/*
* This utility will do a brute force clean of a node. It will
* attempt to clean up any files in the user's session directory.
* It will also look for any orted and orterun processes that are
* not part of this job, and kill them off.
*/
int
main(int argc, char *argv[])
{
int ret = ORTE_SUCCESS;
#if OPAL_ENABLE_FT_CR == 1
char *tmp_env_var;
#endif
/* This is needed so we can print the help message */
if (ORTE_SUCCESS != (ret = opal_init_util(&argc, &argv))) {
return ret;
}
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
return ret;
}
#if OPAL_ENABLE_FT_CR == 1
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1", true, NULL);
free(tmp_env_var);
#endif
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_TOOL))) {
return ret;
}
/*
* Clean out all session directories - we don't have to protect
* our own session directory because (since we are a tool) we
* didn't create one!
*/
if (orte_clean_globals.verbose) {
fprintf(stderr, "orte-clean: cleaning session dir tree %s\n",
orte_process_info.top_session_dir);
}
opal_os_dirpath_destroy(orte_process_info.top_session_dir, true, NULL);
/* now kill any lingering procs, if we can */
kill_procs();
orte_finalize();
return ORTE_SUCCESS;
}
int orte_ess_env_put(orte_std_cntr_t num_procs,
orte_std_cntr_t num_local_procs,
char ***env)
{
char* param;
char* value;
/* tell the ESS to select the PMI component - but don't override
* anything that may have been provided elsewhere
*/
if(OPAL_SUCCESS != mca_base_var_env_name ("ess", ¶m)) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERR_OUT_OF_RESOURCE;
}
opal_setenv(param, "pmi", false, env);
free(param);
/* since we want to pass the name as separate components, make sure
* that the "name" environmental variable is cleared!
*/
if(OPAL_SUCCESS != mca_base_var_env_name ("orte_ess_name", ¶m)) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERR_OUT_OF_RESOURCE;
}
opal_unsetenv(param, env);
free(param);
asprintf(&value, "%ld", (long) num_procs);
if(OPAL_SUCCESS != mca_base_var_env_name ("orte_ess_num_procs", ¶m)) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERR_OUT_OF_RESOURCE;
}
opal_setenv(param, value, true, env);
free(param);
/* although the num_procs is the comm_world size, users
* would appreciate being given a public environmental variable
* that also represents this value - something MPI specific - so
* do that here.
*
* AND YES - THIS BREAKS THE ABSTRACTION BARRIER TO SOME EXTENT.
* We know - just live with it
*/
opal_setenv("OMPI_COMM_WORLD_SIZE", value, true, env);
free(value);
/* users would appreciate being given a public environmental variable
* that also represents this value - something MPI specific - so
* do that here.
*
* AND YES - THIS BREAKS THE ABSTRACTION BARRIER TO SOME EXTENT.
* We know - just live with it
*/
asprintf(&value, "%ld", (long) num_local_procs);
opal_setenv("OMPI_COMM_WORLD_LOCAL_SIZE", value, true, env);
free(value);
return ORTE_SUCCESS;
}
int orte_pre_condition_transports(orte_job_t *jdata, char **key)
{
uint64_t unique_key[2];
int n;
orte_app_context_t *app;
char *string_key, *cs_env;
int fd_rand;
size_t bytes_read;
struct stat buf;
/* put the number here - or else create an appropriate string. this just needs to
* eventually be a string variable
*/
if(0 != stat("/dev/urandom", &buf)) {
/* file doesn't exist! */
orte_pre_condition_transports_use_rand(unique_key);
}
if(-1 == (fd_rand = open("/dev/urandom", O_RDONLY))) {
orte_pre_condition_transports_use_rand(unique_key);
} else {
bytes_read = read(fd_rand, (char *) unique_key, 16);
if(bytes_read != 16) {
orte_pre_condition_transports_use_rand(unique_key);
}
close(fd_rand);
}
if (NULL == (string_key = orte_pre_condition_transports_print(unique_key))) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERR_OUT_OF_RESOURCE;
}
/* record it in case this job executes a dynamic spawn */
if (NULL != jdata) {
orte_set_attribute(&jdata->attributes, ORTE_JOB_TRANSPORT_KEY, ORTE_ATTR_LOCAL, string_key, OPAL_STRING);
if (OPAL_SUCCESS != mca_base_var_env_name ("orte_precondition_transports", &cs_env)) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
free(string_key);
return ORTE_ERR_OUT_OF_RESOURCE;
}
for (n=0; n < jdata->apps->size; n++) {
if (NULL == (app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, n))) {
continue;
}
opal_setenv(cs_env, string_key, true, &app->env);
}
free(cs_env);
free(string_key);
} else if (NULL != key) {
*key = string_key;
} else {
free(string_key);
}
return ORTE_SUCCESS;
}
/*
* Self is a special case. The 'fname' here is the command line that the user
* wishes to execute. This function takes this command line and adds
* -mca crs_self_do_restart 1
* Which will trigger the restart callback once the program has been run.
*
* For example, The user starts their program with:
* $ my_prog arg1 arg2
*
* They checkpoint it:
* $ opal_checkpoint -mca crs self 1234
*
* They restart it:
* $ opal_restart -mca crs self my_prog arg1 arg2
*
* fname is then:
* fname = "my_prog arg1 arg2"
*
* This funciton translates that to the command:
* cmd = "my_prog arg1 arg2 -mca crs self -mca crs_self_do_restart 1"
*
* Which will cause the program "my_prog" to call their restart function
* upon opal_init time.
*
* Note: The user could bypass the opal_restart routine safely by simply calling
* $ my_prog arg1 arg2 -mca crs self -mca crs_self_do_restart 1
* However, for consistency sake, we should not encourage this as it won't work for
* all of the other checkpointers.
*/
static int opal_crs_self_restart_cmd(opal_crs_self_snapshot_t *snapshot, char **cmd)
{
char * tmp_env_var = NULL;
opal_output_verbose(10, mca_crs_self_component.super.output_handle,
"crs:self: restart_cmd(%s, ---)", snapshot->cmd_line);
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
"self",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
(void) mca_base_var_env_name("crs_self_do_restart", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
(void) mca_base_var_env_name("crs_self_prefix", &tmp_env_var);
opal_setenv(tmp_env_var,
mca_crs_self_component.prefix,
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/* Instead of adding it to the command line, we should use the environment
* to pass the values. This allow sthe OPAL application to be braindead
* WRT MCA parameters
* add_args = strdup("-mca crs self -mca crs_self_do_restart 1");
*/
asprintf(cmd, "%s", snapshot->cmd_line);
return OPAL_SUCCESS;
}
static void add_to_env(char **params, char **values, char ***env)
{
int i;
char *name;
/* Loop through all the args that we've gotten and make env
vars of the form OMPI_MCA_*=value. */
for (i = 0; NULL != params && NULL != params[i]; ++i) {
(void) mca_base_var_env_name (params[i], &name);
opal_setenv(name, values[i], true, env);
free(name);
}
}
int opal_cr_refresh_environ(int prev_pid) {
char *file_name;
#if OPAL_ENABLE_CRDEBUG == 1
char *tmp;
#endif
struct stat file_status;
if( 0 >= prev_pid ) {
prev_pid = getpid();
}
/*
* Make sure the file exists. If it doesn't then this means 2 things:
* 1) We have already executed this function, and
* 2) The file has been deleted on the previous round.
*/
asprintf(&file_name, "%s/%s-%d", opal_tmp_directory(), OPAL_CR_BASE_ENV_NAME, prev_pid);
if (NULL == file_name) {
return OPAL_ERR_OUT_OF_RESOURCE;
}
if(0 != stat(file_name, &file_status) ){
free(file_name);
return OPAL_SUCCESS;
}
#if OPAL_ENABLE_CRDEBUG == 1
mca_base_var_env_name ("opal_cr_enable_crdebug", &tmp);
opal_unsetenv(tmp, &environ);
free (tmp);
#endif
extract_env_vars(prev_pid, file_name);
#if OPAL_ENABLE_CRDEBUG == 1
MPIR_debug_with_checkpoint = 0;
(void) mca_base_var_register ("opal", "opal", "cr", "enable_crdebug",
"Enable checkpoint/restart debugging",
MCA_BASE_VAR_TYPE_INT, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_ALL_EQ,
&MPIR_debug_with_checkpoint);
opal_output_verbose(10, opal_cr_output,
"opal_cr: init: C/R Debugging Enabled [%s] (refresh)\n",
(MPIR_debug_with_checkpoint ? "True": "False"));
#endif
free(file_name);
return OPAL_SUCCESS;
}
int opal_crs_none_prelaunch(int32_t rank,
char *base_snapshot_dir,
char **app,
char **cwd,
char ***argv,
char ***env)
{
char * tmp_env_var = NULL;
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"0", true, env);
free(tmp_env_var);
tmp_env_var = NULL;
return OPAL_SUCCESS;
}
int opal_crs_self_prelaunch(int32_t rank,
char *base_snapshot_dir,
char **app,
char **cwd,
char ***argv,
char ***env)
{
char * tmp_env_var = NULL;
/*
* This function should never be called by a tool
*/
if( opal_cr_is_tool ) {
return OPAL_ERR_NOT_SUPPORTED;
}
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"0", true, env);
free(tmp_env_var);
tmp_env_var = NULL;
return OPAL_SUCCESS;
}
/* 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 void launch_daemons(int fd, short args, void *cbdata)
{
orte_job_map_t *map = NULL;
orte_app_context_t *app;
orte_node_t *node;
int proc_vpid_index;
char *param;
char **env = NULL;
char *var;
char **argv = NULL;
char **nodeargv;
int argc = 0;
int rc;
orte_std_cntr_t i;
char *bin_base = NULL, *lib_base = NULL;
tm_event_t *tm_events = NULL;
tm_task_id *tm_task_ids = NULL;
bool failed_launch = true;
mode_t current_umask;
char *nodelist;
char* vpid_string;
orte_job_t *daemons, *jdata;
orte_state_caddy_t *state = (orte_state_caddy_t*)cbdata;
int32_t launchid, *ldptr;
char *prefix_dir = NULL;
jdata = state->jdata;
/* if we are launching debugger daemons, then just go
* do it - no new daemons will be launched
*/
if (ORTE_FLAG_TEST(state->jdata, ORTE_JOB_FLAG_DEBUGGER_DAEMON)) {
jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_DAEMONS_REPORTED);
OBJ_RELEASE(state);
return;
}
/* setup the virtual machine */
daemons = orte_get_job_data_object(ORTE_PROC_MY_NAME->jobid);
if (ORTE_SUCCESS != (rc = orte_plm_base_setup_virtual_machine(jdata))) {
ORTE_ERROR_LOG(rc);
goto cleanup;
}
/* if we don't want to launch, then don't attempt to
* launch the daemons - the user really wants to just
* look at the proposed process map
*/
if (orte_do_not_launch) {
/* set the state to indicate the daemons reported - this
* will trigger the daemons_reported event and cause the
* job to move to the following step
*/
jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_DAEMONS_REPORTED);
OBJ_RELEASE(state);
return;
}
/* Get the map for this job */
if (NULL == (map = daemons->map)) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
rc = ORTE_ERR_NOT_FOUND;
goto cleanup;
}
if (0 == map->num_new_daemons) {
/* set the state to indicate the daemons reported - this
* will trigger the daemons_reported event and cause the
* job to move to the following step
*/
jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_DAEMONS_REPORTED);
OBJ_RELEASE(state);
return;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_base_framework.framework_output,
"%s plm:tm: launching vm",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* Allocate a bunch of TM events to use for tm_spawn()ing */
tm_events = malloc(sizeof(tm_event_t) * map->num_new_daemons);
if (NULL == tm_events) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
tm_task_ids = malloc(sizeof(tm_task_id) * map->num_new_daemons);
if (NULL == tm_task_ids) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
/* add the daemon command (as specified by user) */
orte_plm_base_setup_orted_cmd(&argc, &argv);
/* create a list of nodes in this launch */
nodeargv = NULL;
for (i = 0; i < map->nodes->size; i++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
/* if this daemon already exists, don't launch it! */
if (ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_DAEMON_LAUNCHED)) {
continue;
}
/* add to list */
opal_argv_append_nosize(&nodeargv, node->name);
}
nodelist = opal_argv_join(nodeargv, ',');
opal_argv_free(nodeargv);
/* Add basic orted command line options */
orte_plm_base_orted_append_basic_args(&argc, &argv, "tm",
&proc_vpid_index,
nodelist);
free(nodelist);
if (0 < opal_output_get_verbosity(orte_plm_base_framework.framework_output)) {
param = opal_argv_join(argv, ' ');
OPAL_OUTPUT_VERBOSE((1, orte_plm_base_framework.framework_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);
}
if (!connected) {
if (ORTE_SUCCESS != plm_tm_connect()) {
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);
/* enable local launch by the orteds */
(void) mca_base_var_env_name ("plm", &var);
opal_setenv(var, "rsh", true, &env);
free(var);
/* add our umask -- see big note in orted.c */
current_umask = umask(0);
umask(current_umask);
(void)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
*/
app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, 0);
orte_get_attribute(&app->attributes, ORTE_APP_PREFIX_DIR, (void**)&prefix_dir, OPAL_STRING);
if (NULL != prefix_dir) {
char *newenv;
for (i = 0; NULL != env && NULL != env[i]; ++i) {
/* Reset PATH */
if (0 == strncmp("PATH=", env[i], 5)) {
(void)asprintf(&newenv, "%s/%s:%s",
prefix_dir, bin_base, env[i] + 5);
OPAL_OUTPUT_VERBOSE((1, orte_plm_base_framework.framework_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)) {
(void)asprintf(&newenv, "%s/%s:%s",
prefix_dir, lib_base, env[i] + 16);
OPAL_OUTPUT_VERBOSE((1, orte_plm_base_framework.framework_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);
}
}
free(prefix_dir);
}
/* Iterate through each of the nodes and spin
* up a daemon.
