/**
* Return the integer following an = (actually may only return positive ints)
*/
static int hostfile_parse_int(void)
{
if (ORTE_HOSTFILE_EQUAL != orte_util_hostfile_lex())
return -1;
if (ORTE_HOSTFILE_INT != orte_util_hostfile_lex())
return -1;
return orte_util_hostfile_value.ival;
}
/**
* Return the string following an = (option to a keyword)
*/
static char *hostfile_parse_string(void)
{
int rc;
if (ORTE_HOSTFILE_EQUAL != orte_util_hostfile_lex()){
return NULL;
}
rc = orte_util_hostfile_lex();
if (ORTE_HOSTFILE_STRING != rc){
return NULL;
}
return strdup(orte_util_hostfile_value.sval);
}
static int hostfile_parse(const char *hostfile, opal_list_t* updates, opal_list_t* exclude, bool keep_all)
{
int token;
int rc = ORTE_SUCCESS;
cur_hostfile_name = hostfile;
orte_util_hostfile_done = false;
orte_util_hostfile_in = fopen(hostfile, "r");
if (NULL == orte_util_hostfile_in) {
orte_show_help("help-hostfile.txt", "no-hostfile", true, hostfile);
rc = ORTE_ERR_NOT_FOUND;
goto unlock;
}
while (!orte_util_hostfile_done) {
token = orte_util_hostfile_lex();
switch (token) {
case ORTE_HOSTFILE_DONE:
orte_util_hostfile_done = true;
break;
case ORTE_HOSTFILE_NEWLINE:
break;
/*
* This looks odd, since we have several forms of host-definitions:
* hostname just plain as it is, being a ORTE_HOSTFILE_STRING
* IP4s and user@IPv4s
* hostname.domain and [email protected]
*/
case ORTE_HOSTFILE_STRING:
case ORTE_HOSTFILE_INT:
case ORTE_HOSTFILE_HOSTNAME:
case ORTE_HOSTFILE_IPV4:
case ORTE_HOSTFILE_IPV6:
case ORTE_HOSTFILE_RELATIVE:
case ORTE_HOSTFILE_RANK:
rc = hostfile_parse_line(token, updates, exclude, keep_all);
if (ORTE_SUCCESS != rc) {
goto unlock;
}
break;
default:
hostfile_parse_error(token);
goto unlock;
}
}
fclose(orte_util_hostfile_in);
orte_util_hostfile_in = NULL;
unlock:
cur_hostfile_name = NULL;
return rc;
}
static int hostfile_parse_line(int token, opal_list_t* updates, opal_list_t* exclude, bool keep_all)
{
int rc;
orte_node_t* node;
bool got_count = false;
bool got_max = false;
char* value;
char** argv;
char* node_name = NULL;
char* node_alias = NULL;
char* username = NULL;
int cnt;
int number_of_slots = 0;
char buff[64];
if (ORTE_HOSTFILE_STRING == token ||
ORTE_HOSTFILE_HOSTNAME == token ||
ORTE_HOSTFILE_INT == token ||
ORTE_HOSTFILE_IPV4 == token ||
ORTE_HOSTFILE_IPV6 == token) {
if(ORTE_HOSTFILE_INT == token) {
snprintf(buff, 64, "%d", orte_util_hostfile_value.ival);
value = buff;
} else {
value = orte_util_hostfile_value.sval;
}
argv = opal_argv_split (value, '@');
cnt = opal_argv_count (argv);
if (1 == cnt) {
node_name = strdup(argv[0]);
} else if (2 == cnt) {
username = strdup(argv[0]);
node_name = strdup(argv[1]);
} else {
opal_output(0, "WARNING: Unhandled user@host-combination\n"); /* XXX */
}
opal_argv_free (argv);
/* if the first letter of the name is '^', then this is a node
* to be excluded. Remove the ^ character so the nodename is
* usable, and put it on the exclude list
*/
if ('^' == node_name[0]) {
int i, len;
len = strlen(node_name);
for (i=1; i < len; i++) {
node_name[i-1] = node_name[i];
}
node_name[len-1] = '\0'; /* truncate */
OPAL_OUTPUT_VERBOSE((3, orte_ras_base_framework.framework_output,
"%s hostfile: node %s is being excluded",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node_name));
/* convert this into something globally unique */
if (strcmp(node_name, "localhost") == 0 || opal_ifislocal(node_name)) {
/* Nodename has been allocated, that is for sure */
if (orte_show_resolved_nodenames &&
0 != strcmp(node_name, orte_process_info.nodename)) {
node_alias = strdup(node_name);
}
free (node_name);
node_name = strdup(orte_process_info.nodename);
}
/* Do we need to make a new node object? First check to see
if it's already in the exclude list */
if (NULL == (node = hostfile_lookup(exclude, node_name))) {
node = OBJ_NEW(orte_node_t);
node->name = node_name;
if (NULL != username) {
node->username = strdup(username);
}
}
/* Note that we need to add this back to the exclude list.
