action-if-found], [action-if-not-found])
# -----------------------------------------------------------
AC_DEFUN([MCA_orte_plm_rshd_CONFIG],[
AC_CONFIG_FILES([orte/mca/plm/rshd/Makefile])
AC_CHECK_FUNC([fork], [$1], [$2])
])dnl

77
orte/mca/plm/rshd/help-plm-rshd.txt
Просмотреть файл

@ -1,77 +0,0 @@
# -*- text -*-
#
# 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
#
# $HEADER$
#
# This is the US/English general help file for Open RTE's orterun.
#
[no-local-orted]
The rsh PLS component was not able to find the executable "orted" in
your PATH or in the directory where Open MPI/OpenRTE was initially installed,
and therefore cannot continue.
For reference, your current PATH is:
%s
We also looked for orted in the following directory:
%s
[multiple-prefixes]
Specified multiple application contexts using different
settings for --prefix. Care should be taken, that corresponding
processes are mapped to different nodes. Having multiple prefixes
per node is not allowed.
The previously set prefix was
%s
the prefix to be set overriding:
%s
[concurrency-less-than-zero]
The value of the MCA parameter "pls_rsh_num_concurrent" is less than
or equal to zero (%d). This parameter is used to determine how many
remote agents (typically rsh or ssh) to invoke concurrently while
launching parallel jobs.
This value has automatically be reset to 1; processing will continue.
[deadlock-params]
The rsh launcher has been given a number of %d concurrent daemons to
launch and is in a debug-daemons option. However, the total number of
daemons to launch (%d) is greater than this value. This is a scenario that
will cause the system to deadlock.
To avoid deadlock, either increase the number of concurrent daemons, or
remove the debug-daemons flag.
[unknown-user]
The user (%d) is unknown to the system (i.e. there is no corresponding
entry in the password file). Please contact your system administrator
for a fix.
#
[cannot-resolve-shell-with-prefix]
The rsh launcher has been given a prefix to use, but could not determine
the type of remote shell being used on the remote node. This is a fatal
error as we cannot determine how to construct the cmd line to set your
remote LD_LIBRARY_PATH and PATH environmental variables.
The prefix we were given are:
opal_prefix: %s
prefix_dir: %s

80
orte/mca/plm/rshd/plm_rshd.h
Просмотреть файл

@ -1,80 +0,0 @@
/*
* Copyright (c) 2004-2008 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2009 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 (c) 2008 Sun Microsystems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
/**
* @file:
* Part of the rshd launcher. See plm_rshd.h for an overview of how it works.
*/
#ifndef ORTE_PLM_RSHD_EXPORT_H
#define ORTE_PLM_RSHD_EXPORT_H
#include "orte_config.h"
#ifdef HAVE_SYS_TIME_H
#include <sys/time.h>
#endif
#include "opal/threads/condition.h"
#include "opal/mca/mca.h"
#include "orte/mca/plm/plm.h"
BEGIN_C_DECLS
/*
* Module open / close
*/
int orte_plm_rshd_component_open(void);
int orte_plm_rshd_component_close(void);
int orte_plm_rshd_component_query(mca_base_module_t **module, int *priority);
/*
* Startup / Shutdown
*/
int orte_plm_rshd_finalize(void);
/*
* Interface
*/
int orte_plm_rshd_init(void);
int orte_plm_rshd_launch(orte_job_t *jdata);
int orte_plm_rshd_terminate_job(orte_jobid_t);
int orte_plm_rshd_terminate_orteds(void);
int orte_plm_rshd_signal_job(orte_jobid_t, int32_t);
/**
* PLS Component
*/
struct orte_plm_rshd_component_t {
orte_plm_base_component_t super;
bool force_rsh;
opal_list_t children;
int num_children;
int num_concurrent;
opal_mutex_t lock;
opal_condition_t cond;
};
typedef struct orte_plm_rshd_component_t orte_plm_rshd_component_t;
ORTE_MODULE_DECLSPEC extern orte_plm_rshd_component_t mca_plm_rshd_component;
extern orte_plm_base_module_t orte_plm_rshd_module;
END_C_DECLS
#endif /* ORTE_PLS_RSHD_EXPORT_H */

151
orte/mca/plm/rshd/plm_rshd_component.c
Просмотреть файл

@ -1,151 +0,0 @@
/*
* Copyright (c) 2004-2008 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 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 (c) 2007 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2008-2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2009 Cisco Systems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*
* These symbols are in a file by themselves to provide nice linker
* semantics. Since linkers generally pull in symbols by object
* files, keeping these symbols as the only symbols in this file
* prevents utility programs such as "ompi_info" from having to import
* entire components just to query their version and parameters.
*/
#include "orte_config.h"
#include "orte/constants.h"
#include <stdlib.h>
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
#include <ctype.h>
#include "opal/util/opal_environ.h"
#include "opal/util/output.h"
#include "opal/util/argv.h"
#include "opal/util/path.h"
#include "opal/mca/base/mca_base_param.h"
#include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
#include "orte/util/show_help.h"
#include "orte/mca/plm/plm.h"
#include "orte/mca/plm/base/plm_private.h"
#include "orte/mca/plm/base/plm_base_rsh_support.h"
#include "orte/mca/plm/rshd/plm_rshd.h"
/*
* Public string showing the plm ompi_rshd component version number
*/
const char *mca_plm_rshd_component_version_string =
"Open MPI rshd plm MCA component version " ORTE_VERSION;
/*
* Instantiate the public struct with all of our public information
* and pointers to our public functions in it
*/
orte_plm_rshd_component_t mca_plm_rshd_component = {
{
/* First, the mca_component_t struct containing meta information
about the component itself */
{
ORTE_PLM_BASE_VERSION_2_0_0,
/* Component name and version */
"rshd",
ORTE_MAJOR_VERSION,
ORTE_MINOR_VERSION,
ORTE_RELEASE_VERSION,
/* Component open and close functions */
orte_plm_rshd_component_open,
orte_plm_rshd_component_close,
orte_plm_rshd_component_query
},
{
/* The component is checkpoint ready */
MCA_BASE_METADATA_PARAM_CHECKPOINT
}
}
};
int orte_plm_rshd_component_open(void)
{
int tmp;
mca_base_component_t *c = &mca_plm_rshd_component.super.base_version;
/* initialize globals */
OBJ_CONSTRUCT(&mca_plm_rshd_component.lock, opal_mutex_t);
OBJ_CONSTRUCT(&mca_plm_rshd_component.cond, opal_condition_t);
mca_plm_rshd_component.num_children = 0;
OBJ_CONSTRUCT(&mca_plm_rshd_component.children, opal_list_t);
/* lookup parameters */
mca_base_param_reg_int(c, "num_concurrent",
"How many plm_rsh_agent instances to invoke concurrently (must be > 0)",
false, false, 128, &tmp);
if (tmp <= 0) {
orte_show_help("help-plm-rsh.txt", "concurrency-less-than-zero",
true, tmp);
tmp = 1;
}
mca_plm_rshd_component.num_concurrent = tmp;
mca_base_param_reg_int(c, "force_rsh",
"Force the launcher to always use rsh",
false, false, false, &tmp);
mca_plm_rshd_component.force_rsh = OPAL_INT_TO_BOOL(tmp);
return ORTE_SUCCESS;
}
int orte_plm_rshd_component_query(mca_base_module_t **module, int *priority)
{
if (ORTE_SUCCESS != orte_plm_base_rsh_launch_agent_lookup(NULL, NULL)) {
/* this isn't an error - we just cannot be selected */
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:rshd: unable to be used: cannot find \"%s\" in PATH",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), orte_rsh_agent));
*module = NULL;
return ORTE_ERROR;
}
/* we are good - make ourselves available, but at low priority */
*priority = 0;
*module = (mca_base_module_t *) &orte_plm_rshd_module;
return ORTE_SUCCESS;
}
int orte_plm_rshd_component_close(void)
{
/* cleanup state */
OBJ_DESTRUCT(&mca_plm_rshd_component.lock);
OBJ_DESTRUCT(&mca_plm_rshd_component.cond);
OBJ_DESTRUCT(&mca_plm_rshd_component.children);
return ORTE_SUCCESS;
}