*/
ldptr = &launchid;
for (i = 0; i < map->nodes->size; i++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
/* if this daemon already exists, don't launch it! */
if (ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_DAEMON_LAUNCHED)) {
continue;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_base_framework.framework_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, node->daemon->name.vpid);
if (ORTE_SUCCESS != rc) {
opal_output(0, "plm:tm: unable to get daemon vpid as string");
exit(-1);
}
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_base_framework.framework_output)) {
param = opal_argv_join(argv, ' ');
OPAL_OUTPUT_VERBOSE((1, orte_plm_base_framework.framework_output,
"%s plm:tm: executing:\n\t%s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == param) ? "NULL" : param));
if (NULL != param) free(param);
}
launchid = 0;
if (!orte_get_attribute(&node->attributes, ORTE_NODE_LAUNCH_ID, (void**)&ldptr, OPAL_INT32)) {
orte_show_help("help-plm-tm.txt", "tm-spawn-failed", true, argv[0], node->name, 0);
rc = ORTE_ERROR;
goto cleanup;
}
rc = tm_spawn(argc, argv, env, launchid, tm_task_ids + launched, tm_events + launched);
if (TM_SUCCESS != rc) {
orte_show_help("help-plm-tm.txt", "tm-spawn-failed", true, argv[0], node->name, launchid);
rc = ORTE_ERROR;
goto cleanup;
}
launched++;
}
/* indicate that the daemons for this job were launched */
state->jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
daemons->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
/* flag that launch was successful, so far as we currently know */
failed_launch = false;
OPAL_OUTPUT_VERBOSE((1, orte_plm_base_framework.framework_output,
"%s plm:tm:launch: finished spawning orteds",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
cleanup:
/* cleanup */
OBJ_RELEASE(state);
/* check for failed launch - if so, force terminate */
if (failed_launch) {
ORTE_ACTIVATE_JOB_STATE(daemons, ORTE_JOB_STATE_FAILED_TO_START);
}
}
static int tool_init(int argc, char *argv[]) {
int exit_status = ORTE_SUCCESS, ret;
char * tmp_env_var = NULL;
listener_started = false;
/*
* Make sure to init util before parse_args
* to ensure installdirs is setup properly
* before calling mca_base_open();
*/
if( ORTE_SUCCESS != (ret = opal_init_util(&argc, &argv)) ) {
return ret;
}
/*
* Parse Command Line Arguments
*/
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
return ret;
}
/* Disable the migrate notification routine for this
* tool. As we will never need to migrate this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a migrateer */
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/* we are never allowed to operate as a distributed tool,
* so insist on the ess/tool component */
opal_setenv("OMPI_MCA_ess", "tool", true, &environ);
/***************************
* We need all of OPAL and the TOOLS portion of ORTE - this
* sets us up so we can talk to any HNP over the wire
***************************/
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_TOOL))) {
exit_status = ret;
goto cleanup;
}
/*
* Setup ORTE Output handle from the verbose argument
*/
if( orte_migrate_globals.verbose ) {
orte_migrate_globals.output = opal_output_open(NULL);
opal_output_set_verbosity(orte_migrate_globals.output, orte_migrate_globals.verbose_level);
} else {
orte_migrate_globals.output = 0; /* Default=STDERR */
}
/*
* Start the listener
*/
if( ORTE_SUCCESS != (ret = start_listener() ) ) {
exit_status = ret;
}
cleanup:
return exit_status;
}
static void set(orte_job_t *jobdat,
orte_proc_t *child,
char ***environ_copy,
int write_fd)
{
hwloc_cpuset_t cpuset;
hwloc_obj_t root;
opal_hwloc_topo_data_t *sum;
orte_app_context_t *context;
int rc=ORTE_ERROR;
char *msg, *param;
char *cpu_bitmap;
opal_output_verbose(2, orte_rtc_base_framework.framework_output,
"%s hwloc:set on child %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == child) ? "NULL" : ORTE_NAME_PRINT(&child->name));
if (NULL == jobdat || NULL == child) {
/* nothing for us to do */
opal_output_verbose(2, orte_rtc_base_framework.framework_output,
"%s hwloc:set jobdat %s child %s - nothing to do",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == jobdat) ? "NULL" : ORTE_JOBID_PRINT(jobdat->jobid),
(NULL == child) ? "NULL" : ORTE_NAME_PRINT(&child->name));
return;
}
context = (orte_app_context_t*)opal_pointer_array_get_item(jobdat->apps, child->app_idx);
/* Set process affinity, if given */
cpu_bitmap = NULL;
if (!orte_get_attribute(&child->attributes, ORTE_PROC_CPU_BITMAP, (void**)&cpu_bitmap, OPAL_STRING) ||
NULL == cpu_bitmap || 0 == strlen(cpu_bitmap)) {
/* if the daemon is bound, then we need to "free" this proc */
if (NULL != orte_daemon_cores) {
root = hwloc_get_root_obj(opal_hwloc_topology);
if (NULL == root->userdata) {
orte_rtc_base_send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "incorrectly bound",
orte_process_info.nodename, context->app,
__FILE__, __LINE__);
}
sum = (opal_hwloc_topo_data_t*)root->userdata;
/* bind this proc to all available processors */
rc = hwloc_set_cpubind(opal_hwloc_topology, sum->available, 0);
/* if we got an error and this wasn't a default binding policy, then report it */
if (rc < 0 && OPAL_BINDING_POLICY_IS_SET(jobdat->map->binding)) {
if (errno == ENOSYS) {
msg = "hwloc indicates cpu binding not supported";
} else if (errno == EXDEV) {
msg = "hwloc indicates cpu binding cannot be enforced";
} else {
char *tmp;
(void)hwloc_bitmap_list_asprintf(&tmp, sum->available);
asprintf(&msg, "hwloc_set_cpubind returned \"%s\" for bitmap \"%s\"",
opal_strerror(rc), tmp);
free(tmp);
}
if (OPAL_BINDING_REQUIRED(jobdat->map->binding)) {
/* If binding is required, send an error up the pipe (which exits
-- it doesn't return). */
orte_rtc_base_send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"binding generic error",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
} else {
orte_rtc_base_send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "not bound",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
return;
}
}
}
if (0 == rc && opal_hwloc_report_bindings) {
opal_output(0, "MCW rank %d is not bound (or bound to all available processors)", child->name.vpid);
/* avoid reporting it twice */
(void) mca_base_var_env_name ("hwloc_base_report_bindings", ¶m);
opal_unsetenv(param, environ_copy);
free(param);
}
} else {
/* convert the list to a cpuset */
cpuset = hwloc_bitmap_alloc();
if (0 != (rc = hwloc_bitmap_list_sscanf(cpuset, cpu_bitmap))) {
/* See comment above about "This may be a small memory leak" */
asprintf(&msg, "hwloc_bitmap_sscanf returned \"%s\" for the string \"%s\"",
opal_strerror(rc), cpu_bitmap);
if (NULL == msg) {
msg = "failed to convert bitmap list to hwloc bitmap";
}
if (OPAL_BINDING_REQUIRED(jobdat->map->binding) &&
OPAL_BINDING_POLICY_IS_SET(jobdat->map->binding)) {
/* If binding is required and a binding directive was explicitly
* given (i.e., we are not binding due to a default policy),
* send an error up the pipe (which exits -- it doesn't return).
*/
orte_rtc_base_send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"binding generic error",
orte_process_info.nodename,
context->app, msg,
__FILE__, __LINE__);
} else {
orte_rtc_base_send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "not bound",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
free(cpu_bitmap);
return;
}
}
/* bind as specified */
rc = hwloc_set_cpubind(opal_hwloc_topology, cpuset, 0);
/* if we got an error and this wasn't a default binding policy, then report it */
if (rc < 0 && OPAL_BINDING_POLICY_IS_SET(jobdat->map->binding)) {
char *tmp = NULL;
if (errno == ENOSYS) {
msg = "hwloc indicates cpu binding not supported";
} else if (errno == EXDEV) {
msg = "hwloc indicates cpu binding cannot be enforced";
} else {
asprintf(&msg, "hwloc_set_cpubind returned \"%s\" for bitmap \"%s\"",
opal_strerror(rc), cpu_bitmap);
}
if (OPAL_BINDING_REQUIRED(jobdat->map->binding)) {
/* If binding is required, send an error up the pipe (which exits
-- it doesn't return). */
orte_rtc_base_send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"binding generic error",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
} else {
orte_rtc_base_send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "not bound",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
if (NULL != tmp) {
free(tmp);
free(msg);
}
return;
}
if (NULL != tmp) {
free(tmp);
free(msg);
}
}
if (0 == rc && opal_hwloc_report_bindings) {
char tmp1[1024], tmp2[1024];
hwloc_cpuset_t mycpus;
/* get the cpus we are bound to */
mycpus = hwloc_bitmap_alloc();
if (hwloc_get_cpubind(opal_hwloc_topology,
mycpus,
HWLOC_CPUBIND_PROCESS) < 0) {
opal_output(0, "MCW rank %d is not bound",
child->name.vpid);
} else {
if (OPAL_ERR_NOT_BOUND == opal_hwloc_base_cset2str(tmp1, sizeof(tmp1), opal_hwloc_topology, mycpus)) {
opal_output(0, "MCW rank %d is not bound (or bound to all available processors)", child->name.vpid);
} else {
opal_hwloc_base_cset2mapstr(tmp2, sizeof(tmp2), opal_hwloc_topology, mycpus);
opal_output(0, "MCW rank %d bound to %s: %s",
child->name.vpid, tmp1, tmp2);
}
}
hwloc_bitmap_free(mycpus);
/* avoid reporting it twice */
(void) mca_base_var_env_name ("hwloc_base_report_bindings", ¶m);
opal_unsetenv(param, environ_copy);
free(param);
}
/* set memory affinity policy - if we get an error, don't report
* anything unless the user actually specified the binding policy
*/
rc = opal_hwloc_base_set_process_membind_policy();
if (ORTE_SUCCESS != rc && OPAL_BINDING_POLICY_IS_SET(jobdat->map->binding)) {
if (errno == ENOSYS) {
msg = "hwloc indicates memory binding not supported";
} else if (errno == EXDEV) {
msg = "hwloc indicates memory binding cannot be enforced";
} else {
msg = "failed to bind memory";
}
if (OPAL_HWLOC_BASE_MBFA_ERROR == opal_hwloc_base_mbfa) {
/* If binding is required, send an error up the pipe (which exits
-- it doesn't return). */
orte_rtc_base_send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"memory binding error",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
} else {
orte_rtc_base_send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "memory not bound",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
free(cpu_bitmap);
return;
}
}
}
if (NULL != cpu_bitmap) {
free(cpu_bitmap);
}
}
int opal_register_params(void)
{
int ret;
char *string = NULL;
if (opal_register_done) {
return OPAL_SUCCESS;
}
opal_register_done = true;
/*
* This string is going to be used in opal/util/stacktrace.c
*/
{
int j;
int signals[] = {
#ifdef SIGABRT
SIGABRT,
#endif
#ifdef SIGBUS
SIGBUS,
#endif
#ifdef SIGFPE
SIGFPE,
#endif
#ifdef SIGSEGV
SIGSEGV,
#endif
-1
};
for (j = 0 ; signals[j] != -1 ; ++j) {
if (j == 0) {
opal_asprintf(&string, "%d", signals[j]);
} else {
char *tmp;
opal_asprintf(&tmp, "%s,%d", string, signals[j]);
free(string);
string = tmp;
}
}
opal_signal_string = string;
ret = mca_base_var_register ("opal", "opal", NULL, "signal",
"Comma-delimited list of integer signal numbers to Open MPI to attempt to intercept. Upon receipt of the intercepted signal, Open MPI will display a stack trace and abort. Open MPI will *not* replace signals if handlers are already installed by the time MPI_INIT is invoked. Optionally append \":complain\" to any signal number in the comma-delimited list to make Open MPI complain if it detects another signal handler (and therefore does not insert its own).",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_signal_string);
free (string);
if (0 > ret) {
return ret;
}
}
/*
* Where should the stack trace output be directed
* This string is going to be used in opal/util/stacktrace.c
*/
string = strdup("stderr");
opal_stacktrace_output_filename = string;
ret = mca_base_var_register ("opal", "opal", NULL, "stacktrace_output",
"Specifies where the stack trace output stream goes. "
"Accepts one of the following: none (disabled), stderr (default), stdout, file[:filename]. "
"If 'filename' is not specified, a default filename of 'stacktrace' is used. "
"The 'filename' is appended with either '.PID' or '.RANK.PID', if RANK is available. "
"The 'filename' can be an absolute path or a relative path to the current working directory.",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_LOCAL,
&opal_stacktrace_output_filename);
free (string);
if (0 > ret) {
return ret;
}
#if defined(HAVE_SCHED_YIELD)
opal_progress_yield_when_idle = false;
ret = mca_base_var_register ("opal", "opal", "progress", "yield_when_idle",
"Yield the processor when waiting on progress",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_progress_yield_when_idle);
#endif
#if OPAL_ENABLE_DEBUG
opal_progress_debug = false;
ret = mca_base_var_register ("opal", "opal", "progress", "debug",
"Set to non-zero to debug progress engine features",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_progress_debug);
if (0 > ret) {
return ret;
}
opal_debug_threads = false;
ret = mca_base_var_register ("opal", "opal", "debug", "threads",
"Debug thread usage within OPAL. Reports out "
"when threads are acquired and released.",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_debug_threads);
if (0 > ret) {
return ret;
}
#endif
/* RFC1918 defines
- 10.0.0./8
- 172.16.0.0/12
- 192.168.0.0/16
RFC3330 also mentions
- 169.254.0.0/16 for DHCP onlink iff there's no DHCP server
*/
opal_net_private_ipv4 = "10.0.0.0/8;172.16.0.0/12;192.168.0.0/16;169.254.0.0/16";
ret = mca_base_var_register ("opal", "opal", "net", "private_ipv4",
"Semicolon-delimited list of CIDR notation entries specifying what networks are considered \"private\" (default value based on RFC1918 and RFC3330)",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_ALL_EQ,
&opal_net_private_ipv4);
if (0 > ret) {
return ret;
}
opal_set_max_sys_limits = NULL;
ret = mca_base_var_register ("opal", "opal", NULL, "set_max_sys_limits",
"Set the specified system-imposed limits to the specified value, including \"unlimited\"."
"Supported params: core, filesize, maxmem, openfiles, stacksize, maxchildren",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_ALL_EQ,
&opal_set_max_sys_limits);
if (0 > ret) {
return ret;
}
opal_abort_delay = 0;
ret = mca_base_var_register("opal", "opal", NULL, "abort_delay",
"If nonzero, print out an identifying message when abort operation is invoked (hostname, PID of the process that called abort) and delay for that many seconds before exiting (a negative delay value means to never abort). This allows attaching of a debugger before quitting the job.",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
OPAL_INFO_LVL_5,
MCA_BASE_VAR_SCOPE_READONLY,
&opal_abort_delay);
if (0 > ret) {
return ret;
}
opal_abort_print_stack = false;
ret = mca_base_var_register("opal", "opal", NULL, "abort_print_stack",
"If nonzero, print out a stack trace when abort is invoked",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0,
/* If we do not have stack trace
capability, make this a constant
MCA variable */
#if OPAL_WANT_PRETTY_PRINT_STACKTRACE
0,
OPAL_INFO_LVL_5,
MCA_BASE_VAR_SCOPE_READONLY,
#else
MCA_BASE_VAR_FLAG_DEFAULT_ONLY,
OPAL_INFO_LVL_5,
MCA_BASE_VAR_SCOPE_CONSTANT,
#endif
&opal_abort_print_stack);
if (0 > ret) {
return ret;
}
/* register the envar-forwarding params */
(void)mca_base_var_register ("opal", "mca", "base", "env_list",
"Set SHELL env variables",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0, OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_READONLY, &mca_base_env_list);
mca_base_env_list_sep = MCA_BASE_ENV_LIST_SEP_DEFAULT;
(void)mca_base_var_register ("opal", "mca", "base", "env_list_delimiter",
"Set SHELL env variables delimiter. Default: semicolon ';'",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0, OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_READONLY, &mca_base_env_list_sep);
/* Set OMPI_MCA_mca_base_env_list variable, it might not be set before
* if mca variable was taken from amca conf file. Need to set it
* here because mca_base_var_process_env_list is called from schizo_ompi.c
* only when this env variable was set.
*/
if (NULL != mca_base_env_list) {
char *name = NULL;
(void) mca_base_var_env_name ("mca_base_env_list", &name);
if (NULL != name) {
opal_setenv(name, mca_base_env_list, false, &environ);
free(name);
}
}
/* Register internal MCA variable mca_base_env_list_internal. It can be set only during
* parsing of amca conf file and contains SHELL env variables specified via -x there.
* Its format is the same as for mca_base_env_list.