If it was found, we just removed it (in hostfile_lookup()),
so this puts it back. If it was not found, then we have to
add it to the exclude list anyway. */
opal_list_append(exclude, &node->super);
return ORTE_SUCCESS;
}
/* this is not a node to be excluded, so we need to process it and
* add it to the "include" list. See if this host is actually us.
*/
if (strcmp(node_name, "localhost") == 0 || opal_ifislocal(node_name)) {
/* Nodename has been allocated, that is for sure */
if (orte_show_resolved_nodenames &&
0 != strcmp(node_name, orte_process_info.nodename)) {
node_alias = strdup(node_name);
}
free (node_name);
node_name = strdup(orte_process_info.nodename);
}
OPAL_OUTPUT_VERBOSE((3, orte_ras_base_framework.framework_output,
"%s hostfile: node %s is being included - keep all is %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node_name,
keep_all ? "TRUE" : "FALSE"));
/* Do we need to make a new node object? First check to see
* if we are keeping everything or if it's already in the updates
* list. Because we check keep_all first, if that is set we will
* not do the hostfile_lookup call, and thus won't remove the
* pre-existing node from the updates list
*/
if (keep_all || NULL == (node = hostfile_lookup(updates, node_name))) {
node = OBJ_NEW(orte_node_t);
node->name = node_name;
if (NULL != username) {
node->username = strdup(username);
}
}
/* do we need to record an alias for this node? */
if (NULL != node_alias) {
/* add to list of aliases for this node - only add if unique */
opal_argv_append_unique_nosize(&node->alias, node_alias, false);
free(node_alias);
}
} else if (ORTE_HOSTFILE_RELATIVE == token) {
/* store this for later processing */
node = OBJ_NEW(orte_node_t);
node->name = strdup(orte_util_hostfile_value.sval);
if (NULL != username) {
node->username = strdup(username);
}
} else if (ORTE_HOSTFILE_RANK == token) {
/* we can ignore the rank, but we need to extract the node name. we
* first need to shift over to the other side of the equal sign as
* this is where the node name will be
*/
while (!orte_util_hostfile_done &&
ORTE_HOSTFILE_EQUAL != token) {
token = orte_util_hostfile_lex();
}
if (orte_util_hostfile_done) {
/* bad syntax somewhere */
return ORTE_ERROR;
}
/* next position should be the node name */
token = orte_util_hostfile_lex();
if(ORTE_HOSTFILE_INT == token) {
snprintf(buff, 64, "%d", orte_util_hostfile_value.ival);
value = buff;
} else {
value = orte_util_hostfile_value.sval;
}
argv = opal_argv_split (value, '@');
cnt = opal_argv_count (argv);
if (1 == cnt) {
node_name = strdup(argv[0]);
} else if (2 == cnt) {
username = strdup(argv[0]);
node_name = strdup(argv[1]);
} else {
opal_output(0, "WARNING: Unhandled user@host-combination\n"); /* XXX */
}
opal_argv_free (argv);
/* Do we need to make a new node object? First check to see
* if we are keeping everything or if it's already in the updates
* list. Because we check keep_all first, if that is set we will
* not do the hostfile_lookup call, and thus won't remove the
* pre-existing node from the updates list
*/
if (keep_all || NULL == (node = hostfile_lookup(updates, node_name))) {
node = OBJ_NEW(orte_node_t);
node->name = node_name;
if (NULL != username) {
node->username = strdup(username);
}
}
/* add a slot */
node->slots++;
/* do we need to record an alias for this node? */
if (NULL != node_alias) {
/* add to list of aliases for this node - only add if unique */
opal_argv_append_unique_nosize(&node->alias, node_alias, false);
free(node_alias);
