void osm_update_node_desc(IN osm_opensm_t *osm)
{
CL_PLOCK_ACQUIRE(&osm->lock);
cl_qmap_apply_func(&osm->subn.node_guid_tbl, state_mgr_update_node_desc,
&osm->sm);
CL_PLOCK_RELEASE(&osm->lock);
}
/**********************************************************************
Initiates a thorough sweep of the subnet.
Used when there is suspicion that something on the subnet has changed.
**********************************************************************/
static ib_api_status_t state_mgr_sweep_hop_0(IN osm_sm_t * sm)
{
ib_api_status_t status;
osm_dr_path_t dr_path;
osm_bind_handle_t h_bind;
uint8_t path_array[IB_SUBNET_PATH_HOPS_MAX];
OSM_LOG_ENTER(sm->p_log);
memset(path_array, 0, sizeof(path_array));
/*
* First, get the bind handle.
*/
h_bind = osm_sm_mad_ctrl_get_bind_handle(&sm->mad_ctrl);
if (h_bind != OSM_BIND_INVALID_HANDLE) {
OSM_LOG_MSG_BOX(sm->p_log, OSM_LOG_VERBOSE,
"INITIATING HEAVY SWEEP");
/*
* Start the sweep by clearing the port counts, then
* get our own NodeInfo at 0 hops.
*/
CL_PLOCK_ACQUIRE(sm->p_lock);
cl_qmap_apply_func(&sm->p_subn->node_guid_tbl,
state_mgr_reset_node_count, sm);
cl_qmap_apply_func(&sm->p_subn->port_guid_tbl,
state_mgr_reset_port_count, sm);
cl_qmap_apply_func(&sm->p_subn->sw_guid_tbl,
state_mgr_reset_switch_count, sm);
/* Set the in_sweep_hop_0 flag in subn to be TRUE.
* This will indicate the sweeping not to continue beyond the
* the current node.
* This is relevant for the case of SM on switch, since in the
* switch info we need to signal somehow not to continue
* the sweeping. */
sm->p_subn->in_sweep_hop_0 = TRUE;
CL_PLOCK_RELEASE(sm->p_lock);
osm_dr_path_init(&dr_path, h_bind, 0, path_array);
status = osm_req_get(sm, &dr_path, IB_MAD_ATTR_NODE_INFO, 0,
CL_DISP_MSGID_NONE, NULL);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3305: "
"Request for NodeInfo failed (%s)\n",
ib_get_err_str(status));
} else {
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"No bound ports. Deferring sweep...\n");
status = IB_INVALID_STATE;
}
OSM_LOG_EXIT(sm->p_log);
return status;
}
/**********************************************************************
Returns true if the SM port is down.
The SM's port object must exist in the port_guid table.
**********************************************************************/
static boolean_t state_mgr_is_sm_port_down(IN osm_sm_t * sm)
{
ib_net64_t port_guid;
osm_port_t *p_port;
osm_physp_t *p_physp;
uint8_t state;
OSM_LOG_ENTER(sm->p_log);
port_guid = sm->p_subn->sm_port_guid;
/*
* If we don't know our own port guid yet, assume the port is down.
*/
if (port_guid == 0) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3308: "
"SM port GUID unknown\n");
state = IB_LINK_DOWN;
goto Exit;
}
CL_ASSERT(port_guid);
CL_PLOCK_ACQUIRE(sm->p_lock);
p_port = osm_get_port_by_guid(sm->p_subn, port_guid);
if (!p_port) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3309: "
"SM port with GUID:%016" PRIx64 " is unknown\n",
cl_ntoh64(port_guid));
state = IB_LINK_DOWN;
CL_PLOCK_RELEASE(sm->p_lock);
goto Exit;
}
p_physp = p_port->p_physp;
CL_ASSERT(p_physp);
if (p_port->p_node->sw &&
!ib_switch_info_is_enhanced_port0(&p_port->p_node->sw->switch_info))
state = IB_LINK_ACTIVE; /* base SP0 */
else
state = osm_physp_get_port_state(p_physp);
CL_PLOCK_RELEASE(sm->p_lock);
Exit:
OSM_LOG_EXIT(sm->p_log);
return (state == IB_LINK_DOWN);
}
/**********************************************************************
Clear out all existing port lid assignments
**********************************************************************/
static ib_api_status_t state_mgr_clean_known_lids(IN osm_sm_t * sm)
{
ib_api_status_t status = IB_SUCCESS;
cl_ptr_vector_t *p_vec = &(sm->p_subn->port_lid_tbl);
uint32_t i;
OSM_LOG_ENTER(sm->p_log);
/* we need a lock here! */
CL_PLOCK_ACQUIRE(sm->p_lock);
for (i = 0; i < cl_ptr_vector_get_size(p_vec); i++)
cl_ptr_vector_set(p_vec, i, NULL);
CL_PLOCK_RELEASE(sm->p_lock);
OSM_LOG_EXIT(sm->p_log);
return status;
}
/**********************************************************************
* Send Trap 64 on all new ports.