434
orte/mca/plm/rshd/plm_rshd_module.c
Просмотреть файл

@ -1,434 +0,0 @@
/*
* Copyright (c) 2004-2007 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2007 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2006 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 (c) 2006-2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2007 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2008-2009 Sun Microsystems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*
* These symbols are in a file by themselves to provide nice linker
* semantics. Since linkers generally pull in symbols by object
* files, keeping these symbols as the only symbols in this file
* prevents utility programs such as "ompi_info" from having to import
* entire components just to query their version and parameters.
*/
#include "orte_config.h"
#include "orte/constants.h"
#include <stdlib.h>
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
#include <errno.h>
#include <string.h>
#ifdef HAVE_STRINGS_H
#include <strings.h>
#endif
#ifdef HAVE_SYS_SELECT_H
#include <sys/select.h>
#endif
#ifdef HAVE_SYS_TIME_H
#include <sys/time.h>
#endif
#ifdef HAVE_SYS_TYPES_H
#include <sys/types.h>
#endif
#ifdef HAVE_SYS_STAT_H
#include <sys/stat.h>
#endif
#ifdef HAVE_SYS_WAIT_H
#include <sys/wait.h>
#endif
#include <fcntl.h>
#include <signal.h>
#ifdef HAVE_PWD_H
#include <pwd.h>
#endif
#include "opal/mca/installdirs/installdirs.h"
#include "opal/mca/base/mca_base_param.h"
#include "opal/util/output.h"
#include "opal/util/opal_sos.h"
#include "opal/mca/event/event.h"
#include "opal/util/argv.h"
#include "opal/util/opal_environ.h"
#include "opal/util/basename.h"
#include "opal/util/bit_ops.h"
#include "orte/util/show_help.h"
#include "orte/runtime/orte_wait.h"
#include "orte/runtime/orte_globals.h"
#include "orte/runtime/orte_quit.h"
#include "orte/util/name_fns.h"
#include "orte/util/nidmap.h"
#include "orte/util/proc_info.h"
#include "orte/mca/rml/rml.h"
#include "orte/mca/rml/rml_types.h"
#include "orte/mca/ess/ess.h"
#include "orte/mca/ess/base/base.h"
#include "orte/mca/errmgr/errmgr.h"
#include "orte/mca/rmaps/rmaps.h"
#include "orte/mca/routed/routed.h"
#include "orte/mca/plm/plm.h"
#include "orte/mca/plm/base/base.h"
#include "orte/mca/plm/base/plm_private.h"
#include "orte/mca/plm/base/plm_base_rsh_support.h"
#include "orte/mca/plm/rshd/plm_rshd.h"
static void ssh_child(char *cmd, char **argv) __opal_attribute_noreturn__;
orte_plm_base_module_t orte_plm_rshd_module = {
orte_plm_rshd_init,
orte_plm_base_set_hnp_name,
orte_plm_rshd_launch,
NULL,
orte_plm_rshd_terminate_job,
orte_plm_rshd_terminate_orteds,
NULL,
orte_plm_rshd_signal_job,
orte_plm_rshd_finalize
};
/*
* Local functions
*/
static void set_handler_default(int sig);
/**
* Init the module
*/
int orte_plm_rshd_init(void)
{
int rc;
/* since I was selected, setup the rsh launch agent support */
if (ORTE_SUCCESS != (rc = orte_plm_base_rsh_launch_agent_setup(NULL, NULL))) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (ORTE_SUCCESS != (rc = orte_plm_base_comm_start())) {
ORTE_ERROR_LOG(rc);
}
return rc;
}
/**
* Callback on daemon exit.
*/
static void wait_cb(pid_t pid, int status, void* cbdata)
{
orte_proc_t *proc = (orte_proc_t*)cbdata;
orte_job_t *jdata;
/* get the associated job object */
jdata = orte_get_job_data_object(proc->name.jobid);
if (! WIFEXITED(status) || ! WEXITSTATUS(status) == 0) { /* if abnormal exit */
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s proc %d failed with status %d",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(int)proc->name.vpid, WEXITSTATUS(status)));
}
/* note that this daemon failed */
orte_errmgr.update_state(proc->name.jobid, ORTE_JOB_STATE_FAILED_TO_START,
NULL, ORTE_PROC_STATE_FAILED_TO_START, 0, status);
/* release any waiting threads */
OPAL_THREAD_LOCK(&mca_plm_rshd_component.lock);
/* decrement our #children */
mca_plm_rshd_component.num_children--;
/* see if we can allow launching to continue */
if (mca_plm_rshd_component.num_children <=
mca_plm_rshd_component.num_concurrent ||
mca_plm_rshd_component.num_children == 0) {
opal_condition_signal(&mca_plm_rshd_component.cond);
}
OPAL_THREAD_UNLOCK(&mca_plm_rshd_component.lock);
}
/* actually ssh the child */
static void ssh_child(char *cmd, char **argv)
{
char** env;
char* var;
long fd, fdmax = sysconf(_SC_OPEN_MAX);
int fdin;
sigset_t sigs;
/* setup environment */
env = opal_argv_copy(orte_launch_environ);
/* Don't let ssh slurp all of our stdin! */
fdin = open("/dev/null", O_RDWR);
dup2(fdin, 0);
close(fdin);
/* close all file descriptors w/ exception of stdin/stdout/stderr */
for(fd=3; fd<fdmax; fd++)
close(fd);
/* Set signal handlers back to the default. Do this close
to the execve() because the event library may (and likely
will) reset them. If we don't do this, the event
library may have left some set that, at least on some
OS's, don't get reset via fork() or exec(). Hence, the
orted could be unkillable (for example). */
set_handler_default(SIGTERM);
set_handler_default(SIGINT);
set_handler_default(SIGHUP);
set_handler_default(SIGPIPE);
set_handler_default(SIGCHLD);
/* Unblock all signals, for many of the same reasons that
we set the default handlers, above. This is noticable
on Linux where the event library blocks SIGTERM, but we