}
/* skip to end of line */
while (!orte_util_hostfile_done &&
ORTE_HOSTFILE_NEWLINE != token) {
token = orte_util_hostfile_lex();
}
opal_list_append(updates, &node->super);
return ORTE_SUCCESS;
} else {
hostfile_parse_error(token);
return ORTE_ERROR;
}
got_count = false;
while (!orte_util_hostfile_done) {
token = orte_util_hostfile_lex();
switch (token) {
case ORTE_HOSTFILE_DONE:
goto done;
case ORTE_HOSTFILE_NEWLINE:
goto done;
case ORTE_HOSTFILE_USERNAME:
node->username = hostfile_parse_string();
break;
case ORTE_HOSTFILE_COUNT:
case ORTE_HOSTFILE_CPU:
case ORTE_HOSTFILE_SLOTS:
rc = hostfile_parse_int();
if (rc < 0) {
orte_show_help("help-hostfile.txt", "slots",
true,
cur_hostfile_name, rc);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
node->slots += rc;
got_count = true;
/* Ensure that slots_max >= slots */
if (node->slots_max != 0 && node->slots_max < node->slots) {
node->slots_max = node->slots;
}
break;
case ORTE_HOSTFILE_SLOTS_MAX:
rc = hostfile_parse_int();
if (rc < 0) {
orte_show_help("help-hostfile.txt", "max_slots",
true,
cur_hostfile_name, ((size_t) rc));
OBJ_RELEASE(node);
return ORTE_ERROR;
}
/* Only take this update if it puts us >= node_slots */
if (rc >= node->slots) {
if (node->slots_max != rc) {
node->slots_max = rc;
got_max = true;
}
} else {
orte_show_help("help-hostfile.txt", "max_slots_lt",
true,
cur_hostfile_name, node->slots, rc);
ORTE_ERROR_LOG(ORTE_ERR_BAD_PARAM);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
break;
default:
hostfile_parse_error(token);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
if (number_of_slots > node->slots) {
ORTE_ERROR_LOG(ORTE_ERR_BAD_PARAM);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
}
done:
if (got_count) {
node->slots_given = true;
} else if (got_max) {
node->slots = node->slots_max;
node->slots_given = true;
} else {
/* should be set by obj_new, but just to be clear */
node->slots_given = false;
/* if we weren't give a count or a max, then
* just increment by one to support RMs that
* count slots by listing the node multiple
* times in the file
*/
++node->slots;
}
opal_list_append(updates, &node->super);
return ORTE_SUCCESS;
}
static int hostfile_parse_line(int token, opal_list_t* updates,
opal_list_t* exclude, bool keep_all)
{
int rc;
orte_node_t* node;
bool got_max = false;
char* value;
char **argv;
char* node_name = NULL;
char* username = NULL;
int cnt;
int number_of_slots = 0;
char buff[64];
if (ORTE_HOSTFILE_STRING == token ||
ORTE_HOSTFILE_HOSTNAME == token ||
ORTE_HOSTFILE_INT == token ||
ORTE_HOSTFILE_IPV4 == token ||
ORTE_HOSTFILE_IPV6 == token) {
if(ORTE_HOSTFILE_INT == token) {
snprintf(buff, 64, "%d", orte_util_hostfile_value.ival);
value = buff;
} else {
value = orte_util_hostfile_value.sval;
}
argv = opal_argv_split (value, '@');
cnt = opal_argv_count (argv);
if (1 == cnt) {
node_name = strdup(argv[0]);
} else if (2 == cnt) {
username = argv[0];
node_name = strdup(argv[1]);
} else {
opal_output(0, "WARNING: Unhandled user@host-combination\n"); /* XXX */
}
opal_argv_free (argv);
// Strip off the FQDN if present, ignore IP addresses
if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(node_name) ) {
char *ptr;
if (NULL != (ptr = strchr(node_name, '.'))) {
*ptr = '\0';
}
}
/* if the first letter of the name is '^', then this is a node
* to be excluded. Remove the ^ character so the nodename is
* usable, and put it on the exclude list
*/
if ('^' == node_name[0]) {
int i, len;
len = strlen(node_name);
for (i=1; i < len; i++) {