*/
(void)mca_base_var_register ("opal", "mca", "base", "env_list_internal",
"Store SHELL env variables from amca conf file",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_INTERNAL, OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_READONLY, &mca_base_env_list_internal);
/* dss has parameters */
ret = opal_dss_register_vars ();
if (OPAL_SUCCESS != ret) {
return ret;
}
mca_base_var_register("opal", "opal", NULL, "pmix_verbose",
"Verbosity for OPAL-level PMIx code",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
OPAL_INFO_LVL_5,
MCA_BASE_VAR_SCOPE_READONLY,
&opal_pmix_verbose_output);
return OPAL_SUCCESS;
}
static int parse_args(int argc, char *argv[]) {
int i, ret, len, exit_status = ORTE_SUCCESS ;
opal_cmd_line_t cmd_line;
char **app_env = NULL, **global_env = NULL;
char * tmp_env_var = NULL;
char *argv0 = NULL;
/* Init structure */
memset(&orte_migrate_globals, 0, sizeof(orte_migrate_globals_t));
orte_migrate_globals.help = false;
orte_migrate_globals.pid = -1;
orte_migrate_globals.verbose = false;
orte_migrate_globals.verbose_level = 0;
orte_migrate_globals.status = false;
orte_migrate_globals.output = -1;
orte_migrate_globals.off_nodes = NULL;
orte_migrate_globals.off_procs = NULL;
orte_migrate_globals.onto_nodes = NULL;
#if OPAL_ENABLE_FT_CR == 0
/* Warn and exit if not configured with Migrate/Restart */
{
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-migrate.txt", "usage-no-cr",
true, args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
exit_status = ORTE_ERROR;
goto cleanup;
}
#endif
/* Parse the command line options */
opal_cmd_line_create(&cmd_line, cmd_line_opts);
mca_base_open();
mca_base_cmd_line_setup(&cmd_line);
ret = opal_cmd_line_parse(&cmd_line, false, false, argc, argv);
if (OPAL_SUCCESS != ret) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
exit_status = 1;
goto cleanup;
}
if (orte_migrate_globals.help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-migrate.txt", "usage", true,
args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
exit(0);
}
/**
* Put all of the MCA arguments in the environment
*/
mca_base_cmd_line_process_args(argv, &app_env, &global_env);
len = opal_argv_count(app_env);
for(i = 0; i < len; ++i) {
putenv(app_env[i]);
}
len = opal_argv_count(global_env);
for(i = 0; i < len; ++i) {
putenv(global_env[i]);
}
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/**
* Now start parsing our specific arguments
*/
/* get the remaining bits */
argv0 = strdup(argv[0]);
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
if (NULL == orte_migrate_globals.off_nodes &&
NULL == orte_migrate_globals.off_procs) {
fprintf(stderr, "%s: Nothing to do\n", argv0);
fprintf(stderr, "Type '%s --help' for usage.\n", argv0);
exit_status = 1;
goto cleanup;
}
if(orte_migrate_globals.verbose_level < 0 ) {
orte_migrate_globals.verbose_level = 0;
}
if(orte_migrate_globals.verbose_level > 0) {
orte_migrate_globals.verbose = true;
}
/*
* If the user did not supply an hnp jobid, then they must
* supply the PID of MPIRUN
*/
if(0 >= argc ) {
fprintf(stderr, "%s: Nothing to do\n", argv[0]);
fprintf(stderr, "Type '%s --help' for usage.\n", argv[0]);
exit_status = ORTE_ERROR;
goto cleanup;
}
orte_migrate_globals.pid = atoi(argv[0]);
if ( 0 >= orte_migrate_globals.pid ) {
opal_show_help("help-orte-migrate.txt", "invalid_pid", true,
orte_migrate_globals.pid);
exit_status = ORTE_ERROR;
goto cleanup;
}
if(orte_migrate_globals.verbose) {
orte_migrate_globals.status = true;
}
if(orte_migrate_globals.verbose) {
pretty_print_migration();
}
cleanup:
if (NULL != argv0) {
free(argv0);
}
return exit_status;
}
static int parse_args(int argc, char *argv[]) {
int i, ret, len;
opal_cmd_line_t cmd_line;
char **app_env = NULL, **global_env = NULL;
char * tmp_env_var = NULL;
char *argv0 = NULL;
memset(&opal_checkpoint_globals, 0, sizeof(opal_checkpoint_globals_t));
opal_checkpoint_globals.snapshot_name = NULL;
opal_checkpoint_globals.snapshot_loc = NULL;
/* Parse the command line options */
opal_cmd_line_create(&cmd_line, cmd_line_opts);
mca_base_open();
mca_base_cmd_line_setup(&cmd_line);
ret = opal_cmd_line_parse(&cmd_line, true, false, argc, argv);
if (OPAL_SUCCESS != ret) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
return 1;
}
if (opal_checkpoint_globals.help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-opal-checkpoint.txt", "usage", true,
args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
exit(0);
}
/**
* Put all of the MCA arguments in the environment
*/
mca_base_cmd_line_process_args(&cmd_line, &app_env, &global_env);
len = opal_argv_count(app_env);
for(i = 0; i < len; ++i) {
putenv(app_env[i]);
}
len = opal_argv_count(global_env);
for(i = 0; i < len; ++i) {
putenv(global_env[i]);
}
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/**
* Now start parsing our specific arguments
*/
if( NULL == opal_checkpoint_globals.snapshot_name )
opal_checkpoint_globals.snapshot_name = strdup("");
if( NULL == opal_checkpoint_globals.snapshot_loc ) {
opal_checkpoint_globals.snapshot_loc = strdup("");
}
/* get the remaining bits */
argv0 = strdup(argv[0]);
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
if (0 == argc) {
fprintf(stderr, "%s: Nothing to do\n", argv0);
fprintf(stderr, "Type '%s --help' for usage.\n", argv0);
free(argv0);
return OPAL_ERROR;
}
free(argv0);
opal_checkpoint_globals.pid = atoi(argv[0]);
if ( 0 >= opal_checkpoint_globals.pid ) {
opal_show_help("help-opal-checkpoint.txt", "invalid_pid", true,
opal_checkpoint_globals.pid);
return OPAL_ERROR;
}
return OPAL_SUCCESS;
}
static int orte_ps_init(int argc, char *argv[]) {
int ret;
#if OPAL_ENABLE_FT_CR == 1
char * tmp_env_var = NULL;
#endif
/*
* Make sure to init util before parse_args
* to ensure installdirs is setup properly
* before calling mca_base_open();
*/
if( ORTE_SUCCESS != (ret = opal_init_util(&argc, &argv)) ) {
return ret;
}
/*
* Parse Command Line Arguments
*/
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
return ret;
}
/*
* Setup OPAL Output handle from the verbose argument
*/
if( orte_ps_globals.verbose ) {
orte_ps_globals.output = opal_output_open(NULL);
opal_output_set_verbosity(orte_ps_globals.output, 10);
} else {
orte_ps_globals.output = 0; /* Default=STDERR */
}
#if OPAL_ENABLE_FT_CR == 1
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#endif
/***************************
* We need all of OPAL and the TOOL portion of ORTE
***************************/
ret = orte_init(&argc, &argv, ORTE_PROC_TOOL);
return ret;
}
int orte_daemon(int argc, char *argv[])
{
int ret = 0;
opal_cmd_line_t *cmd_line = NULL;
int i;
opal_buffer_t *buffer;
char hostname[OPAL_MAXHOSTNAMELEN];
#if OPAL_ENABLE_FT_CR == 1
char *tmp_env_var = NULL;
#endif
/* initialize the globals */
memset(&orted_globals, 0, sizeof(orted_globals));
/* initialize the singleton died pipe to an illegal value so we can detect it was set */
orted_globals.singleton_died_pipe = -1;
bucket = OBJ_NEW(opal_buffer_t);
/* setup to check common command line options that just report and die */
cmd_line = OBJ_NEW(opal_cmd_line_t);
if (OPAL_SUCCESS != opal_cmd_line_create(cmd_line, orte_cmd_line_opts)) {
OBJ_RELEASE(cmd_line);
exit(1);
}
mca_base_cmd_line_setup(cmd_line);
if (ORTE_SUCCESS != (ret = opal_cmd_line_parse(cmd_line, false, false,
argc, argv))) {
char *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
fprintf(stderr, "Usage: %s [OPTION]...\n%s\n", argv[0], args);
free(args);
OBJ_RELEASE(cmd_line);
return ret;
}
/*
* Since this process can now handle MCA/GMCA parameters, make sure to
* process them.
*/
mca_base_cmd_line_process_args(cmd_line, &environ, &environ);
/* Ensure that enough of OPAL is setup for us to be able to run */
/*
* NOTE: (JJH)
* We need to allow 'mca_base_cmd_line_process_args()' to process command
* line arguments *before* calling opal_init_util() since the command
* line could contain MCA parameters that affect the way opal_init_util()
* functions. AMCA parameters are one such option normally received on the
* command line that affect the way opal_init_util() behaves.
* It is "safe" to call mca_base_cmd_line_process_args() before
* opal_init_util() since mca_base_cmd_line_process_args() does *not*
* depend upon opal_init_util() functionality.
*/
if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
fprintf(stderr, "OPAL failed to initialize -- orted aborting\n");
exit(1);
}
/* save the environment for launch purposes. This MUST be
* done so that we can pass it to any local procs we
* spawn - otherwise, those local procs won't see any
* non-MCA envars that were set in the enviro when the
* orted was executed - e.g., by .csh
*/
orte_launch_environ = opal_argv_copy(environ);
/* purge any ess/pmix flags set in the environ when we were launched */
opal_unsetenv(OPAL_MCA_PREFIX"ess", &orte_launch_environ);
opal_unsetenv(OPAL_MCA_PREFIX"pmix", &orte_launch_environ);
/* if orte_daemon_debug is set, let someone know we are alive right
* away just in case we have a problem along the way
*/
if (orted_globals.debug) {
gethostname(hostname, sizeof(hostname));
fprintf(stderr, "Daemon was launched on %s - beginning to initialize\n", hostname);
}
/* check for help request */
if (orted_globals.help) {
char *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
orte_show_help("help-orted.txt", "orted:usage", false,
argv[0], args);
free(args);
return 1;
}
#if defined(HAVE_SETSID)
/* see if we were directed to separate from current session */
if (orted_globals.set_sid) {
setsid();
}
#endif
/* see if they want us to spin until they can connect a debugger to us */
i=0;
while (orted_spin_flag) {
i++;
if (1000 < i) i=0;
}
#if OPAL_ENABLE_FT_CR == 1
/* Mark as a tool program */
(void) mca_base_var_env_name ("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#endif
/* detach from controlling terminal
* otherwise, remain attached so output can get to us
*/
if(!orte_debug_flag &&
!orte_debug_daemons_flag &&
orted_globals.daemonize) {
opal_daemon_init(NULL);
}
/* Set the flag telling OpenRTE that I am NOT a
* singleton, but am "infrastructure" - prevents setting
* up incorrect infrastructure that only a singleton would
* require.
*/
if (orted_globals.hnp) {
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_HNP))) {
ORTE_ERROR_LOG(ret);
return ret;
}
} else {
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_DAEMON))) {
ORTE_ERROR_LOG(ret);
return ret;
}
}
/* finalize the OPAL utils. As they are opened again from orte_init->opal_init
* we continue to have a reference count on them. So we have to finalize them twice...
*/
opal_finalize_util();
/* bind ourselves if so directed */
if (NULL != orte_daemon_cores) {
char **cores=NULL, tmp[128];
hwloc_obj_t pu;
hwloc_cpuset_t ours, res;
int core;
/* could be a collection of comma-delimited ranges, so
* use our handy utility to parse it
*/
orte_util_parse_range_options(orte_daemon_cores, &cores);
if (NULL != cores) {
ours = hwloc_bitmap_alloc();
hwloc_bitmap_zero(ours);
res = hwloc_bitmap_alloc();
for (i=0; NULL != cores[i]; i++) {
core = strtoul(cores[i], NULL, 10);
if (NULL == (pu = opal_hwloc_base_get_pu(opal_hwloc_topology, core, OPAL_HWLOC_LOGICAL))) {
/* turn off the show help forwarding as we won't
* be able to cycle the event library to send
*/
orte_show_help_finalize();
/* the message will now come out locally */
orte_show_help("help-orted.txt", "orted:cannot-bind",
true, orte_process_info.nodename,
orte_daemon_cores);
ret = ORTE_ERR_NOT_SUPPORTED;
hwloc_bitmap_free(ours);
hwloc_bitmap_free(res);
goto DONE;
}
hwloc_bitmap_or(res, ours, pu->cpuset);
hwloc_bitmap_copy(ours, res);
}
/* if the result is all zeros, then don't bind */
if (!hwloc_bitmap_iszero(ours)) {
(void)hwloc_set_cpubind(opal_hwloc_topology, ours, 0);
if (opal_hwloc_report_bindings) {
opal_hwloc_base_cset2mapstr(tmp, sizeof(tmp), opal_hwloc_topology, ours);
opal_output(0, "Daemon %s is bound to cores %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), tmp);
}
}
/* cleanup */
hwloc_bitmap_free(ours);
hwloc_bitmap_free(res);
opal_argv_free(cores);
}
}
if ((int)ORTE_VPID_INVALID != orted_debug_failure) {
orted_globals.abort=false;
/* some vpid was ordered to fail. The value can be positive
* or negative, depending upon the desired method for failure,
* so need to check both here
*/
if (0 > orted_debug_failure) {
orted_debug_failure = -1*orted_debug_failure;
orted_globals.abort = true;
}
/* are we the specified vpid? */
if ((int)ORTE_PROC_MY_NAME->vpid == orted_debug_failure) {
/* if the user specified we delay, then setup a timer
* and have it kill us
*/
if (0 < orted_debug_failure_delay) {
ORTE_TIMER_EVENT(orted_debug_failure_delay, 0, shutdown_callback, ORTE_SYS_PRI);
} else {
opal_output(0, "%s is executing clean %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
orted_globals.abort ? "abort" : "abnormal termination");
/* do -not- call finalize as this will send a message to the HNP
* indicating clean termination! Instead, just forcibly cleanup
* the local session_dir tree and exit
*/
orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);
/* if we were ordered to abort, do so */
if (orted_globals.abort) {
abort();
}
/* otherwise, return with non-zero status */
ret = ORTE_ERROR_DEFAULT_EXIT_CODE;
goto DONE;
}
}
}
/* insert our contact info into our process_info struct so we
* have it for later use and set the local daemon field to our name
*/
orte_oob_base_get_addr(&orte_process_info.my_daemon_uri);
if (NULL == orte_process_info.my_daemon_uri) {
/* no way to communicate */
ret = ORTE_ERROR;
goto DONE;
}
ORTE_PROC_MY_DAEMON->jobid = ORTE_PROC_MY_NAME->jobid;
ORTE_PROC_MY_DAEMON->vpid = ORTE_PROC_MY_NAME->vpid;
/* if I am also the hnp, then update that contact info field too */
if (ORTE_PROC_IS_HNP) {
orte_process_info.my_hnp_uri = strdup(orte_process_info.my_daemon_uri);
ORTE_PROC_MY_HNP->jobid = ORTE_PROC_MY_NAME->jobid;
ORTE_PROC_MY_HNP->vpid = ORTE_PROC_MY_NAME->vpid;
}
/* setup the primary daemon command receive function */
orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DAEMON,
ORTE_RML_PERSISTENT, orte_daemon_recv, NULL);
/* output a message indicating we are alive, our name, and our pid
* for debugging purposes
*/
if (orte_debug_daemons_flag) {
fprintf(stderr, "Daemon %s checking in as pid %ld on host %s\n",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), (long)orte_process_info.pid,
orte_process_info.nodename);
}
/* We actually do *not* want the orted to voluntarily yield() the
processor more than necessary. The orted already blocks when
it is doing nothing, so it doesn't use any more CPU cycles than
it should; but when it *is* doing something, we do not want it
to be unnecessarily delayed because it voluntarily yielded the
processor in the middle of its work.
For example: when a message arrives at the orted, we want the
OS to wake up the orted in a timely fashion (which most OS's
seem good about doing) and then we want the orted to process
the message as fast as possible. If the orted yields and lets
aggressive MPI applications get the processor back, it may be a
long time before the OS schedules the orted to run again
(particularly if there is no IO event to wake it up). Hence,
routed OOB messages (for example) may be significantly delayed
before being delivered to MPI processes, which can be
problematic in some scenarios (e.g., COMM_SPAWN, BTL's that
require OOB messages for wireup, etc.). */
opal_progress_set_yield_when_idle(false);
/* Change the default behavior of libevent such that we want to
continually block rather than blocking for the default timeout
and then looping around the progress engine again. There
should be nothing in the orted that cannot block in libevent
until "something" happens (i.e., there's no need to keep
cycling through progress because the only things that should
happen will happen in libevent). This is a minor optimization,
but what the heck... :-) */
opal_progress_set_event_flag(OPAL_EVLOOP_ONCE);
/* if requested, report my uri to the indicated pipe */
if (orted_globals.uri_pipe > 0) {
orte_job_t *jdata;
orte_proc_t *proc;
orte_node_t *node;
orte_app_context_t *app;
char *tmp, *nptr, *sysinfo;
char **singenv=NULL, *string_key, *env_str;
/* setup the singleton's job */
jdata = OBJ_NEW(orte_job_t);
/* default to ompi for now */
opal_argv_append_nosize(&jdata->personality, "ompi");
orte_plm_base_create_jobid(jdata);
opal_hash_table_set_value_uint32(orte_job_data, jdata->jobid, jdata);
/* must create a map for it (even though it has no
* info in it) so that the job info will be picked
* up in subsequent pidmaps or other daemons won't
* know how to route
*/
jdata->map = OBJ_NEW(orte_job_map_t);
/* setup an app_context for the singleton */
app = OBJ_NEW(orte_app_context_t);
app->app = strdup("singleton");
app->num_procs = 1;
opal_pointer_array_add(jdata->apps, app);
jdata->num_apps = 1;
/* setup a proc object for the singleton - since we
* -must- be the HNP, and therefore we stored our
* node on the global node pool, and since the singleton
* -must- be on the same node as us, indicate that
*/
proc = OBJ_NEW(orte_proc_t);
proc->name.jobid = jdata->jobid;
proc->name.vpid = 0;
proc->parent = 0;
ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_ALIVE);
proc->state = ORTE_PROC_STATE_RUNNING;
proc->app_idx = 0;
/* obviously, it is on my node */
node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, 0);
proc->node = node;
OBJ_RETAIN(node); /* keep accounting straight */
opal_pointer_array_add(jdata->procs, proc);
jdata->num_procs = 1;
/* add the node to the job map */
OBJ_RETAIN(node);
opal_pointer_array_add(jdata->map->nodes, node);
jdata->map->num_nodes++;
/* and it obviously is on the node */
OBJ_RETAIN(proc);
opal_pointer_array_add(node->procs, proc);
node->num_procs++;
/* and obviously it is one of my local procs */
OBJ_RETAIN(proc);
opal_pointer_array_add(orte_local_children, proc);
jdata->num_local_procs = 1;
/* set the trivial */
proc->local_rank = 0;
proc->node_rank = 0;
proc->app_rank = 0;
proc->state = ORTE_PROC_STATE_RUNNING;
proc->app_idx = 0;
ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_LOCAL);
/* set the ORTE_JOB_TRANSPORT_KEY from the environment */
orte_pre_condition_transports(jdata, NULL);
/* register the singleton's nspace with our PMIx server */
if (ORTE_SUCCESS != (ret = orte_pmix_server_register_nspace(jdata, false))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* use setup fork to create the envars needed by the singleton */
if (OPAL_SUCCESS != (ret = opal_pmix.server_setup_fork(&proc->name, &singenv))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* append the transport key to the envars needed by the singleton */
if (!orte_get_attribute(&jdata->attributes, ORTE_JOB_TRANSPORT_KEY, (void**)&string_key, OPAL_STRING) || NULL == string_key) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
goto DONE;
}
asprintf(&env_str, OPAL_MCA_PREFIX"orte_precondition_transports=%s", string_key);
opal_argv_append_nosize(&singenv, env_str);
free(env_str);
nptr = opal_argv_join(singenv, '*');
opal_argv_free(singenv);
/* create a string that contains our uri + sysinfo + PMIx server URI envars */
orte_util_convert_sysinfo_to_string(&sysinfo, orte_local_cpu_type, orte_local_cpu_model);
asprintf(&tmp, "%s[%s]%s", orte_process_info.my_daemon_uri, sysinfo, nptr);
free(sysinfo);
free(nptr);
/* pass that info to the singleton */
if (OPAL_SUCCESS != (ret = opal_fd_write(orted_globals.uri_pipe, strlen(tmp)+1, tmp))) { ; /* need to add 1 to get the NULL */
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* cleanup */
free(tmp);
close(orted_globals.uri_pipe);
/* since a singleton spawned us, we need to harvest
* any MCA params from the local environment so
* we can pass them along to any subsequent daemons
* we may start as the result of a comm_spawn
*/
for (i=0; NULL != environ[i]; i++) {
if (0 == strncmp(environ[i], OPAL_MCA_PREFIX, 9)) {
/* make a copy to manipulate */
tmp = strdup(environ[i]);
/* find the equal sign */
nptr = strchr(tmp, '=');
*nptr = '\0';
nptr++;
/* add the mca param to the orted cmd line */
opal_argv_append_nosize(&orted_cmd_line, "-"OPAL_MCA_CMD_LINE_ID);
opal_argv_append_nosize(&orted_cmd_line, &tmp[9]);
opal_argv_append_nosize(&orted_cmd_line, nptr);
free(tmp);
}
}
}
/* if we were given a pipe to monitor for singleton termination, set that up */
if (orted_globals.singleton_died_pipe > 0) {
/* register shutdown handler */
pipe_handler = (opal_event_t*)malloc(sizeof(opal_event_t));
opal_event_set(orte_event_base, pipe_handler,
orted_globals.singleton_died_pipe,
OPAL_EV_READ,
pipe_closed,
pipe_handler);
opal_event_add(pipe_handler, NULL);
}
/* If I have a parent, then save his contact info so
* any messages we send can flow thru him.