don't want that blocked by the orted (or, more
specifically, we don't want it to be blocked by the
orted and then inherited by the ORTE processes that it
forks, making them unkillable by SIGTERM). */
sigprocmask(0, 0, &sigs);
sigprocmask(SIG_UNBLOCK, &sigs, 0);
/* exec the cmd */
var = opal_argv_join(argv, ' ');
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:rshd: executing: (%s) [%s]",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
cmd, (NULL == var) ? "NULL" : var));
if (NULL != var) free(var);
execve(cmd, argv, env);
opal_output(0, "plm:rshd: execv of %s failed with errno=%s(%d)\n",
cmd, strerror(errno), errno);
exit(-1);
}
/**
* Directly launch each specified process.
*/
/* When working in this function, ALWAYS jump to "cleanup" if
* you encounter an error so that orterun will be woken up and
* the job can cleanly terminate
*/
int orte_plm_rshd_launch(orte_job_t *jdata)
{
char **argv = NULL;
char *cmd, *param;
int rc, i;
bool failed_launch = true;
orte_app_context_t *app;
orte_node_t *node;
orte_proc_t *proc;
orte_jobid_t failed_job = ORTE_JOBID_INVALID;
orte_job_state_t job_state = ORTE_JOB_STATE_NEVER_LAUNCHED;
pid_t pid;
if (jdata->controls & ORTE_JOB_CONTROL_LOCAL_SLAVE) {
/* if this is a request to launch a local slave,
* then we will not be launching an orted - we will
* directly ssh the slave process itself. No mapping
* is performed to support this - the caller must
* provide all the info required to launch the job,
* including the target hosts
*/
return orte_plm_base_local_slave_launch(jdata);
}
/* setup the job */
if (ORTE_SUCCESS != (rc = orte_plm_base_setup_job(jdata))) {
ORTE_ERROR_LOG(rc);
goto cleanup;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:rshd: launching job %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_JOBID_PRINT(jdata->jobid)));
/* default to declaring the job launch as having failed */
failed_job = jdata->jobid;
/* launch each proc */
for (i=0; i < jdata->procs->size; i++) {
if (NULL == (proc = (orte_proc_t*)opal_pointer_array_get_item(jdata->procs, i))) {
continue;
}
/* only launch this proc if it isn't already running */
if (ORTE_PROC_STATE_LAUNCHED <= proc->state) {
continue;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:rshd: launching proc %s on node %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(&proc->name), proc->nodename));
if (NULL == (app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, proc->app_idx))) {
continue;
}
node = (orte_node_t*)proc->node;
/* setup the launch */
if (ORTE_SUCCESS != (rc = orte_plm_base_setup_slave_launch(proc->nodename, app,
"orte-bootproxy.sh",
&argv, &cmd))) {
ORTE_ERROR_LOG(rc);
goto cleanup;
}
/* add the bootproxy cmd line options */
if (ORTE_SUCCESS != (rc = orte_plm_base_append_bootproxy_args(app, &argv,
proc->name.jobid, proc->name.vpid,
jdata->map->num_nodes, jdata->num_procs,
proc->node_rank, proc->local_rank,
node->num_procs, jdata->total_slots_alloc,
false))) {
ORTE_ERROR_LOG(rc);
goto cleanup;
}
/* final cmd */
if (0 < opal_output_get_verbosity(orte_plm_globals.output)) {
param = opal_argv_join(argv, ' ');
opal_output(0, "%s plm:rshd: final cmd:\n\t%s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == param) ? "NULL" : param);
if (NULL != param) free(param);
}
/* fork a child to exec the rsh/ssh session */
pid = fork();
if (pid < 0) {
ORTE_ERROR_LOG(ORTE_ERR_SYS_LIMITS_CHILDREN);
rc = ORTE_ERR_SYS_LIMITS_CHILDREN;
goto cleanup;
}
/* child */
if (pid == 0) {
/* do the ssh launch - this will exit if it fails */
ssh_child(cmd, argv);
}
/* father */
/* declare the child launched */
proc->state = ORTE_PROC_STATE_LAUNCHED;
/* track number launched */
OPAL_THREAD_LOCK(&mca_plm_rshd_component.lock);
if (mca_plm_rshd_component.num_children++ >=
mca_plm_rshd_component.num_concurrent) {
opal_condition_wait(&mca_plm_rshd_component.cond, &mca_plm_rshd_component.lock);
}
OPAL_THREAD_UNLOCK(&mca_plm_rshd_component.lock);
/* cleanup */
opal_argv_free(argv);
argv = NULL;
free(cmd);
cmd = NULL;
/* setup callback on sigchild - wait until setup above is complete
* as the callback can occur in the call to orte_wait_cb
*/
orte_wait_cb(pid, wait_cb, (void*)proc);
}
/* flag the launch as successful */
failed_launch = false;
if (jdata->state < ORTE_JOB_STATE_UNTERMINATED) {
jdata->state = ORTE_JOB_STATE_LAUNCHED;
}
cleanup:
if (NULL != argv) {
opal_argv_free(argv);
}
if (NULL != cmd) {
free(cmd);
}
/* check for failed launch - if so, force terminate */
if (failed_launch) {
orte_errmgr.update_state(failed_job, job_state,
NULL, ORTE_PROC_STATE_UNDEF,
0, ORTE_ERROR_DEFAULT_EXIT_CODE);
}
return rc;
}
/**
* Terminate all processes for a given job
*/
int orte_plm_rshd_terminate_job(orte_jobid_t jobid)
{
return ORTE_ERR_NOT_IMPLEMENTED;
}
/**
* No orteds to terminate
*/
int orte_plm_rshd_terminate_orteds(void)
{
orte_quit();
return ORTE_SUCCESS;
}
int orte_plm_rshd_signal_job(orte_jobid_t jobid, int32_t signal)
{
/* no way to do this */
return ORTE_SUCCESS;
}
int orte_plm_rshd_finalize(void)
{
int rc;
/* cleanup any pending recvs */
if (ORTE_SUCCESS != (rc = orte_plm_base_comm_stop())) {
ORTE_ERROR_LOG(rc);
}
return rc;
}
static void set_handler_default(int sig)
{
struct sigaction act;
act.sa_handler = SIG_DFL;
act.sa_flags = 0;
sigemptyset(&act.sa_mask);
sigaction(sig, &act, (struct sigaction *)0);
}