node_name[i-1] = node_name[i];
}
node_name[len-1] = '\0'; /* truncate */
OPAL_OUTPUT_VERBOSE((3, orte_ras_base_framework.framework_output,
"%s hostfile: node %s is being excluded",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node_name));
/* see if this is another name for us */
if (orte_ifislocal(node_name)) {
/* Nodename has been allocated, that is for sure */
free (node_name);
node_name = strdup(orte_process_info.nodename);
}
/* Do we need to make a new node object? First check to see
if it's already in the exclude list */
if (NULL == (node = hostfile_lookup(exclude, node_name))) {
node = OBJ_NEW(orte_node_t);
node->name = node_name;
if (NULL != username) {
orte_set_attribute(&node->attributes, ORTE_NODE_USERNAME, ORTE_ATTR_LOCAL, username, OPAL_STRING);
}
opal_list_append(exclude, &node->super);
} else {
free(node_name);
}
return ORTE_SUCCESS;
}
/* this is not a node to be excluded, so we need to process it and
* add it to the "include" list. See if this host is actually us.
*/
if (orte_ifislocal(node_name)) {
/* Nodename has been allocated, that is for sure */
free (node_name);
node_name = strdup(orte_process_info.nodename);
}
OPAL_OUTPUT_VERBOSE((3, orte_ras_base_framework.framework_output,
"%s hostfile: node %s is being included - keep all is %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node_name,
keep_all ? "TRUE" : "FALSE"));
/* Do we need to make a new node object? */
if (keep_all || NULL == (node = hostfile_lookup(updates, node_name))) {
node = OBJ_NEW(orte_node_t);
node->name = node_name;
node->slots = 1;
if (NULL != username) {
orte_set_attribute(&node->attributes, ORTE_NODE_USERNAME, ORTE_ATTR_LOCAL, username, OPAL_STRING);
}
opal_list_append(updates, &node->super);
} else {
/* this node was already found once - add a slot and mark slots as "given" */
node->slots++;
ORTE_FLAG_SET(node, ORTE_NODE_FLAG_SLOTS_GIVEN);
free(node_name);
}
} else if (ORTE_HOSTFILE_RELATIVE == token) {
/* store this for later processing */
node = OBJ_NEW(orte_node_t);
node->name = strdup(orte_util_hostfile_value.sval);
opal_list_append(updates, &node->super);
} else if (ORTE_HOSTFILE_RANK == token) {
/* we can ignore the rank, but we need to extract the node name. we
* first need to shift over to the other side of the equal sign as
* this is where the node name will be
*/
while (!orte_util_hostfile_done &&
ORTE_HOSTFILE_EQUAL != token) {
token = orte_util_hostfile_lex();
}
if (orte_util_hostfile_done) {
/* bad syntax somewhere */
return ORTE_ERROR;
}
/* next position should be the node name */
token = orte_util_hostfile_lex();
if(ORTE_HOSTFILE_INT == token) {
snprintf(buff, 64, "%d", orte_util_hostfile_value.ival);
value = buff;
} else {
value = orte_util_hostfile_value.sval;
}
argv = opal_argv_split (value, '@');
cnt = opal_argv_count (argv);
if (1 == cnt) {
node_name = strdup(argv[0]);
} else if (2 == cnt) {
username = argv[0];
node_name = strdup(argv[1]);
} else {
opal_output(0, "WARNING: Unhandled user@host-combination\n"); /* XXX */
}
opal_argv_free (argv);
// Strip off the FQDN if present, ignore IP addresses
if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(node_name) ) {
char *ptr;
if (NULL != (ptr = strchr(node_name, '.'))) {
*ptr = '\0';
}
}
/* Do we need to make a new node object? */
if (NULL == (node = hostfile_lookup(updates, node_name))) {
node = OBJ_NEW(orte_node_t);
node->name = node_name;
node->slots = 1;
if (NULL != username) {
orte_set_attribute(&node->attributes, ORTE_NODE_USERNAME, ORTE_ATTR_LOCAL, username, OPAL_STRING);