*/
orte_parent_uri = NULL;
(void) mca_base_var_register ("orte", "orte", NULL, "parent_uri",
"URI for the parent if tree launch is enabled.",
MCA_BASE_VAR_TYPE_STRING, NULL, 0,
MCA_BASE_VAR_FLAG_INTERNAL,
OPAL_INFO_LVL_9,
MCA_BASE_VAR_SCOPE_CONSTANT,
&orte_parent_uri);
if (NULL != orte_parent_uri) {
orte_process_name_t parent;
opal_value_t val;
/* set the contact info into our local database */
ret = orte_rml_base_parse_uris(orte_parent_uri, &parent, NULL);
if (ORTE_SUCCESS != ret) {
ORTE_ERROR_LOG(ret);
free (orte_parent_uri);
orte_parent_uri = NULL;
goto DONE;
}
OBJ_CONSTRUCT(&val, opal_value_t);
val.key = OPAL_PMIX_PROC_URI;
val.type = OPAL_STRING;
val.data.string = orte_parent_uri;
if (OPAL_SUCCESS != (ret = opal_pmix.store_local(&parent, &val))) {
ORTE_ERROR_LOG(ret);
OBJ_DESTRUCT(&val);
goto DONE;
}
val.key = NULL;
val.data.string = NULL;
OBJ_DESTRUCT(&val);
/* don't need this value anymore */
free(orte_parent_uri);
orte_parent_uri = NULL;
/* tell the routed module that we have a path
* back to the HNP
*/
if (ORTE_SUCCESS != (ret = orte_routed.update_route(NULL, ORTE_PROC_MY_HNP, &parent))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* set the lifeline to point to our parent so that we
* can handle the situation if that lifeline goes away
*/
if (ORTE_SUCCESS != (ret = orte_routed.set_lifeline(NULL, &parent))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
}
/* if we are not the HNP...the only time we will be an HNP
* is if we are launched by a singleton to provide support
* for it
*/
if (!ORTE_PROC_IS_HNP) {
orte_process_name_t target;
target.jobid = ORTE_PROC_MY_NAME->jobid;
if (orte_fwd_mpirun_port || orte_static_ports) {
/* setup the rollup callback */
orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_ORTED_CALLBACK,
ORTE_RML_PERSISTENT, rollup, NULL);
target.vpid = ORTE_PROC_MY_NAME->vpid;
/* since we will be waiting for any children to send us
* their rollup info before sending to our parent, save
* a little time in the launch phase by "warming up" the
* connection to our parent while we wait for our children */
buffer = OBJ_NEW(opal_buffer_t); // zero-byte message
if (0 > (ret = orte_rml.send_buffer_nb(orte_mgmt_conduit,
ORTE_PROC_MY_PARENT, buffer,
ORTE_RML_TAG_WARMUP_CONNECTION,
orte_rml_send_callback, NULL))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
} else {
target.vpid = 0;
}
/* send the information to the orted report-back point - this function
* will process the data, but also counts the number of
* orteds that reported back so the launch procedure can continue.
* We need to do this at the last possible second as the HNP
* can turn right around and begin issuing orders to us
*/
buffer = OBJ_NEW(opal_buffer_t);
/* insert our name for rollup purposes */
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, ORTE_PROC_MY_NAME, 1, ORTE_NAME))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
/* get any connection info we may have pushed */
{
opal_value_t *val = NULL, *kv;
opal_list_t *modex;
int32_t flag;
if (OPAL_SUCCESS != (ret = opal_pmix.get(ORTE_PROC_MY_NAME, NULL, NULL, &val)) || NULL == val) {
/* just pack a marker indicating we don't have any to share */
flag = 0;
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &flag, 1, OPAL_INT32))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
} else {
/* the data is returned as a list of key-value pairs in the opal_value_t */
if (OPAL_PTR == val->type) {
modex = (opal_list_t*)val->data.ptr;
flag = (int32_t)opal_list_get_size(modex);
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &flag, 1, OPAL_INT32))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
OPAL_LIST_FOREACH(kv, modex, opal_value_t) {
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &kv, 1, OPAL_VALUE))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
}
OPAL_LIST_RELEASE(modex);
} else {
opal_output(0, "VAL KEY: %s", (NULL == val->key) ? "NULL" : val->key);
/* single value */
flag = 1;
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &flag, 1, OPAL_INT32))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &val, 1, OPAL_VALUE))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
}
OBJ_RELEASE(val);
}
int
main(int argc, char *argv[])
{
int ret, exit_status = OPAL_SUCCESS;
int child_pid;
int prev_pid = 0;
int idx;
opal_crs_base_snapshot_t *snapshot = NULL;
char * tmp_env_var = NULL;
bool select = false;
/***************
* Initialize
***************/
if (OPAL_SUCCESS != (ret = initialize(argc, argv))) {
exit_status = ret;
goto cleanup;
}
/*
* Check for existence of the file, or program in the case of self
*/
if( OPAL_SUCCESS != (ret = check_file() )) {
opal_show_help("help-opal-restart.txt", "invalid_filename", true,
opal_restart_globals.snapshot_ref);
exit_status = ret;
goto cleanup;
}
/* Re-enable the selection of the CRS component, so we can choose the right one */
idx = mca_base_var_find(NULL, "crs", "base", "do_not_select");
if (0 > idx) {
opal_output(opal_restart_globals.output,
"MCA variable opal_crs_base_do_not_select not found\n");
exit_status = OPAL_ERROR;
goto cleanup;
}
ret = mca_base_var_set_value(idx, &select, 0, MCA_BASE_VAR_SOURCE_DEFAULT, NULL);
if (OPAL_SUCCESS != ret) {
exit_status = ret;
goto cleanup;
}
/*
* Make sure we are using the correct checkpointer
*/
if(NULL == expected_crs_comp) {
char * full_metadata_path = NULL;
FILE * metadata = NULL;
opal_asprintf(&full_metadata_path, "%s/%s/%s",
opal_restart_globals.snapshot_loc,
opal_restart_globals.snapshot_ref,
opal_restart_globals.snapshot_metadata);
if( NULL == (metadata = fopen(full_metadata_path, "r")) ) {
opal_show_help("help-opal-restart.txt", "invalid_metadata", true,
opal_restart_globals.snapshot_metadata,
full_metadata_path);
exit_status = OPAL_ERROR;
goto cleanup;
}
if( OPAL_SUCCESS != (ret = opal_crs_base_extract_expected_component(metadata,
&expected_crs_comp,
&prev_pid)) ) {
opal_show_help("help-opal-restart.txt", "invalid_metadata", true,
opal_restart_globals.snapshot_metadata,
full_metadata_path);
exit_status = ret;
goto cleanup;
}
free(full_metadata_path);
full_metadata_path = NULL;
fclose(metadata);
metadata = NULL;
}
opal_output_verbose(10, opal_restart_globals.output,
"Restart Expects checkpointer: (%s)",
expected_crs_comp);
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
expected_crs_comp,
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/* Select this component or don't continue.
* If the selection of this component fails, then we can't
* restart on this node because it doesn't have the proper checkpointer
* available.
*/
if( OPAL_SUCCESS != (ret = opal_crs_base_open(MCA_BASE_OPEN_DEFAULT)) ) {
opal_show_help("help-opal-restart.txt", "comp_select_failure", true,
"crs", ret);
exit_status = ret;
goto cleanup;
}
if( OPAL_SUCCESS != (ret = opal_crs_base_select()) ) {
opal_show_help("help-opal-restart.txt", "comp_select_failure", true,
expected_crs_comp, ret);
exit_status = ret;
goto cleanup;
}
/*
* Make sure we have selected the proper component
*/
if(NULL == expected_crs_comp ||
0 != strncmp(expected_crs_comp,
opal_crs_base_selected_component.base_version.mca_component_name,
strlen(expected_crs_comp)) ) {
opal_show_help("help-opal-restart.txt", "comp_select_mismatch",
true,
expected_crs_comp,
opal_crs_base_selected_component.base_version.mca_component_name,
ret);
exit_status = ret;
goto cleanup;
}
/******************************
* Restart in this process
******************************/
opal_output_verbose(10, opal_restart_globals.output,
"Restarting from file (%s)\n",
opal_restart_globals.snapshot_ref);
snapshot = OBJ_NEW(opal_crs_base_snapshot_t);
snapshot->cold_start = true;
opal_asprintf(&(snapshot->snapshot_directory), "%s/%s",
opal_restart_globals.snapshot_loc,
opal_restart_globals.snapshot_ref);
opal_asprintf(&(snapshot->metadata_filename), "%s/%s",
snapshot->snapshot_directory,
opal_restart_globals.snapshot_metadata);
/* Since some checkpoint/restart systems don't pass along env vars to the
* restarted app, we need to take care of that.
*
* Included here is the creation of any files or directories that need to be
* created before the process is restarted.
*/
if(OPAL_SUCCESS != (ret = post_env_vars(prev_pid, snapshot) ) ) {
exit_status = ret;
goto cleanup;
}
/*
* Do the actual restart
*/
ret = opal_crs.crs_restart(snapshot,
false,
&child_pid);
if (OPAL_SUCCESS != ret) {
opal_show_help("help-opal-restart.txt", "restart_cmd_failure", true,
opal_restart_globals.snapshot_ref,
ret,
opal_crs_base_selected_component.base_version.mca_component_name);
exit_status = ret;
goto cleanup;
}
/* Should never get here, since crs_restart calls exec */
/***************
* Cleanup
***************/
cleanup:
if (OPAL_SUCCESS != (ret = finalize())) {
return ret;
}
if(NULL != snapshot )
OBJ_DESTRUCT(snapshot);
return exit_status;
}
int orte_ess_base_orted_setup(char **hosts)
{
int ret = ORTE_ERROR;
int fd;
char log_file[PATH_MAX];
char *jobidstring;
char *error = NULL;
orte_job_t *jdata;
orte_proc_t *proc;
orte_app_context_t *app;
orte_node_t *node;
char *param;
plm_in_use = false;
/* setup callback for SIGPIPE */
setup_sighandler(SIGPIPE, &epipe_handler, epipe_signal_callback);
/* Set signal handlers to catch kill signals so we can properly clean up
* after ourselves.
*/
setup_sighandler(SIGTERM, &term_handler, shutdown_signal);
setup_sighandler(SIGINT, &int_handler, shutdown_signal);
/** setup callbacks for signals we should ignore */
setup_sighandler(SIGUSR1, &sigusr1_handler, signal_callback);
setup_sighandler(SIGUSR2, &sigusr2_handler, signal_callback);
signals_set = true;
#if OPAL_HAVE_HWLOC
{
hwloc_obj_t obj;
unsigned i, j;
/* get the local topology */
if (NULL == opal_hwloc_topology) {
if (OPAL_SUCCESS != opal_hwloc_base_get_topology()) {
error = "topology discovery";
goto error;
}
}
/* remove the hostname from the topology. Unfortunately, hwloc
* decided to add the source hostname to the "topology", thus
* rendering it unusable as a pure topological description. So
* we remove that information here.
*/
obj = hwloc_get_root_obj(opal_hwloc_topology);
for (i=0; i < obj->infos_count; i++) {
if (NULL == obj->infos[i].name ||
NULL == obj->infos[i].value) {
continue;
}
if (0 == strncmp(obj->infos[i].name, "HostName", strlen("HostName"))) {
free(obj->infos[i].name);
free(obj->infos[i].value);
/* left justify the array */
for (j=i; j < obj->infos_count-1; j++) {
obj->infos[j] = obj->infos[j+1];
}
obj->infos[obj->infos_count-1].name = NULL;
obj->infos[obj->infos_count-1].value = NULL;
obj->infos_count--;
break;
}
}
if (4 < opal_output_get_verbosity(orte_ess_base_framework.framework_output)) {
opal_output(0, "%s Topology Info:", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
opal_dss.dump(0, opal_hwloc_topology, OPAL_HWLOC_TOPO);
}
}
#endif
/* setup the global nidmap/pidmap object */
orte_nidmap.bytes = NULL;
orte_nidmap.size = 0;
orte_pidmap.bytes = NULL;
orte_pidmap.size = 0;
/* open and setup the opal_pstat framework so we can provide
* process stats if requested
*/
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&opal_pstat_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "opal_pstat_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = opal_pstat_base_select())) {
ORTE_ERROR_LOG(ret);
error = "opal_pstat_base_select";
goto error;
}
/* open and setup the state machine */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_state_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_state_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_state_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_state_base_select";
goto error;
}
/* open the errmgr */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_errmgr_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_errmgr_base_open";
goto error;
}
/* some environments allow remote launches - e.g., ssh - so
* open and select something -only- if we are given
* a specific module to use
*/
(void) mca_base_var_env_name("plm", ¶m);
plm_in_use = !!(getenv(param));
free (param);
if (plm_in_use) {
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_plm_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_plm_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_plm_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_plm_base_select";
goto error;
}
}
/* Setup the communication infrastructure */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_oob_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_oob_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_oob_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_oob_base_select";
goto error;
}
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_rml_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_rml_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_rml_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_rml_base_select";
goto error;
}
/* select the errmgr */
if (ORTE_SUCCESS != (ret = orte_errmgr_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_errmgr_base_select";
goto error;
}
/* Routed system */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_routed_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_rml_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_routed_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_routed_base_select";
goto error;
}
/* database */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&opal_db_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_db_base_open";
goto error;
}
/* always restrict daemons to local database components */
if (ORTE_SUCCESS != (ret = opal_db_base_select(true))) {
ORTE_ERROR_LOG(ret);
error = "orte_db_base_select";
goto error;
}
/* set our id */
opal_db.set_id((opal_identifier_t*)ORTE_PROC_MY_NAME);
/*
* Group communications
*/
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_grpcomm_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_grpcomm_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_grpcomm_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_grpcomm_base_select";
goto error;
}
/* Open/select the odls */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_odls_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_odls_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_odls_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_odls_base_select";
goto error;
}
/* enable communication with the rml */
if (ORTE_SUCCESS != (ret = orte_rml.enable_comm())) {
ORTE_ERROR_LOG(ret);
error = "orte_rml.enable_comm";
goto error;
}
/* initialize the nidmaps */
if (ORTE_SUCCESS != (ret = orte_util_nidmap_init(NULL))) {
ORTE_ERROR_LOG(ret);
error = "orte_util_nidmap_init";
goto error;
}
#if ORTE_ENABLE_STATIC_PORTS
/* if we are using static ports, then we need to setup
* the daemon info so the RML can function properly
* without requiring a wireup stage. This must be done
* after we enable_comm as that function determines our
* own port, which we need in order to construct the nidmap
*/
if (orte_static_ports) {
/* define the routing tree so we know the pattern
* if we are trying to setup common or static ports
*/
orte_routed.update_routing_plan();
/* extract the node info from the environment and
* build a nidmap from it
*/
if (ORTE_SUCCESS != (ret = orte_util_build_daemon_nidmap(hosts))) {
ORTE_ERROR_LOG(ret);
error = "construct daemon map from static ports";
goto error;
}
}
#endif
/* be sure to update the routing tree so the initial "phone home"
* to mpirun goes through the tree if static ports were enabled - still
* need to do it anyway just to initialize things
*/
orte_routed.update_routing_plan();
/* Now provide a chance for the PLM
* to perform any module-specific init functions. This
* needs to occur AFTER the communications are setup
* as it may involve starting a non-blocking recv
* Do this only if a specific PLM was given to us - the
* orted has no need of the proxy PLM at all
*/
if (plm_in_use) {
if (ORTE_SUCCESS != (ret = orte_plm.init())) {
ORTE_ERROR_LOG(ret);
error = "orte_plm_init";
goto error;
}
}
/* setup my session directory */
if (orte_create_session_dirs) {
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_framework.framework_output,
"%s setting up session dir with\n\ttmpdir: %s\n\thost %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == orte_process_info.tmpdir_base) ? "UNDEF" : orte_process_info.tmpdir_base,
orte_process_info.nodename));
/* take a pass thru the session directory code to fillin the
* tmpdir names - don't create anything yet
*/
if (ORTE_SUCCESS != (ret = orte_session_dir(false,
orte_process_info.tmpdir_base,
orte_process_info.nodename, NULL,
ORTE_PROC_MY_NAME))) {
ORTE_ERROR_LOG(ret);
error = "orte_session_dir define";
goto error;
}
/* clear the session directory just in case there are
* stale directories laying around
*/
orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);
/* now actually create the directory tree */
if (ORTE_SUCCESS != (ret = orte_session_dir(true,
orte_process_info.tmpdir_base,
orte_process_info.nodename, NULL,
ORTE_PROC_MY_NAME))) {
ORTE_ERROR_LOG(ret);
error = "orte_session_dir";
goto error;
}
/* Once the session directory location has been established, set
the opal_output env file location to be in the
proc-specific session directory. */
opal_output_set_output_file_info(orte_process_info.proc_session_dir,
"output-", NULL, NULL);
/* setup stdout/stderr */
if (orte_debug_daemons_file_flag) {
/* if we are debugging to a file, then send stdout/stderr to
* the orted log file
*/
/* get my jobid */
if (ORTE_SUCCESS != (ret = orte_util_convert_jobid_to_string(&jobidstring,
ORTE_PROC_MY_NAME->jobid))) {
ORTE_ERROR_LOG(ret);
error = "convert_jobid";
goto error;
}
/* define a log file name in the session directory */
snprintf(log_file, PATH_MAX, "output-orted-%s-%s.log",
jobidstring, orte_process_info.nodename);
log_path = opal_os_path(false,
orte_process_info.tmpdir_base,
orte_process_info.top_session_dir,
log_file,
NULL);
fd = open(log_path, O_RDWR|O_CREAT|O_TRUNC, 0640);
if (fd < 0) {
/* couldn't open the file for some reason, so
* just connect everything to /dev/null
*/
fd = open("/dev/null", O_RDWR|O_CREAT|O_TRUNC, 0666);
} else {
dup2(fd, STDOUT_FILENO);
dup2(fd, STDERR_FILENO);
if(fd != STDOUT_FILENO && fd != STDERR_FILENO) {
close(fd);
}
}
}
}
/* setup the global job and node arrays */
orte_job_data = OBJ_NEW(opal_pointer_array_t);
if (ORTE_SUCCESS != (ret = opal_pointer_array_init(orte_job_data,
1,
ORTE_GLOBAL_ARRAY_MAX_SIZE,
1))) {
ORTE_ERROR_LOG(ret);
error = "setup job array";
goto error;
}
orte_node_pool = OBJ_NEW(opal_pointer_array_t);
if (ORTE_SUCCESS != (ret = opal_pointer_array_init(orte_node_pool,
ORTE_GLOBAL_ARRAY_BLOCK_SIZE,
ORTE_GLOBAL_ARRAY_MAX_SIZE,
ORTE_GLOBAL_ARRAY_BLOCK_SIZE))) {
ORTE_ERROR_LOG(ret);
error = "setup node array";
goto error;
}
orte_node_topologies = OBJ_NEW(opal_pointer_array_t);
if (ORTE_SUCCESS != (ret = opal_pointer_array_init(orte_node_topologies,
ORTE_GLOBAL_ARRAY_BLOCK_SIZE,
ORTE_GLOBAL_ARRAY_MAX_SIZE,
ORTE_GLOBAL_ARRAY_BLOCK_SIZE))) {
ORTE_ERROR_LOG(ret);
error = "setup node topologies array";
goto error;
}
/* Setup the job data object for the daemons */
/* create and store the job data object */
jdata = OBJ_NEW(orte_job_t);
jdata->jobid = ORTE_PROC_MY_NAME->jobid;
opal_pointer_array_set_item(orte_job_data, 0, jdata);
/* every job requires at least one app */
app = OBJ_NEW(orte_app_context_t);
opal_pointer_array_set_item(jdata->apps, 0, app);
jdata->num_apps++;
/* create and store a node object where we are */
node = OBJ_NEW(orte_node_t);
node->name = strdup(orte_process_info.nodename);
node->index = opal_pointer_array_set_item(orte_node_pool, ORTE_PROC_MY_NAME->vpid, node);
#if OPAL_HAVE_HWLOC
/* point our topology to the one detected locally */
node->topology = opal_hwloc_topology;
#endif
/* create and store a proc object for us */
proc = OBJ_NEW(orte_proc_t);
proc->name.jobid = ORTE_PROC_MY_NAME->jobid;
proc->name.vpid = ORTE_PROC_MY_NAME->vpid;
proc->pid = orte_process_info.pid;
proc->rml_uri = orte_rml.get_contact_info();
proc->state = ORTE_PROC_STATE_RUNNING;
opal_pointer_array_set_item(jdata->procs, proc->name.vpid, proc);
/* record that the daemon (i.e., us) is on this node
* NOTE: we do not add the proc object to the node's
* proc array because we are not an application proc.