0
orte/mca/plm/tmd/.ompi_ignore
Просмотреть файл

1
orte/mca/plm/tmd/.ompi_unignore
Просмотреть файл

@ -1 +0,0 @@
rhc

53
orte/mca/plm/tmd/Makefile.am
Просмотреть файл

@ -1,53 +0,0 @@
#
# 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 (c) 2010 Cisco Systems, Inc. All rights reserved.
# $COPYRIGHT$
#
# Additional copyrights may follow
#
# $HEADER$
#
AM_CPPFLAGS = $(plm_tmd_CPPFLAGS)
dist_pkgdata_DATA = help-plm-tmd.txt
sources = \
plm_tmd.h \
plm_tmd_component.c \
plm_tmd_module.c
# Make the output library in this directory, and name it either
# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
# (for static builds).
if MCA_BUILD_orte_plm_tmd_DSO
lib =
lib_sources =
component = mca_plm_tmd.la
component_sources = $(sources)
else
lib = libmca_plm_tmd.la
lib_sources = $(sources)
component =
component_sources =
endif
mcacomponentdir = $(pkglibdir)
mcacomponent_LTLIBRARIES = $(component)
mca_plm_tmd_la_SOURCES = $(component_sources)
mca_plm_tmd_la_LDFLAGS = -module -avoid-version $(plm_tmd_LDFLAGS)
noinst_LTLIBRARIES = $(lib)
libmca_plm_tmd_la_SOURCES = $(lib_sources)
libmca_plm_tmd_la_LDFLAGS = -module -avoid-version $(plm_tmd_LDFLAGS)
libmca_plm_tmd_la_LIBADD = $(plm_tmd_LIBS)

40
orte/mca/plm/tmd/configure.m4
Просмотреть файл

@ -1,40 +0,0 @@
# -*- shell-script -*-
#
# 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 (c) 2009-2010 Cisco Systems, Inc. All rights reserved.
# $COPYRIGHT$
#
# Additional copyrights may follow
#
# $HEADER$
#
# MCA_plm_tmd_CONFIG([action-if-found], [action-if-not-found])
# -----------------------------------------------------------
AC_DEFUN([MCA_orte_plm_tmd_CONFIG],[
AC_CONFIG_FILES([orte/mca/plm/tmd/Makefile])
ORTE_CHECK_TM([plm_tmd], [plm_tmd_good=1], [plm_tmd_good=0])
# if check worked, set wrapper flags if so.
# Evaluate succeed / fail
AS_IF([test "$plm_tmd_good" = "1"],
[plm_tmd_WRAPPER_EXTRA_LDFLAGS="$plm_tmd_LDFLAGS"
plm_tmd_WRAPPER_EXTRA_LIBS="$plm_tmd_LIBS"
$1],
[$2])
# set build flags to use in makefile
AC_SUBST([plm_tmd_CPPFLAGS])
AC_SUBST([plm_tmd_LDFLAGS])
AC_SUBST([plm_tmd_LIBS])
])dnl

52
orte/mca/plm/tmd/help-plm-tmd.txt
Просмотреть файл

@ -1,52 +0,0 @@
# -*- text -*-
#
# 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
#
# $HEADER$
#
[tmd-bad-launchid]
The TM (PBS / Torque) process starter cannot spawn the specified
application on a remote node due to an invalid launch_id.
This is most likely due to use of the "--hostfile" option to the
command line with one or more hosts in that file not having
been allocated to this job.
Removing "--hostfile" from the command line will likely allow the
application to be launched.
#
[multiple-prefixes]
Multiple different --prefix options were specified to mpirun for the
same node. This is a fatal error for the TM (PBS / Torque) process
starter in Open MPI.
The first two prefix values supplied for node %s were:
%s
and %s
#
[tmd-spawn-failed]
The TM (PBS / Torque) process starter failed to spawn a daemon (orted)
on a remote node.
Command line: %s
Node name: %s
Launch id: %d
If you do not understand this error mesage, please try the following:
1. Ensure that the executable "orted" is in your PATH
2. Use the --prefix option to indicate where we can
find that executable
3. Talk to your local system administrator

43
orte/mca/plm/tmd/plm_tmd.h
Просмотреть файл

@ -1,43 +0,0 @@
/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 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 (c) 2006 Cisco Systems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#ifndef ORTE_PLM_TMD_EXPORT_H
#define ORTE_PLM_TMD_EXPORT_H
#include "orte_config.h"
#include "opal/mca/mca.h"
#include "orte/mca/plm/plm.h"
BEGIN_C_DECLS
struct orte_plm_tmd_component_t {
orte_plm_base_component_t super;
bool want_path_check;
char **checked_paths;
};
typedef struct orte_plm_tmd_component_t orte_plm_tmd_component_t;
/* Globally exported variables */
ORTE_DECLSPEC extern orte_plm_tmd_component_t mca_plm_tmd_component;
extern orte_plm_base_module_t orte_plm_tmd_module;
END_C_DECLS
#endif /* ORTE_PLM_TMD_EXPORT_H */