}
opal_list_append(updates, &node->super);
} else {
/* add a slot */
node->slots++;
free(node_name);
}
OPAL_OUTPUT_VERBOSE((1, orte_ras_base_framework.framework_output,
"%s hostfile: node %s slots %d nodes-given %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node->name, node->slots,
ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_SLOTS_GIVEN) ? "TRUE" : "FALSE"));
/* mark the slots as "given" since we take them as being the
* number specified via the rankfile
*/
ORTE_FLAG_SET(node, ORTE_NODE_FLAG_SLOTS_GIVEN);
/* skip to end of line */
while (!orte_util_hostfile_done &&
ORTE_HOSTFILE_NEWLINE != token) {
token = orte_util_hostfile_lex();
}
return ORTE_SUCCESS;
} else {
hostfile_parse_error(token);
return ORTE_ERROR;
}
free(username);
while (!orte_util_hostfile_done) {
token = orte_util_hostfile_lex();
switch (token) {
case ORTE_HOSTFILE_DONE:
goto done;
case ORTE_HOSTFILE_NEWLINE:
goto done;
case ORTE_HOSTFILE_USERNAME:
username = hostfile_parse_string();
if (NULL != username) {
orte_set_attribute(&node->attributes, ORTE_NODE_USERNAME, ORTE_ATTR_LOCAL, username, OPAL_STRING);
free(username);
}
break;
case ORTE_HOSTFILE_PORT:
rc = hostfile_parse_int();
if (rc < 0) {
orte_show_help("help-hostfile.txt", "port",
true,
cur_hostfile_name, rc);
return ORTE_ERROR;
}
orte_set_attribute(&node->attributes, ORTE_NODE_PORT, ORTE_ATTR_LOCAL, &rc, OPAL_INT);
break;
case ORTE_HOSTFILE_COUNT:
case ORTE_HOSTFILE_CPU:
case ORTE_HOSTFILE_SLOTS:
rc = hostfile_parse_int();
if (rc < 0) {
orte_show_help("help-hostfile.txt", "slots",
true,
cur_hostfile_name, rc);
opal_list_remove_item(updates, &node->super);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
if (ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_SLOTS_GIVEN)) {
/* multiple definitions were given for the
* slot count - this is not allowed
*/
orte_show_help("help-hostfile.txt", "slots-given",
true,
cur_hostfile_name, node->name);
opal_list_remove_item(updates, &node->super);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
node->slots = rc;
ORTE_FLAG_SET(node, ORTE_NODE_FLAG_SLOTS_GIVEN);
/* Ensure that slots_max >= slots */
if (node->slots_max != 0 && node->slots_max < node->slots) {
node->slots_max = node->slots;
}
break;
case ORTE_HOSTFILE_SLOTS_MAX:
rc = hostfile_parse_int();
if (rc < 0) {
orte_show_help("help-hostfile.txt", "max_slots",
true,
cur_hostfile_name, ((size_t) rc));
opal_list_remove_item(updates, &node->super);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
/* Only take this update if it puts us >= node_slots */
if (rc >= node->slots) {
if (node->slots_max != rc) {
node->slots_max = rc;
got_max = true;
}
} else {
orte_show_help("help-hostfile.txt", "max_slots_lt",
true,
cur_hostfile_name, node->slots, rc);
ORTE_ERROR_LOG(ORTE_ERR_BAD_PARAM);
opal_list_remove_item(updates, &node->super);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
break;
case ORTE_HOSTFILE_STRING:
case ORTE_HOSTFILE_INT:
/* just ignore it */
break;
default:
hostfile_parse_error(token);
opal_list_remove_item(updates, &node->super);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
if (number_of_slots > node->slots) {
ORTE_ERROR_LOG(ORTE_ERR_BAD_PARAM);
opal_list_remove_item(updates, &node->super);
OBJ_RELEASE(node);
return ORTE_ERROR;
}
}
done:
if (got_max && !ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_SLOTS_GIVEN)) {
node->slots = node->slots_max;
ORTE_FLAG_SET(node, ORTE_NODE_FLAG_SLOTS_GIVEN);
}
return ORTE_SUCCESS;
}