* Instead, we record it in the daemon field of the
* node object
*/
OBJ_RETAIN(proc); /* keep accounting straight */
node->daemon = proc;
node->daemon_launched = true;
node->state = ORTE_NODE_STATE_UP;
/* now point our proc node field to the node */
OBJ_RETAIN(node); /* keep accounting straight */
proc->node = node;
/* record that the daemon job is running */
jdata->num_procs = 1;
jdata->state = ORTE_JOB_STATE_RUNNING;
/* obviously, we have "reported" */
jdata->num_reported = 1;
/* setup the routed info - the selected routed component
* will know what to do.
*/
if (ORTE_SUCCESS != (ret = orte_routed.init_routes(ORTE_PROC_MY_NAME->jobid, NULL))) {
ORTE_ERROR_LOG(ret);
error = "orte_routed.init_routes";
goto error;
}
/* setup I/O forwarding system - must come after we init routes */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_iof_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_iof_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_iof_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_iof_base_select";
goto error;
}
/* setup the FileM */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_filem_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_filem_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_filem_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_filem_base_select";
goto error;
}
#if OPAL_ENABLE_FT_CR == 1
/*
* Setup the SnapC
*/
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_snapc_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_snapc_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_sstore_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_sstore_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_snapc_base_select(!ORTE_PROC_IS_HNP, ORTE_PROC_IS_DAEMON))) {
ORTE_ERROR_LOG(ret);
error = "orte_snapc_base_select";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_sstore_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_sstore_base_select";
goto error;
}
/* For daemons, ORTE doesn't need the OPAL CR stuff */
opal_cr_set_enabled(false);
#else
opal_cr_set_enabled(false);
#endif
/*
* Initalize the CR setup
* Note: Always do this, even in non-FT builds.
* If we don't some user level tools may hang.
*/
if (ORTE_SUCCESS != (ret = orte_cr_init())) {
ORTE_ERROR_LOG(ret);
error = "orte_cr_init";
goto error;
}
/* setup the DFS framework */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_dfs_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_dfs_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_dfs_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_dfs_select";
goto error;
}
return ORTE_SUCCESS;
error:
orte_show_help("help-orte-runtime.txt",
"orte_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
return ORTE_ERR_SILENT;
}
static int initialize(int argc, char *argv[])
{
int ret, exit_status = OPAL_SUCCESS;
char * tmp_env_var = NULL;
/*
* Make sure to init util before parse_args
* to ensure installdirs is setup properly
* before calling mca_base_open();
*/
if( OPAL_SUCCESS != (ret = opal_init_util(&argc, &argv)) ) {
return ret;
}
/*
* Parse Command line arguments
*/
if (OPAL_SUCCESS != (ret = parse_args(argc, argv))) {
exit_status = ret;
goto cleanup;
}
/*
* Setup OPAL Output handle from the verbose argument
*/
if( opal_restart_globals.verbose ) {
opal_restart_globals.output = opal_output_open(NULL);
opal_output_set_verbosity(opal_restart_globals.output, 10);
} else {
opal_restart_globals.output = 0; /* Default=STDOUT */
}
/*
* Turn off the selection of the CRS component,
* we need to do that later
*/
(void) mca_base_var_env_name("crs_base_do_not_select", &tmp_env_var);
opal_setenv(tmp_env_var,
"1", /* turn off the selection */
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/*
* Make sure we select the proper compress component.
*/
if( NULL != opal_restart_globals.snapshot_compress ) {
(void) mca_base_var_env_name("compress", &tmp_env_var);
opal_setenv(tmp_env_var,
opal_restart_globals.snapshot_compress,
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
}
/*
* Initialize the OPAL layer
*/
if (OPAL_SUCCESS != (ret = opal_init(&argc, &argv))) {
exit_status = ret;
goto cleanup;
}
/*
* If the checkpoint was compressed, then decompress it before continuing
*/
if( NULL != opal_restart_globals.snapshot_compress ) {
char * zip_dir = NULL;
char * tmp_str = NULL;
/* Make sure to clear the selection for the restart,
* this way the user can swich compression mechanism
* across restart
*/
(void) mca_base_var_env_name("compress", &tmp_env_var);
opal_unsetenv(tmp_env_var, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
opal_asprintf(&zip_dir, "%s/%s%s",
opal_restart_globals.snapshot_loc,
opal_restart_globals.snapshot_ref,
opal_restart_globals.snapshot_compress_postfix);
if (0 > (ret = access(zip_dir, F_OK)) ) {
opal_output(opal_restart_globals.output,
"Error: Unable to access the file [%s]!",
zip_dir);
exit_status = OPAL_ERROR;
goto cleanup;
}
opal_output_verbose(10, opal_restart_globals.output,
"Decompressing (%s)",
zip_dir);
opal_compress.decompress(zip_dir, &tmp_str);
if( NULL != zip_dir ) {
free(zip_dir);
zip_dir = NULL;
}
if( NULL != tmp_str ) {
free(tmp_str);
tmp_str = NULL;
}
}
/*
* If a cache directory has been suggested, see if it exists
*/
if( NULL != opal_restart_globals.snapshot_cache ) {
if(0 == (ret = access(opal_restart_globals.snapshot_cache, F_OK)) ) {
opal_output_verbose(10, opal_restart_globals.output,
"Using the cached snapshot (%s) instead of (%s)",
opal_restart_globals.snapshot_cache,
opal_restart_globals.snapshot_loc);
if( NULL != opal_restart_globals.snapshot_loc ) {
free(opal_restart_globals.snapshot_loc);
opal_restart_globals.snapshot_loc = NULL;
}
opal_restart_globals.snapshot_loc = opal_dirname(opal_restart_globals.snapshot_cache);
} else {
opal_show_help("help-opal-restart.txt", "cache_not_avail", true,
opal_restart_globals.snapshot_cache,
opal_restart_globals.snapshot_loc);
}
}
/*
* Mark this process as a tool
*/
opal_cr_is_tool = true;
cleanup:
return exit_status;
}
static int parse_args(int argc, char *argv[]) {
int i, ret, len, exit_status = ORTE_SUCCESS ;
opal_cmd_line_t cmd_line;
char **app_env = NULL, **global_env = NULL;
char * tmp_env_var = NULL;
char *argv0 = NULL;
/* Init structure */
memset(&orte_checkpoint_globals, 0, sizeof(orte_checkpoint_globals_t));
orte_checkpoint_globals.help = false;
orte_checkpoint_globals.pid = -1;
orte_checkpoint_globals.verbose = false;
orte_checkpoint_globals.verbose_level = 0;
orte_checkpoint_globals.req_hnp = ORTE_JOBID_INVALID;
orte_checkpoint_globals.nowait = false;
orte_checkpoint_globals.status = false;
orte_checkpoint_globals.output = -1;
orte_checkpoint_globals.ckpt_status = ORTE_SNAPC_CKPT_STATE_NONE;
orte_checkpoint_globals.list_only = false;
#if OPAL_ENABLE_CRDEBUG == 1
orte_checkpoint_globals.enable_crdebug = false;
#endif
orte_checkpoint_globals.options = OBJ_NEW(opal_crs_base_ckpt_options_t);
orte_checkpoint_globals.term = false;
orte_checkpoint_globals.stop = false;
#if OPAL_ENABLE_CRDEBUG == 1
orte_checkpoint_globals.attach_debugger = false;
orte_checkpoint_globals.detach_debugger = false;
#endif
#if OPAL_ENABLE_FT_CR == 0
/* Warn and exit if not configured with Checkpoint/Restart */
{
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-checkpoint.txt", "usage-no-cr",
true, args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
exit_status = ORTE_ERROR;
goto cleanup;
}
#endif
/* Parse the command line options */
opal_cmd_line_create(&cmd_line, cmd_line_opts);
mca_base_open();
mca_base_cmd_line_setup(&cmd_line);
ret = opal_cmd_line_parse(&cmd_line, true, argc, argv);
if (OPAL_SUCCESS != ret) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
exit_status = 1;
goto cleanup;
}
if (orte_checkpoint_globals.help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-checkpoint.txt", "usage", true,
args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
exit(0);
}
/**
* Put all of the MCA arguments in the environment
*/
mca_base_cmd_line_process_args(&cmd_line, &app_env, &global_env);
len = opal_argv_count(app_env);
for(i = 0; i < len; ++i) {
putenv(app_env[i]);
}
len = opal_argv_count(global_env);
for(i = 0; i < len; ++i) {
putenv(global_env[i]);
}
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/**
* Now start parsing our specific arguments
*/
/* get the remaining bits */
argv0 = strdup(argv[0]);
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
if(orte_checkpoint_globals.list_only ) {
exit_status = ORTE_SUCCESS;
goto cleanup;
}
if (0 >= argc && ORTE_JOBID_INVALID == orte_checkpoint_globals.req_hnp) {
fprintf(stderr, "%s: Nothing to do\n", argv0);
fprintf(stderr, "Type '%s --help' for usage.\n", argv0);
exit_status = 1;
goto cleanup;
}
orte_checkpoint_globals.options->term = orte_checkpoint_globals.term;
orte_checkpoint_globals.options->stop = orte_checkpoint_globals.stop;
#if OPAL_ENABLE_CRDEBUG == 1
orte_checkpoint_globals.options->attach_debugger = orte_checkpoint_globals.attach_debugger;
orte_checkpoint_globals.options->detach_debugger = orte_checkpoint_globals.detach_debugger;
#endif
if(orte_checkpoint_globals.verbose_level < 0 ) {
orte_checkpoint_globals.verbose_level = 0;
}
if(orte_checkpoint_globals.verbose_level > 0) {
orte_checkpoint_globals.verbose = true;
}
/*
* If the user did not supply an hnp jobid, then they must
* supply the PID of MPIRUN
*/
if(0 >= argc &&
ORTE_JOBID_INVALID != orte_checkpoint_globals.req_hnp) {
exit_status = ORTE_SUCCESS;
goto cleanup;
}
orte_checkpoint_globals.pid = atoi(argv[0]);
if ( 0 >= orte_checkpoint_globals.pid ) {
opal_show_help("help-orte-checkpoint.txt", "invalid_pid", true,
orte_checkpoint_globals.pid);
exit_status = ORTE_ERROR;
goto cleanup;
}
/*
* JJH: No wait is currently not implemented or tested
*/
if(orte_checkpoint_globals.nowait) {
orte_checkpoint_globals.nowait = false;
opal_show_help("help-orte-checkpoint.txt", "not_impl",
true,
"Disconnected checkpoint");
}
if(orte_checkpoint_globals.verbose) {
orte_checkpoint_globals.status = true;
}
cleanup:
if (NULL != argv0) {
free(argv0);
}
return exit_status;
}
static int parse_args(int argc, char *argv[])
{
int i, ret, len;
opal_cmd_line_t cmd_line;
char **app_env = NULL, **global_env = NULL;
char * tmp_env_var = NULL;
char *argv0 = NULL;
orte_restart_globals_t tmp = { false, /* help */
NULL, /* filename */
NULL, /* appfile */
false, /* verbose */
false, /* forked */
-1, /* seq_number */
NULL, /* hostfile */
-1, /* output*/
false, /* info only */
false, /* app only */
false, /* showme */
NULL}; /* mpirun_opts */
orte_restart_globals = tmp;
#if OPAL_ENABLE_CRDEBUG == 1
orte_restart_globals.enable_crdebug = false;
#endif
#if OPAL_ENABLE_FT_CR == 0
/* Warn and exit if not configured with Checkpoint/Restart */
{
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-restart.txt", "usage-no-cr",
true, args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
return ORTE_ERROR;
}
#endif
/* Parse the command line options */
opal_cmd_line_create(&cmd_line, cmd_line_opts);
mca_base_open();
mca_base_cmd_line_setup(&cmd_line);
ret = opal_cmd_line_parse(&cmd_line, true, argc, argv);
if (OPAL_SUCCESS != ret) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
return 1;
}
if (orte_restart_globals.help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-restart.txt", "usage", true,
args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
exit(0);
}
/**
* Put all of the MCA arguments in the environment
*/
mca_base_cmd_line_process_args(&cmd_line, &app_env, &global_env);
len = opal_argv_count(app_env);
for(i = 0; i < len; ++i) {
putenv(app_env[i]);
}
len = opal_argv_count(global_env);
for(i = 0; i < len; ++i) {
putenv(global_env[i]);
}
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/**
* Now start parsing our specific arguments
*/
/* get the remaining bits */
argv0 = strdup(argv[0]);
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
if (0 == argc) {
fprintf(stderr, "%s: Nothing to do\n", argv0);
fprintf(stderr, "Type '%s --help' for usge.\n", argv0);
free(argv0);
return ORTE_ERROR;
}
free(argv0);
orte_restart_globals.snapshot_ref = strdup(argv[0]);
if ( NULL == orte_restart_globals.snapshot_ref ||
0 >= strlen(orte_restart_globals.snapshot_ref) ) {
opal_show_help("help-orte-restart.txt", "invalid_filename", true,
"<none provided>");
return ORTE_ERROR;
}
/* If we have arguments after the command, then assume they
* need to be grouped together.