128
orte/mca/plm/tmd/plm_tmd_component.c
Просмотреть файл

@ -1,128 +0,0 @@
/*
* Copyright (c) 2004-2008 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 (c) 2006 Cisco Systems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*
* These symbols are in a file by themselves to provide nice linker
* semantics. Since linkers generally pull in symbols by object
* files, keeping these symbols as the only symbols in this file
* prevents utility programs such as "ompi_info" from having to import
* entire components just to query their version and parameters.
*/
#include "orte_config.h"
#include "orte/constants.h"
#include "opal/mca/base/mca_base_param.h"
#include "opal/util/argv.h"
#include "orte/mca/plm/plm.h"
#include "orte/mca/plm/base/base.h"
#include "orte/mca/plm/base/plm_private.h"
#include "plm_tmd.h"
/*
* Public string showing the plm ompi_tm component version number
*/
const char *mca_plm_tmd_component_version_string =
"Open MPI tmd plm MCA component version " ORTE_VERSION;
/*
* Local function
*/
static int plm_tmd_open(void);
static int plm_tmd_close(void);
static int orte_plm_tmd_component_query(mca_base_module_t **module, int *priority);
/*
* Instantiate the public struct with all of our public information
* and pointers to our public functions in it
*/
orte_plm_tmd_component_t mca_plm_tmd_component = {
{
/* First, the mca_component_t struct containing meta information
about the component itself */
{
ORTE_PLM_BASE_VERSION_2_0_0,
/* Component name and version */
"tmd",
ORTE_MAJOR_VERSION,
ORTE_MINOR_VERSION,
ORTE_RELEASE_VERSION,
/* Component open and close functions */
plm_tmd_open,
plm_tmd_close,
orte_plm_tmd_component_query
},
{
/* The component is checkpoint ready */
MCA_BASE_METADATA_PARAM_CHECKPOINT
}
}
};
static int plm_tmd_open(void)
{
int tmp;
mca_base_component_t *comp = &mca_plm_tmd_component.super.base_version;
mca_base_param_reg_int(comp, "want_path_check",
"Whether the launching process should check for the plm_tmd_orted executable in the PATH before launching (the TM API does not give an indication of failure; this is a somewhat-lame workaround; non-zero values enable this check)",
false, false, (int) true, &tmp);
mca_plm_tmd_component.want_path_check = OPAL_INT_TO_BOOL(tmp);
mca_plm_tmd_component.checked_paths = NULL;
return ORTE_SUCCESS;
}
static int plm_tmd_close(void)
{
if (NULL != mca_plm_tmd_component.checked_paths) {
opal_argv_free(mca_plm_tmd_component.checked_paths);
}
return ORTE_SUCCESS;
}
static int orte_plm_tmd_component_query(mca_base_module_t **module, int *priority)
{
/* Are we running under a TM job? */
if (NULL != getenv("PBS_ENVIRONMENT") &&
NULL != getenv("PBS_JOBID")) {
*priority = 2;
*module = (mca_base_module_t *) &orte_plm_tmd_module;
return ORTE_SUCCESS;
}
/* Sadly, no */
*module = NULL;
return ORTE_ERROR;
}