**********************************************************************/
static void state_mgr_report_new_ports(IN osm_sm_t * sm)
{
ib_gid_t port_gid;
ib_mad_notice_attr_t notice;
ib_api_status_t status;
ib_net64_t port_guid;
cl_map_item_t *p_next;
osm_port_t *p_port;
uint16_t min_lid_ho;
uint16_t max_lid_ho;
OSM_LOG_ENTER(sm->p_log);
CL_PLOCK_ACQUIRE(sm->p_lock);
p_next = cl_qmap_head(&sm->p_subn->port_guid_tbl);
while (p_next != cl_qmap_end(&sm->p_subn->port_guid_tbl)) {
p_port = (osm_port_t *) p_next;
p_next = cl_qmap_next(p_next);
if (!p_port->is_new)
continue;
port_guid = osm_port_get_guid(p_port);
/* issue a notice - trap 64 */
/* details of the notice */
notice.generic_type = 0x83; /* is generic subn mgt type */
ib_notice_set_prod_type_ho(¬ice, 4); /* A Class Manager generator */
/* endport becomes reachable */
notice.g_or_v.generic.trap_num = CL_HTON16(64);
/* The sm_base_lid is saved in network order already. */
notice.issuer_lid = sm->p_subn->sm_base_lid;
/* following C14-72.1.1 and table 119 p739 */
/* we need to provide the GID */
port_gid.unicast.prefix = sm->p_subn->opt.subnet_prefix;
port_gid.unicast.interface_id = port_guid;
memcpy(&(notice.data_details.ntc_64_67.gid), &(port_gid),
sizeof(ib_gid_t));
/* According to page 653 - the issuer gid in this case of trap
* is the SM gid, since the SM is the initiator of this trap. */
notice.issuer_gid.unicast.prefix =
sm->p_subn->opt.subnet_prefix;
notice.issuer_gid.unicast.interface_id =
sm->p_subn->sm_port_guid;
status = osm_report_notice(sm->p_log, sm->p_subn, ¬ice);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3318: "
"Error sending trap reports on GUID:0x%016"
PRIx64 " (%s)\n", port_gid.unicast.interface_id,
ib_get_err_str(status));
osm_port_get_lid_range_ho(p_port, &min_lid_ho, &max_lid_ho);
OSM_LOG(sm->p_log, OSM_LOG_INFO,
"Discovered new port with GUID:0x%016" PRIx64
" LID range [%u,%u] of node: %s\n",
cl_ntoh64(port_gid.unicast.interface_id),
min_lid_ho, max_lid_ho,
p_port->p_node ? p_port->p_node->
print_desc : "UNKNOWN");
p_port->is_new = 0;
}
CL_PLOCK_RELEASE(sm->p_lock);
OSM_LOG_EXIT(sm->p_log);
}
/**********************************************************************
Initiates a lightweight sweep of the subnet.
Used during normal sweeps after the subnet is up.
**********************************************************************/
static ib_api_status_t state_mgr_light_sweep_start(IN osm_sm_t * sm)
{
ib_api_status_t status = IB_SUCCESS;
osm_bind_handle_t h_bind;
cl_qmap_t *p_sw_tbl;
cl_map_item_t *p_next;
osm_node_t *p_node;
osm_physp_t *p_physp;
uint8_t port_num;
OSM_LOG_ENTER(sm->p_log);
p_sw_tbl = &sm->p_subn->sw_guid_tbl;
/*
* First, get the bind handle.