*/
if(argc > 1) {
orte_restart_globals.snapshot_ref = strdup(opal_argv_join(argv, ' '));
}
return ORTE_SUCCESS;
}
static int initialize(int argc, char *argv[]) {
int ret, exit_status = ORTE_SUCCESS;
char * tmp_env_var = NULL;
/*
* Make sure to init util before parse_args
* to ensure installdirs is setup properly
* before calling mca_base_open();
*/
if( ORTE_SUCCESS != (ret = opal_init_util(&argc, &argv)) ) {
return ret;
}
/*
* Parse command line arguments
*/
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
exit_status = ret;
goto cleanup;
}
/*
* Setup OPAL Output handle from the verbose argument
*/
if( orte_restart_globals.verbose ) {
orte_restart_globals.output = opal_output_open(NULL);
opal_output_set_verbosity(orte_restart_globals.output, 10);
} else {
orte_restart_globals.output = 0; /* Default=STDERR */
}
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
"none",
true, &environ);
/* Don't free the environment variable name. It is used again below */
/*
* Setup any ORTE stuff we might need
*/
if (OPAL_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_TOOL))) {
exit_status = ret;
goto cleanup;
}
/* Unset these now that we no longer need them */
opal_unsetenv(tmp_env_var, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_unsetenv(tmp_env_var, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
cleanup:
return exit_status;
}
static int do_child(orte_app_context_t* context,
orte_proc_t *child,
char **environ_copy,
orte_job_t *jobdat, int write_fd,
orte_iof_base_io_conf_t opts)
{
int i, rc;
sigset_t sigs;
long fd, fdmax = sysconf(_SC_OPEN_MAX);
char *param, *msg;
if (orte_forward_job_control) {
/* Set a new process group for this child, so that a
SIGSTOP can be sent to it without being sent to the
orted. */
setpgid(0, 0);
}
/* Setup the pipe to be close-on-exec */
fcntl(write_fd, F_SETFD, FD_CLOEXEC);
if (NULL != child) {
/* setup stdout/stderr so that any error messages that we
may print out will get displayed back at orterun.
NOTE: Definitely do this AFTER we check contexts so
that any error message from those two functions doesn't
come out to the user. IF we didn't do it in this order,
THEN a user who gives us a bad executable name or
working directory would get N error messages, where
N=num_procs. This would be very annoying for large
jobs, so instead we set things up so that orterun
always outputs a nice, single message indicating what
happened
*/
if (ORTE_SUCCESS != (i = orte_iof_base_setup_child(&opts,
&environ_copy))) {
ORTE_ERROR_LOG(i);
send_error_show_help(write_fd, 1,
"help-orte-odls-default.txt",
"iof setup failed",
orte_process_info.nodename, context->app);
/* Does not return */
}
#if OPAL_HAVE_HWLOC
{
hwloc_cpuset_t cpuset;
hwloc_obj_t root;
opal_hwloc_topo_data_t *sum;
/* Set process affinity, if given */
if (NULL == child->cpu_bitmap) {
/* if the daemon is bound, then we need to "free" this proc */
if (NULL != orte_daemon_cores) {
root = hwloc_get_root_obj(opal_hwloc_topology);
if (NULL == root->userdata) {
send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "incorrectly bound",
orte_process_info.nodename, context->app,
__FILE__, __LINE__);
}
sum = (opal_hwloc_topo_data_t*)root->userdata;
/* bind this proc to all available processors */
hwloc_set_cpubind(opal_hwloc_topology, sum->available, 0);
}
if (opal_hwloc_report_bindings) {
opal_output(0, "MCW rank %d is not bound (or bound to all available processors)", child->name.vpid);
/* avoid reporting it twice */
(void) mca_base_var_env_name ("hwloc_base_report_bindings", ¶m);
opal_unsetenv(param, &environ_copy);
free(param);
}
} else {
if (0 == strlen(child->cpu_bitmap)) {
/* this proc is not bound */
if (opal_hwloc_report_bindings) {
opal_output(0, "MCW rank %d is not bound (or bound to all available processors)", child->name.vpid);
/* avoid reporting it twice */
(void) mca_base_var_env_name ("hwloc_base_report_bindings", ¶m);
opal_unsetenv(param, &environ_copy);
free(param);
}
/* if the daemon is bound, then we need to "free" this proc */
if (NULL != orte_daemon_cores) {
root = hwloc_get_root_obj(opal_hwloc_topology);
if (NULL == root->userdata) {
send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "incorrectly bound",
orte_process_info.nodename, context->app,
__FILE__, __LINE__);
}
sum = (opal_hwloc_topo_data_t*)root->userdata;
/* bind this proc to all available processors */
hwloc_set_cpubind(opal_hwloc_topology, sum->available, 0);
}
/* provide a nice string representation of what we bound to */
(void) mca_base_var_env_name ("orte_base_applied_binding", ¶m);
opal_setenv(param, child->cpu_bitmap, true, &environ_copy);
free(param);
goto PROCEED;
}
/* convert the list to a cpu bitmap */
cpuset = hwloc_bitmap_alloc();
if (0 != (rc = hwloc_bitmap_list_sscanf(cpuset, child->cpu_bitmap))) {
/* See comment above about "This may be a small memory leak" */
asprintf(&msg, "hwloc_bitmap_sscanf returned \"%s\" for the string \"%s\"",
opal_strerror(rc), child->cpu_bitmap);
if (NULL == msg) {
msg = "failed to convert bitmap list to hwloc bitmap";
}
if (OPAL_BINDING_REQUIRED(jobdat->map->binding) &&
OPAL_BINDING_POLICY_IS_SET(jobdat->map->binding)) {
/* If binding is required and a binding directive was explicitly
* given (i.e., we are not binding due to a default policy),
* send an error up the pipe (which exits -- it doesn't return).
*/
send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"binding generic error",
orte_process_info.nodename,
context->app, msg,
__FILE__, __LINE__);
} else {
send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "not bound",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
goto PROCEED;
}
}
/* bind as specified */
rc = hwloc_set_cpubind(opal_hwloc_topology, cpuset, 0);
/* if we got an error and this wasn't a default binding policy, then report it */
if (rc < 0 && OPAL_BINDING_POLICY_IS_SET(jobdat->map->binding)) {
char *tmp = NULL;
if (errno == ENOSYS) {
msg = "hwloc indicates cpu binding not supported";
} else if (errno == EXDEV) {
msg = "hwloc indicates cpu binding cannot be enforced";
} else {
asprintf(&msg, "hwloc_set_cpubind returned \"%s\" for bitmap \"%s\"",
opal_strerror(rc), child->cpu_bitmap);
}
if (OPAL_BINDING_REQUIRED(jobdat->map->binding)) {
/* If binding is required, send an error up the pipe (which exits
-- it doesn't return). */
send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"binding generic error",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
} else {
send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "not bound",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
if (NULL != tmp) {
free(tmp);
free(msg);
}
goto PROCEED;
}
if (NULL != tmp) {
free(tmp);
free(msg);
}
}
if (0 == rc && opal_hwloc_report_bindings) {
char tmp1[1024], tmp2[1024];
hwloc_cpuset_t mycpus;
/* get the cpus we are bound to */
mycpus = hwloc_bitmap_alloc();
if (hwloc_get_cpubind(opal_hwloc_topology,
mycpus,
HWLOC_CPUBIND_PROCESS) < 0) {
opal_output(0, "MCW rank %d is not bound",
child->name.vpid);
} else {
if (OPAL_ERR_NOT_BOUND == opal_hwloc_base_cset2str(tmp1, sizeof(tmp1), opal_hwloc_topology, mycpus)) {
opal_output(0, "MCW rank %d is not bound (or bound to all available processors)", child->name.vpid);
} else {
opal_hwloc_base_cset2mapstr(tmp2, sizeof(tmp2), opal_hwloc_topology, mycpus);
opal_output(0, "MCW rank %d bound to %s: %s",
child->name.vpid, tmp1, tmp2);
}
}
hwloc_bitmap_free(mycpus);
/* avoid reporting it twice */
(void) mca_base_var_env_name ("hwloc_base_report_bindings", ¶m);
opal_unsetenv(param, &environ_copy);
free(param);
}
/* set memory affinity policy - if we get an error, don't report
* anything unless the user actually specified the binding policy
*/
rc = opal_hwloc_base_set_process_membind_policy();
if (ORTE_SUCCESS != rc && OPAL_BINDING_POLICY_IS_SET(jobdat->map->binding)) {
if (errno == ENOSYS) {
msg = "hwloc indicates memory binding not supported";
} else if (errno == EXDEV) {
msg = "hwloc indicates memory binding cannot be enforced";
} else {
msg = "failed to bind memory";
}
if (OPAL_HWLOC_BASE_MBFA_ERROR == opal_hwloc_base_mbfa) {
/* If binding is required, send an error up the pipe (which exits
-- it doesn't return). */
send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"memory binding error",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
} else {
send_warn_show_help(write_fd,
"help-orte-odls-default.txt", "memory not bound",
orte_process_info.nodename, context->app, msg,
__FILE__, __LINE__);
goto PROCEED;
}
}
}
}
#endif
} else if (!(ORTE_JOB_CONTROL_FORWARD_OUTPUT & jobdat->controls)) {
/* tie stdin/out/err/internal to /dev/null */
int fdnull;
for (i=0; i < 3; i++) {
fdnull = open("/dev/null", O_RDONLY, 0);
if (fdnull > i && i != write_fd) {
dup2(fdnull, i);
}
close(fdnull);
}
fdnull = open("/dev/null", O_RDONLY, 0);
if (fdnull > opts.p_internal[1]) {
dup2(fdnull, opts.p_internal[1]);
}
close(fdnull);
}
#if OPAL_HAVE_HWLOC
PROCEED:
#endif
/* if the user requested it, set the system resource limits */
if (OPAL_SUCCESS != (rc = opal_util_init_sys_limits(&msg))) {
send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"set limit",
orte_process_info.nodename, context->app,
__FILE__, __LINE__, msg);
}
/* ensure we only do this once */
(void) mca_base_var_env_name("opal_set_max_sys_limits", ¶m);
opal_unsetenv(param, &environ_copy);
free(param);
/* close all file descriptors w/ exception of stdin/stdout/stderr,
the pipe used for the IOF INTERNAL messages, and the pipe up to
the parent. */
for(fd=3; fd<fdmax; fd++) {
if (fd != opts.p_internal[1] && fd != write_fd) {
close(fd);
}
}
if (context->argv == NULL) {
context->argv = malloc(sizeof(char*)*2);
context->argv[0] = strdup(context->app);
context->argv[1] = NULL;
}
/* Set signal handlers back to the default. Do this close to
the exev() 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 launched process
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 launched process. */
sigprocmask(0, 0, &sigs);
sigprocmask(SIG_UNBLOCK, &sigs, 0);
/* Exec the new executable */
if (10 < opal_output_get_verbosity(orte_odls_base_framework.framework_output)) {
int jout;
opal_output(0, "%s STARTING %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), context->app);
for (jout=0; NULL != context->argv[jout]; jout++) {
opal_output(0, "%s\tARGV[%d]: %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), jout, context->argv[jout]);
}
for (jout=0; NULL != environ_copy[jout]; jout++) {
opal_output(0, "%s\tENVIRON[%d]: %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), jout, environ_copy[jout]);
}
}
execve(context->app, context->argv, environ_copy);
send_error_show_help(write_fd, 1,
"help-orte-odls-default.txt", "execve error",
orte_process_info.nodename, context->app, strerror(errno));
/* Does not return */
}
int main(int argc, char *argv[])
{
int ret = 0;
opal_cmd_line_t *cmd_line = NULL;
char *rml_uri;
#if OPAL_ENABLE_FT_CR == 1
char * tmp_env_var = NULL;
#endif
/* init enough of opal to process cmd lines */
if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
fprintf(stderr, "OPAL failed to initialize -- orted aborting\n");
exit(1);
}
/* setup to check common command line options that just report and die */
cmd_line = OBJ_NEW(opal_cmd_line_t);
opal_cmd_line_create(cmd_line, orte_server_cmd_line_opts);
mca_base_cmd_line_setup(cmd_line);
if (OPAL_SUCCESS != (ret = opal_cmd_line_parse(cmd_line, false, false,
argc, argv))) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
return 1;
}
/* check for help request */
if (help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
str = opal_show_help_string("help-orte-server.txt",
"orteserver:usage", false,
argv[0], args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
return 0;
}
/*
* Since this process can now handle MCA/GMCA parameters, make sure to
* process them.
*/
mca_base_cmd_line_process_args(cmd_line, &environ, &environ);
/* if debug is set, then set orte_debug_flag so that the data server
* code will output
*/
if (debug) {
putenv(OPAL_MCA_PREFIX"orte_debug=1");
}
/* detach from controlling terminal
* otherwise, remain attached so output can get to us
*/
if(debug == false &&
no_daemonize == false) {
opal_daemon_init(NULL);
}
#if OPAL_ENABLE_FT_CR == 1
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/* Mark as a tool program */
(void) mca_base_var_env_name("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#endif
/* don't want session directories */
orte_create_session_dirs = false;
/* Perform the standard init, but flag that we are an HNP */
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_HNP))) {
fprintf(stderr, "orte-server: failed to initialize -- aborting\n");
exit(1);
}
/* report out our URI, if we were requested to do so, using syntax
* proposed in an email thread by Jeff Squyres
*/
if (NULL != report_uri) {
orte_oob_base_get_addr(&rml_uri);
if (0 == strcmp(report_uri, "-")) {
/* if '-', then output to stdout */
printf("%s\n", rml_uri);
} else if (0 == strcmp(report_uri, "+")) {
/* if '+', output to stderr */
fprintf(stderr, "%s\n", rml_uri);
} else {
/* treat it as a filename and output into it */
FILE *fp;
fp = fopen(report_uri, "w");
if (NULL == fp) {
fprintf(stderr, "orte-server: failed to open designated file %s -- aborting\n", report_uri);
orte_finalize();
exit(1);
}
fprintf(fp, "%s\n", rml_uri);
fclose(fp);
}
free(rml_uri);
}
/* setup the data server to listen for commands */
if (ORTE_SUCCESS != (ret = orte_data_server_init())) {
fprintf(stderr, "orte-server: failed to start data server -- aborting\n");
orte_finalize();
exit(1);
}
/* setup to listen for commands sent specifically to me */
orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DAEMON,
ORTE_RML_NON_PERSISTENT, orte_daemon_recv, NULL);
/* Set signal handlers to catch kill signals so we can properly clean up
* after ourselves.
*/
opal_event_set(orte_event_base, &term_handler, SIGTERM, OPAL_EV_SIGNAL,
shutdown_callback, NULL);
opal_event_add(&term_handler, NULL);
opal_event_set(orte_event_base, &int_handler, SIGINT, OPAL_EV_SIGNAL,
shutdown_callback, NULL);
opal_event_add(&int_handler, NULL);
/* We actually do *not* want the server to voluntarily yield() the
processor more than necessary. The server already blocks when
it is doing nothing, so it doesn't use any more CPU cycles than
it should; but when it *is* doing something, we do not want it
to be unnecessarily delayed because it voluntarily yielded the
processor in the middle of its work.
For example: when a message arrives at the server, we want the
OS to wake up the server in a timely fashion (which most OS's
seem good about doing) and then we want the server to process
the message as fast as possible. If the server yields and lets
aggressive MPI applications get the processor back, it may be a
long time before the OS schedules the server to run again
(particularly if there is no IO event to wake it up). Hence,
publish and lookup (for example) may be significantly delayed
before being delivered to MPI processes, which can be
problematic in some scenarios (e.g., COMM_SPAWN). */
opal_progress_set_yield_when_idle(false);
/* Change the default behavior of libevent such that we want to
continually block rather than blocking for the default timeout
and then looping around the progress engine again. There
should be nothing in the server that cannot block in libevent
until "something" happens (i.e., there's no need to keep
cycling through progress because the only things that should
happen will happen in libevent). This is a minor optimization,
but what the heck... :-) */
opal_progress_set_event_flag(OPAL_EVLOOP_ONCE);
if (debug) {
opal_output(0, "%s orte-server: up and running!", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
}
/* wait to hear we are done */
while (orte_event_base_active) {
opal_event_loop(orte_event_base, OPAL_EVLOOP_ONCE);
}
ORTE_ACQUIRE_OBJECT(orte_event_base_active);
/* should never get here, but if we do... */
/* Finalize and clean up ourselves */
orte_finalize();
return orte_exit_status;
}
int orte_daemon(int argc, char *argv[])
{
int ret = 0;
opal_cmd_line_t *cmd_line = NULL;
char *rml_uri;
int i;
opal_buffer_t *buffer;
char hostname[100];
#if OPAL_ENABLE_FT_CR == 1
char *tmp_env_var = NULL;
#endif
/* initialize the globals */
memset(&orted_globals, 0, sizeof(orted_globals));
/* initialize the singleton died pipe to an illegal value so we can detect it was set */
orted_globals.singleton_died_pipe = -1;
/* init the failure orted vpid to an invalid value */
orted_globals.fail = ORTE_VPID_INVALID;
/* setup to check common command line options that just report and die */
cmd_line = OBJ_NEW(opal_cmd_line_t);
if (OPAL_SUCCESS != opal_cmd_line_create(cmd_line, orte_cmd_line_opts)) {
OBJ_RELEASE(cmd_line);
exit(1);
}
mca_base_cmd_line_setup(cmd_line);
if (ORTE_SUCCESS != (ret = opal_cmd_line_parse(cmd_line, false,
argc, argv))) {
char *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
fprintf(stderr, "Usage: %s [OPTION]...\n%s\n", argv[0], args);
free(args);
OBJ_RELEASE(cmd_line);
return ret;
}
/*
* Since this process can now handle MCA/GMCA parameters, make sure to
* process them.