834
orte/mca/plm/tmd/plm_tmd_module.c
Просмотреть файл

@ -1,834 +0,0 @@
/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2006 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 (c) 2006 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2007 Los Alamos National Security, LLC. All rights
* reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*
* These symbols are in a file by themselves to provide nice linker
* semantics. Since linkers generally pull in symbols by object
* files, keeping these symbols as the only symbols in this file
* prevents utility programs such as "ompi_info" from having to import
* entire components just to query their version and parameters.
*/
#include "orte_config.h"
#include "orte/constants.h"
#include "orte/types.h"
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif
#include <signal.h>
#ifdef HAVE_SYS_TYPES_H
#include <sys/types.h>
#endif
#ifdef HAVE_SYS_STAT_H
#include <sys/stat.h>
#endif
#ifdef HAVE_SYS_WAIT_H
#include <sys/wait.h>
#endif
#ifdef HAVE_SCHED_H
#include <sched.h>
#endif
#ifdef HAVE_SYS_TIME_H
#include <sys/time.h>
#endif
#include <errno.h>
#include <string.h>
#include <tm.h>
#include "opal/mca/installdirs/installdirs.h"
#include "opal/mca/event/event.h"
#include "opal/util/argv.h"
#include "opal/util/output.h"
#include "orte/util/show_help.h"
#include "opal/util/opal_environ.h"
#include "opal/util/basename.h"
#include "opal/mca/base/mca_base_param.h"
#include "opal/runtime/opal_progress.h"
#include "orte/util/name_fns.h"
#include "orte/runtime/orte_globals.h"
#include "orte/runtime/orte_wait.h"
#include "orte/mca/errmgr/errmgr.h"
#include "orte/mca/rmaps/rmaps.h"
#include "orte/runtime/orte_quit.h"
#include "orte/mca/plm/plm.h"
#include "orte/mca/plm/base/plm_private.h"
#include "plm_tmd.h"
/*
* Local functions
*/
static int plm_tmd_init(void);
static int plm_tmd_launch_job(orte_job_t *jdata);
static int plm_tmd_terminate_job(orte_jobid_t jobid);
static int plm_tmd_terminate_orteds(void);
static int plm_tmd_signal_job(orte_jobid_t jobid, int32_t signal);
static int plm_tmd_finalize(void);
static int plm_tmd_connect(void);
static void failed_start(int fd, short event, void *arg);
static int obit_submit(int tid);
/*
* Local "global" variables
*/
static opal_event_t *ev=NULL;
static bool connected;
static tm_event_t *events_spawn = NULL;
static tm_event_t *events_obit = NULL;
static tm_task_id *tm_task_ids = NULL;
static int *evs = NULL;
static bool time_is_up;
/*
* Global variable
*/
orte_plm_base_module_t orte_plm_tmd_module = {
plm_tmd_init,
orte_plm_base_set_hnp_name,
plm_tmd_launch_job,
NULL,
plm_tmd_terminate_job,
plm_tmd_terminate_orteds,
NULL,
plm_tmd_signal_job,
plm_tmd_finalize
};
/* catch timeout to allow cmds to progress */
static void timer_cb(int fd, short event, void *cbdata)
{
opal_event_t *ev = (opal_event_t*)cbdata;
/* free event */
if (NULL != ev) {
free(ev);
}
/* declare time is up */
time_is_up = true;
}
/**
* Init the module
*/
static int plm_tmd_init(void)
{
int rc;
if (ORTE_SUCCESS != (rc = orte_plm_base_comm_start())) {
ORTE_ERROR_LOG(rc);
}
return rc;
}
/* When working in this function, ALWAYS jump to "cleanup" if
* you encounter an error so that orterun will be woken up and
* the job can cleanly terminate
*/
static int plm_tmd_launch_job(orte_job_t *jdata)
{
orte_job_t *jdatorted;
orte_job_map_t *map = NULL;
orte_app_context_t **apps;
orte_node_t **nodes;
int proc_vpid_index;
char *param;
char **env = NULL;
char *var;
char **argv = NULL;
int argc = 0;
int rc;
orte_std_cntr_t launched = 0, i;
char *bin_base = NULL, *lib_base = NULL;
int local_err;
bool failed_launch = true;
mode_t current_umask;
orte_jobid_t failed_job;
orte_job_state_t job_state = ORTE_JOB_STATE_NEVER_LAUNCHED;
int offset;
tm_event_t eventpolled;
orte_std_cntr_t num_daemons;
opal_event_t *timerev;
int j;
/* default to declaring the daemons as failed */
failed_job = ORTE_PROC_MY_NAME->jobid;
connected = false;
/* setup the job */
if (ORTE_SUCCESS != (rc = orte_plm_base_setup_job(jdata))) {
ORTE_ERROR_LOG(rc);
goto cleanup;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: launching job %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_JOBID_PRINT(jdata->jobid)));
/* Get the map for this job */
if (NULL == (map = orte_rmaps.get_job_map(jdata->jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
rc = ORTE_ERR_NOT_FOUND;
goto cleanup;
}
apps = (orte_app_context_t**)jdata->apps->addr;
nodes = (orte_node_t**)map->nodes->addr;
if (0 == map->num_new_daemons) {
/* have all the daemons we need - launch app */
goto launch_apps;
}
/* lookup the daemon job object - must do this -after- the job is
* setup so the number of required daemons has been updated
*/
if (NULL == (jdatorted = orte_get_job_data_object(ORTE_PROC_MY_NAME->jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
rc = ORTE_ERR_NOT_FOUND;
goto cleanup;
}
num_daemons = jdatorted->num_procs - 1; /* do not include myself as I am already here! */
if (0 >= num_daemons) {
/* this won't work */
rc = ORTE_ERR_BAD_PARAM;
goto cleanup;
}
/* Allocate a bunch of TM events to use */
if (NULL == events_spawn) {
/* spawn events for first launch */
events_spawn = (tm_event_t*)malloc(num_daemons * sizeof(tm_event_t));
if (NULL == events_spawn) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
} else {
/* comm_spawn launch */
events_spawn = (tm_event_t*)realloc(events_spawn, sizeof(tm_event_t) * num_daemons);
if (NULL == events_spawn) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
}
if (NULL == events_obit) {
/* obit events for first launch */
events_obit = (tm_event_t*)malloc(num_daemons * sizeof(tm_event_t));
if (NULL == events_obit) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
} else {
/* comm_spawn launch */
events_obit = (tm_event_t*)realloc(events_obit, sizeof(tm_event_t) * num_daemons);
if (NULL == events_obit) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
}
if (NULL == evs) {
/* evs for first launch */
evs = (int*)malloc(num_daemons * sizeof(tm_event_t));
if (NULL == evs) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
} else {
/* comm_spawn launch */
evs = (int*)realloc(evs, sizeof(int) * num_daemons);
if (NULL == evs) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
}
/* allocate task ids for the orteds */
if (NULL == tm_task_ids) {
/* first launch */
tm_task_ids = (tm_task_id*)malloc(num_daemons * sizeof(tm_task_id));
if (NULL == tm_task_ids) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
} else {
/* comm_spawn launch */
tm_task_ids = (tm_task_id*)realloc(tm_task_ids, sizeof(tm_task_id) * num_daemons);
if (NULL == tm_task_ids) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
}
/* compute the offset into the event/task arrays */
offset = num_daemons - map->num_new_daemons;
/* initialize them */
for (i=0; i < map->num_new_daemons; i++) {
*(tm_task_ids + offset + i) = TM_NULL_TASK;
*(events_spawn + offset + i) = TM_NULL_EVENT;
