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

If we are given a TCP oob address that doesn't match any active module, it is still possible that we could route to the address if a router is in the system. No harm in trying, so arbitrarily pick the first connection in the active module list and assign the peer to it. If that module can't reach it, we'll follow the usual failover mechanism until finally concluding that nobody can get there.

Просмотр исходного кода 
cmr=v1.7.5:reviewer=jsquyres:subject=handle non-matching addresses

This commit was SVN r30719.
Этот коммит содержится в:
Ralph Castain 2014-02-13 23:37:22 +00:00
родитель 449cd8f3d7
Коммит 4e1c07cbf2
2 изменённых файлов: 92 добавлений и 4 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

90
orte/mca/oob/tcp/oob_tcp_component.c
Просмотреть файл

@ -799,7 +799,7 @@ static int component_set_addr(orte_process_name_t *peer,
mca_oob_tcp_component_peer_t *pr;
uint16_t af_family = AF_UNSPEC;
uint64_t ui64;
bool found;
bool found, assigned;
memcpy(&ui64, (char*)peer, sizeof(uint64_t));
/* cycle across component parts and see if one belongs to us */
@ -866,10 +866,27 @@ static int component_set_addr(orte_process_name_t *peer,
/* cycle across the provided addrs */
assigned = false;
for (j=0; NULL != addrs[j]; j++) {
/* if they gave us "localhost", then just take the first conn on our list */
if (0 == strcasecmp(addrs[j], "localhost")) {
#if OPAL_ENABLE_IPV6
if (AF_INET6 == af_family) {
if (NULL == mca_oob_tcp_component.ipv6conns ||
NULL == mca_oob_tcp_component.ipv6conns[0]) {
continue;
}
host = mca_oob_tcp_component.ipv6conns[0];
} else {
#endif
if (NULL == mca_oob_tcp_component.ipv4conns ||
NULL == mca_oob_tcp_component.ipv4conns[0]) {
continue;
}
host = mca_oob_tcp_component.ipv4conns[0];
#if OPAL_ENABLE_IPV6
}
#endif
} else {
host = addrs[j];
}
@ -918,6 +935,77 @@ static int component_set_addr(orte_process_name_t *peer,
mod->api.set_peer((struct mca_oob_tcp_module_t*)mod,
peer, af_family, host, ports);
found = true;
assigned = true;
}
/* if we cycled thru all their addresses without finding a match in our
* interfaces, then it remains possible that they have a routed system
* that can still route messages to the destination. So give it a chance
* to succeed by assigning our first module to try, and let the normal
* failure progression ultimately determine if we can reach this peer
*/
if (!assigned) {
#if OPAL_ENABLE_IPV6
if (AF_INET6 == af_family) {
if (NULL == mca_oob_tcp_component.ipv6conns ||
NULL == mca_oob_tcp_component.ipv6conns[0]) {
continue;
}
host = mca_oob_tcp_component.ipv6conns[0];
} else {
#endif
if (NULL == mca_oob_tcp_component.ipv4conns ||
NULL == mca_oob_tcp_component.ipv4conns[0]) {
continue;
}
host = mca_oob_tcp_component.ipv4conns[0];
#if OPAL_ENABLE_IPV6
}
#endif
/* lookup the kernel index of this address */
if (0 >= (k = opal_ifaddrtokindex(host))) {
opal_output_verbose(OOB_TCP_DEBUG_CONNECT, orte_oob_base_framework.framework_output,
"%s UNFOUND KERNEL INDEX %d FOR ADDRESS %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), k, host);
/* we don't have an interface on this subnet - ignore it */
continue;
}
if (NULL == (mod = (mca_oob_tcp_module_t*)opal_pointer_array_get_item(&mca_oob_tcp_component.modules, k))) {
opal_output_verbose(OOB_TCP_DEBUG_CONNECT, orte_oob_base_framework.framework_output,
"%s NO MODULE AT KINDEX %d FOR ADDRESS %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), k, host);
continue;
}
/* record that this peer may be reachable via this module, but don't assign
* the peer to this module until later when we actually connect
*/
opal_output_verbose(OOB_TCP_DEBUG_CONNECT, orte_oob_base_framework.framework_output,
"%s PEER %s MAY BE REACHABLE BY ROUTING - ASSIGNING MODULE AT KINDEX %d INTERFACE %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(peer), k, mod->if_name);
if (OPAL_SUCCESS != opal_hash_table_get_value_uint64(&mca_oob_tcp_component.peers,
ui64, (void**)&pr) || NULL == pr) {
pr = OBJ_NEW(mca_oob_tcp_component_peer_t);
if (OPAL_SUCCESS != (rc = opal_hash_table_set_value_uint64(&mca_oob_tcp_component.peers,
ui64, (void*)pr))) {
ORTE_ERROR_LOG(rc);
return rc;
}
}
opal_bitmap_set_bit(&pr->reachable, k);
/* pass this proc, and its ports, to the
* module for handling - this module will be responsible
* for communicating with the proc via this network.
* Note that the modules are *not* necessarily running
* on our event base - thus, the modules will push this
* call into their own event base for processing.
*/
opal_output_verbose(OOB_TCP_DEBUG_CONNECT, orte_oob_base_framework.framework_output,
"%s PASSING ADDR %s TO INTERFACE %s AT KERNEL INDEX %d",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), host,
mod->if_name, k);
mod->api.set_peer((struct mca_oob_tcp_module_t*)mod,
peer, af_family, host, ports);
found = true;
}
if (NULL != addrs) {
opal_argv_free(addrs);

6
orte/mca/oob/tcp/oob_tcp_listener.c
Просмотреть файл

@ -192,7 +192,7 @@ static int create_listen(void)
opal_argv_append_nosize(&ports, "0");
orte_static_ports = false;
}
} else if (ORTE_PROC_IS_HNP) {
} else if (ORTE_PROC_IS_HNP || ORTE_PROC_IS_SCHEDULER) {
if (NULL != mca_oob_tcp_component.tcp_static_ports) {
/* if static ports were provided, take the
* first entry in the list
@ -357,9 +357,9 @@ static int create_listen(void)
opal_argv_append_nosize(&mca_oob_tcp_component.ipv4ports, tconn);
free(tconn);
if (OOB_TCP_DEBUG_CONNECT <= opal_output_get_verbosity(orte_oob_base_framework.framework_output)) {
port = ntohs(((struct sockaddr_in*) &inaddr)->sin_port);
opal_output(0, "%s assigned IPv4 port %d",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
((struct sockaddr_in*) &inaddr)->sin_port);
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), port);
}
if (!ORTE_PROC_IS_HNP) {