*/
mca_base_cmd_line_process_args(cmd_line, &environ, &environ);
/* Ensure that enough of OPAL is setup for us to be able to run */
/*
* NOTE: (JJH)
* We need to allow 'mca_base_cmd_line_process_args()' to process command
* line arguments *before* calling opal_init_util() since the command
* line could contain MCA parameters that affect the way opal_init_util()
* functions. AMCA parameters are one such option normally received on the
* command line that affect the way opal_init_util() behaves.
* It is "safe" to call mca_base_cmd_line_process_args() before
* opal_init_util() since mca_base_cmd_line_process_args() does *not*
* depend upon opal_init_util() functionality.
*/
if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
fprintf(stderr, "OPAL failed to initialize -- orted aborting\n");
exit(1);
}
/* save the environment for launch purposes. This MUST be
* done so that we can pass it to any local procs we
* spawn - otherwise, those local procs won't see any
* non-MCA envars that were set in the enviro when the
* orted was executed - e.g., by .csh
*/
orte_launch_environ = opal_argv_copy(environ);
/* purge any ess flag set in the environ when we were launched */
opal_unsetenv(OPAL_MCA_PREFIX"ess", &orte_launch_environ);
/* if orte_daemon_debug is set, let someone know we are alive right
* away just in case we have a problem along the way
*/
if (orted_globals.debug) {
gethostname(hostname, 100);
fprintf(stderr, "Daemon was launched on %s - beginning to initialize\n", hostname);
}
/* check for help request */
if (orted_globals.help) {
char *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
orte_show_help("help-orted.txt", "orted:usage", false,
argv[0], args);
free(args);
return 1;
}
#if defined(HAVE_SETSID)
/* see if we were directed to separate from current session */
if (orted_globals.set_sid) {
setsid();
}
#endif
/* see if they want us to spin until they can connect a debugger to us */
i=0;
while (orted_spin_flag) {
i++;
if (1000 < i) i=0;
}
#if OPAL_ENABLE_FT_CR == 1
/* Mark as a tool program */
(void) mca_base_var_env_name ("opal_cr_is_tool", &tmp_env_var);
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#endif
/* if mapreduce set, flag it */
if (orted_globals.mapreduce) {
orte_map_reduce = true;
}
/* detach from controlling terminal
* otherwise, remain attached so output can get to us
*/
if(!orte_debug_flag &&
!orte_debug_daemons_flag &&
orted_globals.daemonize) {
opal_daemon_init(NULL);
}
/* Set the flag telling OpenRTE that I am NOT a
* singleton, but am "infrastructure" - prevents setting
* up incorrect infrastructure that only a singleton would
* require.
*/
if (orted_globals.hnp) {
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_HNP))) {
ORTE_ERROR_LOG(ret);
return ret;
}
} else {
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_DAEMON))) {
ORTE_ERROR_LOG(ret);
return ret;
}
}
/* finalize the OPAL utils. As they are opened again from orte_init->opal_init
* we continue to have a reference count on them. So we have to finalize them twice...
*/
opal_finalize_util();
#if OPAL_HAVE_HWLOC
/* bind ourselves if so directed */
if (NULL != orte_daemon_cores) {
char **cores=NULL, tmp[128];
hwloc_obj_t pu;
hwloc_cpuset_t ours, pucpus, res;
int core;
/* could be a collection of comma-delimited ranges, so
* use our handy utility to parse it
*/
orte_util_parse_range_options(orte_daemon_cores, &cores);
if (NULL != cores) {
ours = hwloc_bitmap_alloc();
hwloc_bitmap_zero(ours);
pucpus = hwloc_bitmap_alloc();
res = hwloc_bitmap_alloc();
for (i=0; NULL != cores[i]; i++) {
core = strtoul(cores[i], NULL, 10);
if (NULL == (pu = opal_hwloc_base_get_pu(opal_hwloc_topology, core, OPAL_HWLOC_LOGICAL))) {
/* turn off the show help forwarding as we won't
* be able to cycle the event library to send
*/
orte_show_help_finalize();
/* the message will now come out locally */
orte_show_help("help-orted.txt", "orted:cannot-bind",
true, orte_process_info.nodename,
orte_daemon_cores);
ret = ORTE_ERR_NOT_SUPPORTED;
goto DONE;
}
hwloc_bitmap_and(pucpus, pu->online_cpuset, pu->allowed_cpuset);
hwloc_bitmap_or(res, ours, pucpus);
hwloc_bitmap_copy(ours, res);
}
/* if the result is all zeros, then don't bind */
if (!hwloc_bitmap_iszero(ours)) {
(void)hwloc_set_cpubind(opal_hwloc_topology, ours, 0);
if (opal_hwloc_report_bindings) {
opal_hwloc_base_cset2mapstr(tmp, sizeof(tmp), opal_hwloc_topology, ours);
opal_output(0, "Daemon %s is bound to cores %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), tmp);
}
}
/* cleanup */
hwloc_bitmap_free(ours);
hwloc_bitmap_free(pucpus);
hwloc_bitmap_free(res);
opal_argv_free(cores);
}
}
#endif
if ((int)ORTE_VPID_INVALID != orted_globals.fail) {
orted_globals.abort=false;
/* some vpid was ordered to fail. The value can be positive
* or negative, depending upon the desired method for failure,
* so need to check both here
*/
if (0 > orted_globals.fail) {
orted_globals.fail = -1*orted_globals.fail;
orted_globals.abort = true;
}
/* are we the specified vpid? */
if ((int)ORTE_PROC_MY_NAME->vpid == orted_globals.fail) {
/* if the user specified we delay, then setup a timer
* and have it kill us
*/
if (0 < orted_globals.fail_delay) {
ORTE_TIMER_EVENT(orted_globals.fail_delay, 0, shutdown_callback, ORTE_SYS_PRI);
} else {
opal_output(0, "%s is executing clean %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
orted_globals.abort ? "abort" : "abnormal termination");
/* do -not- call finalize as this will send a message to the HNP
* indicating clean termination! Instead, just forcibly cleanup
* the local session_dir tree and exit
*/
orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);
/* if we were ordered to abort, do so */
if (orted_globals.abort) {
abort();
}
/* otherwise, return with non-zero status */
ret = ORTE_ERROR_DEFAULT_EXIT_CODE;
goto DONE;
}
}
}
/* insert our contact info into our process_info struct so we
* have it for later use and set the local daemon field to our name
*/
orte_process_info.my_daemon_uri = orte_rml.get_contact_info();
ORTE_PROC_MY_DAEMON->jobid = ORTE_PROC_MY_NAME->jobid;
ORTE_PROC_MY_DAEMON->vpid = ORTE_PROC_MY_NAME->vpid;
/* if I am also the hnp, then update that contact info field too */
if (ORTE_PROC_IS_HNP) {
orte_process_info.my_hnp_uri = orte_rml.get_contact_info();
ORTE_PROC_MY_HNP->jobid = ORTE_PROC_MY_NAME->jobid;
ORTE_PROC_MY_HNP->vpid = ORTE_PROC_MY_NAME->vpid;
}
/* setup the primary daemon command receive function */
orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DAEMON,
ORTE_RML_PERSISTENT, orte_daemon_recv, NULL);
/* output a message indicating we are alive, our name, and our pid
* for debugging purposes
*/
if (orte_debug_daemons_flag) {
fprintf(stderr, "Daemon %s checking in as pid %ld on host %s\n",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), (long)orte_process_info.pid,
orte_process_info.nodename);
}
/* We actually do *not* want the orted to voluntarily yield() the
processor more than necessary. The orted already blocks when
it is doing nothing, so it doesn't use any more CPU cycles than
it should; but when it *is* doing something, we do not want it
to be unnecessarily delayed because it voluntarily yielded the
processor in the middle of its work.
For example: when a message arrives at the orted, we want the
OS to wake up the orted in a timely fashion (which most OS's
seem good about doing) and then we want the orted to process
the message as fast as possible. If the orted yields and lets
aggressive MPI applications get the processor back, it may be a
long time before the OS schedules the orted to run again
(particularly if there is no IO event to wake it up). Hence,
routed OOB messages (for example) may be significantly delayed
before being delivered to MPI processes, which can be
problematic in some scenarios (e.g., COMM_SPAWN, BTL's that
require OOB messages for wireup, etc.). */
opal_progress_set_yield_when_idle(false);
/* Change the default behavior of libevent such that we want to
continually block rather than blocking for the default timeout
and then looping around the progress engine again. There
should be nothing in the orted that cannot block in libevent
until "something" happens (i.e., there's no need to keep
cycling through progress because the only things that should
happen will happen in libevent). This is a minor optimization,
but what the heck... :-) */
opal_progress_set_event_flag(OPAL_EVLOOP_ONCE);
/* if requested, report my uri to the indicated pipe */
if (orted_globals.uri_pipe > 0) {
orte_job_t *jdata;
orte_proc_t *proc;
orte_node_t *node;
orte_app_context_t *app;
char *tmp, *nptr, *sysinfo;
int32_t ljob;
/* setup the singleton's job */
jdata = OBJ_NEW(orte_job_t);
orte_plm_base_create_jobid(jdata);
ljob = ORTE_LOCAL_JOBID(jdata->jobid);
opal_pointer_array_set_item(orte_job_data, ljob, jdata);
/* must create a map for it (even though it has no
* info in it) so that the job info will be picked
* up in subsequent pidmaps or other daemons won't
* know how to route
*/
jdata->map = OBJ_NEW(orte_job_map_t);
/* setup an app_context for the singleton */
app = OBJ_NEW(orte_app_context_t);
app->app = strdup("singleton");
app->num_procs = 1;
opal_pointer_array_add(jdata->apps, app);
/* setup a proc object for the singleton - since we
* -must- be the HNP, and therefore we stored our
* node on the global node pool, and since the singleton
* -must- be on the same node as us, indicate that
*/
proc = OBJ_NEW(orte_proc_t);
proc->name.jobid = jdata->jobid;
proc->name.vpid = 0;
ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_ALIVE);
proc->state = ORTE_PROC_STATE_RUNNING;
proc->app_idx = 0;
/* obviously, it is on my node */
node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, 0);
proc->node = node;
OBJ_RETAIN(node); /* keep accounting straight */
opal_pointer_array_add(jdata->procs, proc);
jdata->num_procs = 1;
/* and it obviously is on the node */
OBJ_RETAIN(proc);
opal_pointer_array_add(node->procs, proc);
node->num_procs++;
/* and obviously it is one of my local procs */
OBJ_RETAIN(proc);
opal_pointer_array_add(orte_local_children, proc);
jdata->num_local_procs = 1;
/* set the trivial */
proc->local_rank = 0;
proc->node_rank = 0;
proc->app_rank = 0;
proc->state = ORTE_PROC_STATE_RUNNING;
proc->app_idx = 0;
ORTE_FLAG_SET(proc, ORTE_PROC_FLAG_LOCAL);
/* create a string that contains our uri + sysinfo + PMIx server URI */
orte_util_convert_sysinfo_to_string(&sysinfo, orte_local_cpu_type, orte_local_cpu_model);
asprintf(&tmp, "%s[%s]%s", orte_process_info.my_daemon_uri, sysinfo, pmix_server_uri);
free(sysinfo);
/* pass that info to the singleton */
write(orted_globals.uri_pipe, tmp, strlen(tmp)+1); /* need to add 1 to get the NULL */
/* cleanup */
free(tmp);
/* since a singleton spawned us, we need to harvest
* any MCA params from the local environment so
* we can pass them along to any subsequent daemons
* we may start as the result of a comm_spawn
*/
for (i=0; NULL != environ[i]; i++) {
if (0 == strncmp(environ[i], OPAL_MCA_PREFIX, 9)) {
/* make a copy to manipulate */
tmp = strdup(environ[i]);
/* find the equal sign */
nptr = strchr(tmp, '=');
*nptr = '\0';
nptr++;
/* add the mca param to the orted cmd line */
opal_argv_append_nosize(&orted_cmd_line, "-"OPAL_MCA_CMD_LINE_ID);
opal_argv_append_nosize(&orted_cmd_line, &tmp[9]);
opal_argv_append_nosize(&orted_cmd_line, nptr);
free(tmp);
}
}
}
/* if we were given a pipe to monitor for singleton termination, set that up */
if (orted_globals.singleton_died_pipe > 0) {
/* register shutdown handler */
pipe_handler = (opal_event_t*)malloc(sizeof(opal_event_t));
opal_event_set(orte_event_base, pipe_handler,
orted_globals.singleton_died_pipe,
OPAL_EV_READ,
pipe_closed,
pipe_handler);
opal_event_add(pipe_handler, NULL);
}
/* If I have a parent, then save his contact info so
* any messages we send can flow thru him.
*/
orte_parent_uri = NULL;
(void) mca_base_var_register ("orte", "orte", NULL, "parent_uri",
"URI for the parent if tree launch is enabled.",
MCA_BASE_VAR_TYPE_STRING, NULL, 0,
MCA_BASE_VAR_FLAG_INTERNAL,
OPAL_INFO_LVL_9,
MCA_BASE_VAR_SCOPE_CONSTANT,
&orte_parent_uri);
if (NULL != orte_parent_uri) {
orte_process_name_t parent;
/* set the contact info into the hash table */
orte_rml.set_contact_info(orte_parent_uri);
ret = orte_rml_base_parse_uris(orte_parent_uri, &parent, NULL);
if (ORTE_SUCCESS != ret) {
ORTE_ERROR_LOG(ret);
free (orte_parent_uri);
orte_parent_uri = NULL;
goto DONE;
}
/* don't need this value anymore */
free(orte_parent_uri);
orte_parent_uri = NULL;
/* tell the routed module that we have a path
* back to the HNP
*/
if (ORTE_SUCCESS != (ret = orte_routed.update_route(ORTE_PROC_MY_HNP, &parent))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* set the lifeline to point to our parent so that we
* can handle the situation if that lifeline goes away
*/
if (ORTE_SUCCESS != (ret = orte_routed.set_lifeline(&parent))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
}
/* if we are not the HNP...the only time we will be an HNP
* is if we are launched by a singleton to provide support
* for it
*/
if (!ORTE_PROC_IS_HNP) {
/* send the information to the orted report-back point - this function
* will process the data, but also counts the number of
* orteds that reported back so the launch procedure can continue.