*(events_obit + offset + i) = TM_NULL_EVENT;
*(evs + offset + i) = 0;
}
/* add the daemon command (as specified by user) */
orte_plm_base_setup_orted_cmd(&argc, &argv);
/* Add basic orted command line options */
orte_plm_base_orted_append_basic_args(&argc, &argv, "env",
&proc_vpid_index,
true, NULL);
if (0 < opal_output_get_verbosity(orte_plm_globals.output)) {
param = opal_argv_join(argv, ' ');
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: final top-level argv:\n\t%s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == param) ? "NULL" : param));
if (NULL != param) free(param);
}
rc = plm_tmd_connect();
if (ORTE_SUCCESS != rc) {
goto cleanup;
}
connected = true;
/* Figure out the basenames for the libdir and bindir. There is a
lengthy comment about this in plm_rsh_module.c explaining all
the rationale for how / why we're doing this. */
lib_base = opal_basename(opal_install_dirs.libdir);
bin_base = opal_basename(opal_install_dirs.bindir);
/* setup environment */
env = opal_argv_copy(orte_launch_environ);
/* add our umask -- see big note in orted.c */
current_umask = umask(0);
umask(current_umask);
asprintf(&var, "0%o", current_umask);
opal_setenv("ORTE_DAEMON_UMASK_VALUE", var, true, &env);
free(var);
/* If we have a prefix, then modify the PATH and
LD_LIBRARY_PATH environment variables. We only allow
a single prefix to be specified. Since there will
always be at least one app_context, we take it from
there
*/
if (NULL != apps[0]->prefix_dir) {
char *newenv;
for (i = 0; NULL != env && NULL != env[i]; ++i) {
/* Reset PATH */
if (0 == strncmp("PATH=", env[i], 5)) {
asprintf(&newenv, "%s/%s:%s",
apps[0]->prefix_dir, bin_base, env[i] + 5);
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: resetting PATH: %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
newenv));
opal_setenv("PATH", newenv, true, &env);
free(newenv);
}
/* Reset LD_LIBRARY_PATH */
else if (0 == strncmp("LD_LIBRARY_PATH=", env[i], 16)) {
asprintf(&newenv, "%s/%s:%s",
apps[0]->prefix_dir, lib_base, env[i] + 16);
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: resetting LD_LIBRARY_PATH: %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
newenv));
opal_setenv("LD_LIBRARY_PATH", newenv, true, &env);
free(newenv);
}
}
}
/* set the job state to indicate we attempted to launch */
job_state = ORTE_JOB_STATE_FAILED_TO_START;
/* Iterate through each of the nodes and spin
* up a daemon.
*/
for (i = 0; i < map->num_nodes; i++) {
orte_node_t* node = nodes[i];
char* vpid_string;
/* if this daemon already exists, don't launch it! */
if (node->daemon_launched) {
continue;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: launching on node %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
node->name));
/* setup process name */
rc = orte_util_convert_vpid_to_string(&vpid_string, nodes[i]->daemon->name.vpid);
if (ORTE_SUCCESS != rc) {
opal_output(0, "plm:tm: unable to get daemon vpid as string");
goto cleanup;
}
free(argv[proc_vpid_index]);
argv[proc_vpid_index] = strdup(vpid_string);
free(vpid_string);
/* exec the daemon */
if (0 < opal_output_get_verbosity(orte_plm_globals.output)) {
param = opal_argv_join(argv, ' ');
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: executing:\n\t%s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == param) ? "NULL" : param));
if (NULL != param) free(param);
}
rc = tm_spawn(argc, argv, env, node->launch_id, tm_task_ids + offset + launched, events_spawn + offset + launched);
if (TM_SUCCESS != rc) {
orte_show_help("help-plm-tm.txt", "tm-spawn-failed",
true, argv[0], node->name, node->launch_id);
rc = ORTE_ERROR;
goto cleanup;
}
launched++;
/* Allow some progress to occur */
opal_event_loop(opal_event_base, OPAL_EVLOOP_NONBLOCK);
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm:launch: finished spawning orteds",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* setup a timer to give the cmd a chance to be sent */
time_is_up = false;
ORTE_DETECT_TIMEOUT(&timerev, launched,
100, -1, timer_cb);
ORTE_PROGRESSED_WAIT(time_is_up, 0, 1);
/* TM poll for all the spawns */
while (0 < launched) {
rc = tm_poll(TM_NULL_EVENT, &eventpolled, (int)false, &local_err);
if (TM_SUCCESS != rc) {
opal_output(0, "plm:tm: event poll for spawned daemon failed, return status = %d", rc);
rc = ORTE_ERROR;
goto cleanup;
}
/* if we get back the NULL event, then just continue */
if (eventpolled == TM_NULL_EVENT) {
continue;
}
/* look for the spawned event */
for (j=0; j < map->num_new_daemons; j++) {
if (eventpolled == *(events_spawn + offset + j)) {
/* got the event - check returned code */
if (local_err) {
/* this orted failed to launch! */
orte_show_help("help-plm-tm.txt", "tm-spawn-failed",
true, argv[0], nodes[j]->name, nodes[j]->launch_id);
rc = ORTE_ERROR;
goto cleanup;
}
/* register the corresponding obit so we can detect when this
* orted terminates
*/
if (ORTE_SUCCESS != (rc = obit_submit(offset+j))) {
ORTE_ERROR_LOG(rc);
goto cleanup;
}
/* all done with this event */
goto MOVEON;
}
}
/* if we get here, then we failed to find the event */
opal_output(0, "TM FAILED TO FIND SPAWN EVENT WHEN LAUNCHING");
rc = ORTE_ERROR;
goto cleanup;
MOVEON:
launched--;
}
/* set a timer to tell us if one or more daemon's fails to start - use the
* millisec/daemon timeout provided by the user to compute time
*/
if (0 < orte_startup_timeout) {
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: setting startup timer for %d milliseconds",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
orte_startup_timeout));
ORTE_DETECT_TIMEOUT(&ev, map->num_new_daemons,
orte_startup_timeout*1000,
-1, failed_start);
}
/* wait for daemons to callback */
if (ORTE_SUCCESS != (rc = orte_plm_base_daemon_callback(map->num_new_daemons))) {
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: daemon launch failed for job %s on error %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_JOBID_PRINT(jdata->jobid), ORTE_ERROR_NAME(rc)));
goto cleanup;
}
/* if issued, cancel the failed-to-start timer */
if (NULL != ev) {
opal_event_del(ev);
}
launch_apps:
/* since the daemons have launched, any failures now will be for the
* application job
*/
failed_job = jdata->jobid;
if (ORTE_SUCCESS != (rc = orte_plm_base_launch_apps(jdata->jobid))) {
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: launch of apps failed for job %s on error %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_JOBID_PRINT(jdata->jobid), ORTE_ERROR_NAME(rc)));
goto cleanup;
}
/* if we get here, then everything launched okay - record that fact */
failed_launch = false;
cleanup:
if (NULL != argv) {
opal_argv_free(argv);
}
if (NULL != env) {
opal_argv_free(env);
}
if (NULL != lib_base) {
free(lib_base);
}
if (NULL != bin_base) {
free(bin_base);
}
/* check for failed launch - if so, force terminate */
if (failed_launch) {
orte_errmgr.update_state(failed_job, job_state,