* We need to do this at the last possible second as the HNP
* can turn right around and begin issuing orders to us
*/
buffer = OBJ_NEW(opal_buffer_t);
/* insert our name for rollup purposes */
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, ORTE_PROC_MY_NAME, 1, ORTE_NAME))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
/* for now, always include our contact info, even if we are using
* static ports. Eventually, this will be removed
*/
rml_uri = orte_rml.get_contact_info();
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &rml_uri, 1, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
/* include our node name */
opal_dss.pack(buffer, &orte_process_info.nodename, 1, OPAL_STRING);
/* if requested, include any non-loopback aliases for this node */
if (orte_retain_aliases) {
char **aliases=NULL;
uint8_t naliases, ni;
char hostname[ORTE_MAX_HOSTNAME_SIZE];
/* if we stripped the prefix or removed the fqdn,
* include full hostname as an alias
*/
gethostname(hostname, ORTE_MAX_HOSTNAME_SIZE);
if (strlen(orte_process_info.nodename) < strlen(hostname)) {
opal_argv_append_nosize(&aliases, hostname);
}
opal_ifgetaliases(&aliases);
naliases = opal_argv_count(aliases);
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &naliases, 1, OPAL_UINT8))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
for (ni=0; ni < naliases; ni++) {
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &aliases[ni], 1, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
}
opal_argv_free(aliases);
}
#if OPAL_HAVE_HWLOC
{
char *coprocessors;
/* add the local topology */
if (NULL != opal_hwloc_topology &&
(1 == ORTE_PROC_MY_NAME->vpid || orte_hetero_nodes)) {
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &opal_hwloc_topology, 1, OPAL_HWLOC_TOPO))) {
ORTE_ERROR_LOG(ret);
}
}
/* detect and add any coprocessors */
coprocessors = opal_hwloc_base_find_coprocessors(opal_hwloc_topology);
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &coprocessors, 1, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
}
/* see if I am on a coprocessor */
coprocessors = opal_hwloc_base_check_on_coprocessor();
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &coprocessors, 1, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
}
}
#endif
/* send to the HNP's callback - will be routed if routes are available */
if (0 > (ret = orte_rml.send_buffer_nb(ORTE_PROC_MY_HNP, buffer,
ORTE_RML_TAG_ORTED_CALLBACK,
orte_rml_send_callback, NULL))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
}
/* if we are tree-spawning, then we need to capture the MCA params
* from our cmd line so we can pass them along to the daemons we spawn -
* otherwise, only the first layer of daemons will ever see them
*/
if (orted_globals.tree_spawn) {
int j, k;
bool ignore;
char *no_keep[] = {
"orte_hnp_uri",
"orte_ess_jobid",
"orte_ess_vpid",
"orte_ess_num_procs",
"orte_parent_uri",
"mca_base_env_list",
NULL
};
for (i=0; i < argc; i++) {
if (0 == strcmp("-"OPAL_MCA_CMD_LINE_ID, argv[i]) ||
0 == strcmp("--"OPAL_MCA_CMD_LINE_ID, argv[i]) ) {
ignore = false;
/* see if this is something we cannot pass along */
for (k=0; NULL != no_keep[k]; k++) {
if (0 == strcmp(no_keep[k], argv[i+1])) {
ignore = true;
break;
}
}
if (!ignore) {
/* see if this is already present so we at least can
* avoid growing the cmd line with duplicates
*/
if (NULL != orted_cmd_line) {
for (j=0; NULL != orted_cmd_line[j]; j++) {
if (0 == strcmp(argv[i+1], orted_cmd_line[j])) {
/* already here - ignore it */
ignore = true;
break;
}
}
}
if (!ignore) {
opal_argv_append_nosize(&orted_cmd_line, argv[i]);
opal_argv_append_nosize(&orted_cmd_line, argv[i+1]);
opal_argv_append_nosize(&orted_cmd_line, argv[i+2]);
}
}
i += 2;
}
}
}
if (orte_debug_daemons_flag) {
opal_output(0, "%s orted: up and running - waiting for commands!", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
}
ret = ORTE_SUCCESS;
/* loop the event lib until an exit event is detected */
while (orte_event_base_active) {
opal_event_loop(orte_event_base, OPAL_EVLOOP_ONCE);
}
/* ensure all local procs are dead */
orte_odls.kill_local_procs(NULL);
DONE:
/* update the exit status, in case it wasn't done */
ORTE_UPDATE_EXIT_STATUS(ret);
/* cleanup and leave */
orte_finalize();
if (orte_debug_flag) {
fprintf(stderr, "exiting with status %d\n", orte_exit_status);
}
exit(orte_exit_status);
}
static int initialize(int argc, char *argv[]) {
int ret, exit_status = OPAL_SUCCESS;
char * tmp_env_var = NULL;
/*
* Make sure to init util before parse_args
* to ensure installdirs is setup properly
* before calling mca_base_open();
*/
if( OPAL_SUCCESS != (ret = opal_init_util(&argc, &argv)) ) {
return ret;
}
/*
* Parse Command Line Arguments
*/
if (OPAL_SUCCESS != (ret = parse_args(argc, argv))) {
exit_status = ret;
goto cleanup;
}
/*
* Setup OPAL Output handle from the verbose argument
*/
if( opal_checkpoint_globals.verbose ) {
opal_checkpoint_globals.quiet = false; /* Automaticly turn off quiet if it is set */
opal_checkpoint_globals.output = opal_output_open(NULL);
opal_output_set_verbosity(opal_checkpoint_globals.output, 10);
} else {
opal_checkpoint_globals.output = 0; /* Default=STDOUT */
}
/*
* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/*
* Select the 'none' CRS component,
* since we don't actually use a checkpointer
*/
(void) mca_base_var_env_name("crs", &tmp_env_var);
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/*
* Initialize OPAL
*/
if (OPAL_SUCCESS != (ret = opal_init(&argc, &argv))) {
exit_status = ret;
goto cleanup;
}
cleanup:
return exit_status;
}
static int do_child(orte_app_context_t* context,
orte_proc_t *child,
char **environ_copy,
orte_job_t *jobdat, int write_fd,
orte_iof_base_io_conf_t opts)
{
int i, rc;
sigset_t sigs;
long fd, fdmax = sysconf(_SC_OPEN_MAX);
char *param, *msg;
if (orte_forward_job_control) {
/* Set a new process group for this child, so that a
SIGSTOP can be sent to it without being sent to the
orted. */
setpgid(0, 0);
}
/* Setup the pipe to be close-on-exec */
opal_fd_set_cloexec(write_fd);
if (NULL != child) {
/* setup stdout/stderr so that any error messages that we
may print out will get displayed back at orterun.
NOTE: Definitely do this AFTER we check contexts so
that any error message from those two functions doesn't
come out to the user. IF we didn't do it in this order,
THEN a user who gives us a bad executable name or
working directory would get N error messages, where
N=num_procs. This would be very annoying for large
jobs, so instead we set things up so that orterun
always outputs a nice, single message indicating what
happened
*/
if (ORTE_SUCCESS != (i = orte_iof_base_setup_child(&opts,
&environ_copy))) {
ORTE_ERROR_LOG(i);
send_error_show_help(write_fd, 1,
"help-orte-odls-default.txt",
"iof setup failed",
orte_process_info.nodename, context->app);
/* Does not return */
}
/* now set any child-level controls such as binding */
orte_rtc.set(jobdat, child, &environ_copy, write_fd);
} else if (!ORTE_FLAG_TEST(jobdat, ORTE_JOB_FLAG_FORWARD_OUTPUT)) {
/* tie stdin/out/err/internal to /dev/null */
int fdnull;
for (i=0; i < 3; i++) {
fdnull = open("/dev/null", O_RDONLY, 0);
if (fdnull > i && i != write_fd) {
dup2(fdnull, i);
}
close(fdnull);
}
fdnull = open("/dev/null", O_RDONLY, 0);
if (fdnull > opts.p_internal[1]) {
dup2(fdnull, opts.p_internal[1]);
}
close(fdnull);
}
/* if the user requested it, set the system resource limits */
if (OPAL_SUCCESS != (rc = opal_util_init_sys_limits(&msg))) {
send_error_show_help(write_fd, 1, "help-orte-odls-default.txt",
"set limit",
orte_process_info.nodename, context->app,
__FILE__, __LINE__, msg);
}
/* ensure we only do this once */
(void) mca_base_var_env_name("opal_set_max_sys_limits", ¶m);
opal_unsetenv(param, &environ_copy);
free(param);
/* close all file descriptors w/ exception of stdin/stdout/stderr,
the pipe used for the IOF INTERNAL messages, and the pipe up to
the parent. */
for(fd=3; fd<fdmax; fd++) {
if (fd != opts.p_internal[1] && fd != write_fd) {
close(fd);
}
}
if (context->argv == NULL) {
context->argv = malloc(sizeof(char*)*2);
context->argv[0] = strdup(context->app);
context->argv[1] = NULL;
}
/* Set signal handlers back to the default. Do this close to
the exev() 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 launched process
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 launched process. */
sigprocmask(0, 0, &sigs);
sigprocmask(SIG_UNBLOCK, &sigs, 0);
/* Exec the new executable */
if (10 < opal_output_get_verbosity(orte_odls_base_framework.framework_output)) {
int jout;
opal_output(0, "%s STARTING %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), context->app);
for (jout=0; NULL != context->argv[jout]; jout++) {
opal_output(0, "%s\tARGV[%d]: %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), jout, context->argv[jout]);
}
for (jout=0; NULL != environ_copy[jout]; jout++) {
opal_output(0, "%s\tENVIRON[%d]: %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), jout, environ_copy[jout]);
}
}
execve(context->app, context->argv, environ_copy);
send_error_show_help(write_fd, 1,
"help-orte-odls-default.txt", "execve error",
orte_process_info.nodename, context->app, strerror(errno));
/* Does not return */
}
int opal_register_params(void)
{
int ret;
char *string = NULL;
if (opal_register_done) {
return OPAL_SUCCESS;
}
opal_register_done = true;
/*
* This string is going to be used in opal/util/stacktrace.c
*/
{
int j;
int signals[] = {
#ifdef SIGABRT
SIGABRT,
#endif
#ifdef SIGBUS
SIGBUS,
#endif
#ifdef SIGFPE
SIGFPE,
#endif
#ifdef SIGSEGV
SIGSEGV,
#endif
-1
};
for (j = 0 ; signals[j] != -1 ; ++j) {
if (j == 0) {
opal_asprintf(&string, "%d", signals[j]);
} else {
char *tmp;
opal_asprintf(&tmp, "%s,%d", string, signals[j]);
free(string);
string = tmp;
}
}
opal_signal_string = string;
ret = mca_base_var_register ("opal", "opal", NULL, "signal",
"Comma-delimited list of integer signal numbers to Open MPI to attempt to intercept. Upon receipt of the intercepted signal, Open MPI will display a stack trace and abort. Open MPI will *not* replace signals if handlers are already installed by the time MPI_INIT is invoked. Optionally append \":complain\" to any signal number in the comma-delimited list to make Open MPI complain if it detects another signal handler (and therefore does not insert its own).",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_signal_string);
free (string);
if (0 > ret) {
return ret;
}
}
/*
* Where should the stack trace output be directed
* This string is going to be used in opal/util/stacktrace.c
*/
string = strdup("stderr");
opal_stacktrace_output_filename = string;
ret = mca_base_var_register ("opal", "opal", NULL, "stacktrace_output",
"Specifies where the stack trace output stream goes. "
"Accepts one of the following: none (disabled), stderr (default), stdout, file[:filename]. "
"If 'filename' is not specified, a default filename of 'stacktrace' is used. "
"The 'filename' is appended with either '.PID' or '.RANK.PID', if RANK is available. "
"The 'filename' can be an absolute path or a relative path to the current working directory.",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_LOCAL,
&opal_stacktrace_output_filename);
free (string);
if (0 > ret) {
return ret;
}
#if defined(HAVE_SCHED_YIELD)
opal_progress_yield_when_idle = false;
ret = mca_base_var_register ("opal", "opal", "progress", "yield_when_idle",
"Yield the processor when waiting on progress",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_progress_yield_when_idle);
#endif
#if OPAL_ENABLE_DEBUG
opal_progress_debug = false;
ret = mca_base_var_register ("opal", "opal", "progress", "debug",
"Set to non-zero to debug progress engine features",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_progress_debug);
if (0 > ret) {
return ret;
}
opal_debug_threads = false;
ret = mca_base_var_register ("opal", "opal", "debug", "threads",
"Debug thread usage within OPAL. Reports out "
"when threads are acquired and released.",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_debug_threads);
if (0 > ret) {
return ret;
}
#endif
#if OPAL_ENABLE_FT_CR == 1
opal_base_distill_checkpoint_ready = false;
ret = mca_base_var_register("opal", "opal", "base", "distill_checkpoint_ready",
"Distill only those components that are Checkpoint Ready",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_8, MCA_BASE_VAR_SCOPE_LOCAL,
&opal_base_distill_checkpoint_ready);
if (0 > ret) {
return ret;
}
#endif
/* RFC1918 defines
- 10.0.0./8
- 172.16.0.0/12
- 192.168.0.0/16
RFC3330 also mentions
- 169.254.0.0/16 for DHCP onlink iff there's no DHCP server
*/
opal_net_private_ipv4 = "10.0.0.0/8;172.16.0.0/12;192.168.0.0/16;169.254.0.0/16";
ret = mca_base_var_register ("opal", "opal", "net", "private_ipv4",
"Semicolon-delimited list of CIDR notation entries specifying what networks are considered \"private\" (default value based on RFC1918 and RFC3330)",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_ALL_EQ,
&opal_net_private_ipv4);
if (0 > ret) {
return ret;
}
opal_set_max_sys_limits = NULL;
ret = mca_base_var_register ("opal", "opal", NULL, "set_max_sys_limits",
"Set the specified system-imposed limits to the specified value, including \"unlimited\"."
"Supported params: core, filesize, maxmem, openfiles, stacksize, maxchildren",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_ALL_EQ,
&opal_set_max_sys_limits);
if (0 > ret) {
return ret;
}
ret = mca_base_var_register("opal", "opal", NULL, "built_with_cuda_support",
"Whether CUDA GPU buffer support is built into library or not",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_DEFAULT_ONLY,
OPAL_INFO_LVL_4, MCA_BASE_VAR_SCOPE_CONSTANT,
&opal_built_with_cuda_support);
if (0 > ret) {
return ret;
}
/* Current default is to enable CUDA support if it is built into library */
opal_cuda_support = opal_built_with_cuda_support;
ret = mca_base_var_register ("opal", "opal", NULL, "cuda_support",
"Whether CUDA GPU buffer support is enabled or not",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_ALL_EQ,
&opal_cuda_support);
if (0 > ret) {
return ret;
}
opal_warn_on_missing_libcuda = true;
ret = mca_base_var_register ("opal", "opal", NULL, "warn_on_missing_libcuda",
"Whether to print a message when CUDA support is enabled but libcuda is not found",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, MCA_BASE_VAR_FLAG_SETTABLE,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_ALL_EQ,
&opal_warn_on_missing_libcuda);
if (0 > ret) {
return ret;
}
/* Leave pinned parameter */
opal_leave_pinned = -1;
ret = mca_base_var_register("ompi", "mpi", NULL, "leave_pinned",
"Whether to use the \"leave pinned\" protocol or not. Enabling this setting can help bandwidth performance when repeatedly sending and receiving large messages with the same buffers over RDMA-based networks (false = do not use \"leave pinned\" protocol, true = use \"leave pinned\" protocol, auto = allow network to choose at runtime).",
MCA_BASE_VAR_TYPE_INT, &mca_base_var_enum_auto_bool, 0, 0,
OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_READONLY,
&opal_leave_pinned);
mca_base_var_register_synonym(ret, "opal", "opal", NULL, "leave_pinned",
MCA_BASE_VAR_SYN_FLAG_DEPRECATED);
opal_leave_pinned_pipeline = false;
ret = mca_base_var_register("ompi", "mpi", NULL, "leave_pinned_pipeline",
"Whether to use the \"leave pinned pipeline\" protocol or not.",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, 0,
OPAL_INFO_LVL_9,
MCA_BASE_VAR_SCOPE_READONLY,
&opal_leave_pinned_pipeline);
mca_base_var_register_synonym(ret, "opal", "opal", NULL, "leave_pinned_pipeline",
MCA_BASE_VAR_SYN_FLAG_DEPRECATED);
if (opal_leave_pinned > 0 && opal_leave_pinned_pipeline) {
opal_leave_pinned_pipeline = 0;
opal_show_help("help-opal-runtime.txt",
"mpi-params:leave-pinned-and-pipeline-selected",
true);
}
opal_warn_on_fork = true;
(void) mca_base_var_register("ompi", "mpi", NULL, "warn_on_fork",
"If nonzero, issue a warning if program forks under conditions that could cause system errors",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, 0,
OPAL_INFO_LVL_9,
MCA_BASE_VAR_SCOPE_READONLY,
&opal_warn_on_fork);
opal_abort_delay = 0;
ret = mca_base_var_register("opal", "opal", NULL, "abort_delay",
"If nonzero, print out an identifying message when abort operation is invoked (hostname, PID of the process that called abort) and delay for that many seconds before exiting (a negative delay value means to never abort). This allows attaching of a debugger before quitting the job.",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
OPAL_INFO_LVL_5,
MCA_BASE_VAR_SCOPE_READONLY,
&opal_abort_delay);
if (0 > ret) {
return ret;
}
opal_abort_print_stack = false;
ret = mca_base_var_register("opal", "opal", NULL, "abort_print_stack",
"If nonzero, print out a stack trace when abort is invoked",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0,
/* If we do not have stack trace
capability, make this a constant
MCA variable */
#if OPAL_WANT_PRETTY_PRINT_STACKTRACE
0,
OPAL_INFO_LVL_5,
MCA_BASE_VAR_SCOPE_READONLY,
#else
MCA_BASE_VAR_FLAG_DEFAULT_ONLY,
OPAL_INFO_LVL_5,
MCA_BASE_VAR_SCOPE_CONSTANT,
#endif
&opal_abort_print_stack);
if (0 > ret) {
return ret;
}
/* register the envar-forwarding params */
(void)mca_base_var_register ("opal", "mca", "base", "env_list",
"Set SHELL env variables",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0, OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_READONLY, &mca_base_env_list);
mca_base_env_list_sep = MCA_BASE_ENV_LIST_SEP_DEFAULT;
(void)mca_base_var_register ("opal", "mca", "base", "env_list_delimiter",
"Set SHELL env variables delimiter. Default: semicolon ';'",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0, OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_READONLY, &mca_base_env_list_sep);
/* Set OMPI_MCA_mca_base_env_list variable, it might not be set before
* if mca variable was taken from amca conf file. Need to set it
* here because mca_base_var_process_env_list is called from schizo_ompi.c
* only when this env variable was set.
*/
if (NULL != mca_base_env_list) {
char *name = NULL;
(void) mca_base_var_env_name ("mca_base_env_list", &name);
if (NULL != name) {
opal_setenv(name, mca_base_env_list, false, &environ);
free(name);
}
}
/* Register internal MCA variable mca_base_env_list_internal. It can be set only during
* parsing of amca conf file and contains SHELL env variables specified via -x there.
* Its format is the same as for mca_base_env_list.
*/
(void)mca_base_var_register ("opal", "mca", "base", "env_list_internal",
"Store SHELL env variables from amca conf file",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, MCA_BASE_VAR_FLAG_INTERNAL, OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_READONLY, &mca_base_env_list_internal);
/* Number of threads allowed in opal_progress. This might increase multithreaded performance. */
(void)mca_base_var_register ("opal", "opal", NULL, "max_thread_in_progress",
"Number of thread allowed in opal_progress. Default: 1",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0, OPAL_INFO_LVL_8,
MCA_BASE_VAR_SCOPE_READONLY, &opal_max_thread_in_progress);
/* The ddt engine has a few parameters */
ret = opal_datatype_register_params();
if (OPAL_SUCCESS != ret) {
return ret;
}
/* dss has parameters */
ret = opal_dss_register_vars ();
if (OPAL_SUCCESS != ret) {
return ret;
}
opal_finalize_register_cleanup (opal_deregister_params);
return OPAL_SUCCESS;
}