NULL, ORTE_PROC_STATE_UNDEF,
0, ORTE_ERROR_DEFAULT_EXIT_CODE);
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm:launch: finished",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
return rc;
}
static int plm_tmd_terminate_job(orte_jobid_t jobid)
{
return ORTE_ERR_NOT_IMPLEMENTED;
}
/* quick timeout loop */
static bool timer_fired;
static void quicktime_cb(int fd, short event, void *cbdata)
{
/* declare it fired */
timer_fired = true;
}
/**
* Terminate the orteds for a given job
*/
int plm_tmd_terminate_orteds(void)
{
int rc;
orte_job_t *jdata;
orte_proc_t **daemons;
tm_event_t eventpolled;
orte_vpid_t j, alive;
int local_err;
opal_event_t *timerev=NULL;
opal_event_t *quicktime=NULL;
struct timeval quicktimeval;
bool aborted;
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: terminating orteds",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* lookup the daemon job object */
if (NULL == (jdata = orte_get_job_data_object(ORTE_PROC_MY_NAME->jobid))) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
}
alive = jdata->num_procs - 1; /* do not include myself! */
daemons = (orte_proc_t**)jdata->procs->addr;
aborted = false;
/* tell them to die! */
if (ORTE_SUCCESS != (rc = orte_plm_base_orted_exit(ORTE_DAEMON_EXIT_CMD))) {
ORTE_ERROR_LOG(rc);
}
/* if there are more than just me... */
if (0 < alive) {
/* setup a max time for the daemons to die */
time_is_up = false;
ORTE_DETECT_TIMEOUT(&timerev, alive,
1000000, 60000000, timer_cb);
/* give the cmds a chance to get out */
quicktimeval.tv_sec = 0;
quicktimeval.tv_usec = 100;
timer_fired = false;
ORTE_DETECT_TIMEOUT(&quicktime, alive, 1000, 10000, quicktime_cb);
ORTE_PROGRESSED_WAIT(timer_fired, 0, 1);
/* now begin polling to see if daemons have terminated */
while (!time_is_up && 0 < alive) {
OPAL_OUTPUT_VERBOSE((10, orte_plm_globals.output,
"%s plm:tm: polling for daemon termination",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
rc = tm_poll(TM_NULL_EVENT, &eventpolled, (int)false, &local_err);
if (TM_SUCCESS != rc) {
errno = local_err;
opal_output(0, "plm:tm: event poll for daemon termination failed, return status = %d", rc);
continue; /* we will wait for timeout to tell us to quit */
}
/* if we get back the NULL event, then just continue */
if (eventpolled == TM_NULL_EVENT) {
OPAL_OUTPUT_VERBOSE((10, orte_plm_globals.output,
"%s plm:tm: got null event",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* give system a little time to progress */
timer_fired = false;
opal_event_evtimer_add(quicktime, &quicktimeval);
ORTE_PROGRESSED_WAIT(timer_fired, 0, 1);
continue;
}
/* look for the obit event */
for (j=0; j < jdata->num_procs-1; j++) {
if (eventpolled == *(events_obit + j)) {
/* got the event - check returned code */
if (local_err == TM_ESYSTEM) {
OPAL_OUTPUT_VERBOSE((10, orte_plm_globals.output,
"%s plm:tm: got TM_ESYSTEM on obit - resubmitting",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
if (ORTE_SUCCESS != (rc = obit_submit(j))) {
ORTE_ERROR_LOG(rc);
goto MOVEON;
}
/* give system a little time to progress */
timer_fired = false;
opal_event_evtimer_add(quicktime, &quicktimeval);
ORTE_PROGRESSED_WAIT(timer_fired, 0, 1);
}
if (0 != local_err) {
OPAL_OUTPUT_VERBOSE((10, orte_plm_globals.output,
"%s plm:tm: got error %d on obit for task %d",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), local_err, j));
rc = ORTE_ERROR;
goto MOVEON;
}
/* this daemon has terminated */
*(tm_task_ids+j) = TM_NULL_TASK;
*(events_obit+j) = TM_NULL_EVENT;
OPAL_OUTPUT_VERBOSE((10, orte_plm_globals.output,
"%s plm:tm: task %d exited with status %d",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), j, *(evs+j)));
/* update the termination status for this daemon */
daemons[j+1]->exit_code = *(evs+j);
if (0 != daemons[j+1]->exit_code) {
daemons[j+1]->state = ORTE_PROC_STATE_ABORTED;
aborted = true;
} else {
daemons[j+1]->state = ORTE_PROC_STATE_TERMINATED;
}
jdata->num_terminated++;
/* all done with this event */
goto MOVEON;
}
}
/* if we get here, then we failed to find the event */
opal_output(0, "TM FAILED TO FIND OBIT EVENT");
MOVEON:
alive--;
}
/* release event if not already done */
if (NULL != quicktime) {
free(quicktime);
}
if (NULL != timerev) {
opal_event_del(timerev);
free(timerev);
}
} else {
/* still need to give the cmds a chance to get out so I can process
* them myself!
*/
timer_fired = false;
ORTE_DETECT_TIMEOUT(&quicktime, 1, 1000, 10000, quicktime_cb);
ORTE_PROGRESSED_WAIT(timer_fired, 0, 1);
}
/* declare the daemons done */
if (aborted || 0 < alive) {
jdata->state = ORTE_JOB_STATE_ABORTED;
} else {
jdata->state = ORTE_JOB_STATE_TERMINATED;
}
orte_quit();
return rc;
}
static int plm_tmd_signal_job(orte_jobid_t jobid, int32_t signal)
{
int rc;
/* order them to pass this signal to their local procs */
if (ORTE_SUCCESS != (rc = orte_plm_base_orted_signal_local_procs(jobid, signal))) {
ORTE_ERROR_LOG(rc);
}
return rc;
}
/*
* Free stuff
*/
static int plm_tmd_finalize(void)
{
int rc;
/* cleanup any pending recvs */
if (ORTE_SUCCESS != (rc = orte_plm_base_comm_stop())) {
ORTE_ERROR_LOG(rc);
}
if (connected) {
tm_finalize();
}
/* cleanup data arrays */
if (NULL != events_spawn) {
free(events_spawn);
}
if (NULL != events_obit) {
free(events_obit);
}
if (NULL != tm_task_ids) {
free(tm_task_ids);
}
if (NULL != evs) {
free(evs);
}
return ORTE_SUCCESS;
}
static int plm_tmd_connect(void)
{
int ret;
struct tm_roots tm_root;
int count, progress;
/* try a couple times to connect - might get busy signals every
now and then */
for (count = 0 ; count < 10; ++count) {
ret = tm_init(NULL, &tm_root);
if (TM_SUCCESS == ret) {
return ORTE_SUCCESS;
}
for (progress = 0 ; progress < 10 ; ++progress) {
opal_progress();
#if HAVE_SCHED_YIELD
sched_yield();
#endif
}
}
return ORTE_ERR_RESOURCE_BUSY;
}
/* call this function if the timer fires indicating that one
* or more daemons failed to start
*/
static void failed_start(int fd, short dummy, void *arg)
{
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm:failed_start",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* if we are aborting, ignore this */
if (orte_abnormal_term_ordered) {
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm:failed_start - abnormal term in progress",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
return;
}
orte_errmgr.update_state(ORTE_PROC_MY_NAME->jobid, ORTE_JOB_STATE_FAILED_TO_START,
NULL, ORTE_PROC_STATE_UNDEF, 0, ORTE_ERROR_DEFAULT_EXIT_CODE);
}
static int obit_submit(int tid)
{
int rc;
if (TM_SUCCESS != (rc = tm_obit(*(tm_task_ids+tid), evs+tid, events_obit+tid))) {
opal_output(0, "failed to register termination notice for task %d", tid);
rc = ORTE_ERROR;
return rc;
}
if (*(events_obit+tid) == TM_NULL_EVENT) {
opal_output(0, "task %d is already dead", tid);
} else if (*(events_obit+tid) == TM_ERROR_EVENT) {
opal_output(0, "Error on obit return - got error event for task %d", tid);
}
return ORTE_SUCCESS;
}