*/
h_bind = osm_sm_mad_ctrl_get_bind_handle(&sm->mad_ctrl);
if (h_bind == OSM_BIND_INVALID_HANDLE) {
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"No bound ports. Deferring sweep...\n");
status = IB_INVALID_STATE;
goto _exit;
}
OSM_LOG_MSG_BOX(sm->p_log, OSM_LOG_VERBOSE, "INITIATING LIGHT SWEEP");
CL_PLOCK_ACQUIRE(sm->p_lock);
cl_qmap_apply_func(p_sw_tbl, state_mgr_get_sw_info, sm);
CL_PLOCK_RELEASE(sm->p_lock);
CL_PLOCK_ACQUIRE(sm->p_lock);
cl_qmap_apply_func(&sm->p_subn->node_guid_tbl, state_mgr_get_node_desc,
sm);
CL_PLOCK_RELEASE(sm->p_lock);
/* now scan the list of physical ports that were not down but have no remote port */
CL_PLOCK_ACQUIRE(sm->p_lock);
p_next = cl_qmap_head(&sm->p_subn->node_guid_tbl);
while (p_next != cl_qmap_end(&sm->p_subn->node_guid_tbl)) {
p_node = (osm_node_t *) p_next;
p_next = cl_qmap_next(p_next);
for (port_num = 1; port_num < osm_node_get_num_physp(p_node);
port_num++) {
p_physp = osm_node_get_physp_ptr(p_node, port_num);
if (p_physp && (osm_physp_get_port_state(p_physp) !=
IB_LINK_DOWN)
&& !osm_physp_get_remote(p_physp)) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3315: "
"Unknown remote side for node 0x%016"
PRIx64
" (%s) port %u. Adding to light sweep sampling list\n",
cl_ntoh64(osm_node_get_node_guid
(p_node)),
p_node->print_desc, port_num);
osm_dump_dr_path(sm->p_log,
osm_physp_get_dr_path_ptr
(p_physp), OSM_LOG_ERROR);
state_mgr_get_remote_port_info(sm, p_physp);
}
}
}
cl_qmap_apply_func(&sm->p_subn->sm_guid_tbl, query_sm_info, sm);
CL_PLOCK_RELEASE(sm->p_lock);
_exit:
OSM_LOG_EXIT(sm->p_log);
return status;
}
/**********************************************************************
* Send SubnSet(SMInfo) SMP with HANDOVER attribute to the
* remote_sm indicated.
**********************************************************************/
static void state_mgr_send_handover(IN osm_sm_t * sm, IN osm_remote_sm_t * p_sm)
{
uint8_t payload[IB_SMP_DATA_SIZE];
ib_sm_info_t *p_smi = (ib_sm_info_t *) payload;
osm_madw_context_t context;
const osm_port_t *p_port;
ib_api_status_t status;
OSM_LOG_ENTER(sm->p_log);
/*
* Send a query of SubnSet(SMInfo) HANDOVER to the remote sm given.
*/
memset(&context, 0, sizeof(context));
p_port = osm_get_port_by_guid(sm->p_subn, p_sm->smi.guid);
if (p_port == NULL) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3316: "
"No port object on given remote_sm object\n");
goto Exit;
}
/* update the master_guid in the sm_state_mgr object according to */
/* the guid of the port where the new Master SM should reside. */
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Handing over mastership. Updating sm_state_mgr master_guid: %016"
PRIx64 " (node %s)\n", cl_ntoh64(p_port->guid),
p_port->p_node ? p_port->p_node->print_desc : "UNKNOWN");
sm->master_sm_guid = p_port->guid;
context.smi_context.port_guid = p_port->guid;
context.smi_context.set_method = TRUE;
p_smi->guid = sm->p_subn->sm_port_guid;
p_smi->act_count = cl_hton32(sm->p_subn->p_osm->stats.qp0_mads_sent);
p_smi->pri_state = (uint8_t) (sm->p_subn->sm_state |
sm->p_subn->opt.sm_priority << 4);
/*
* Return 0 for the SM key unless we authenticate the requester
* as the master SM.
*/
if (ib_sminfo_get_state(&p_sm->smi) == IB_SMINFO_STATE_MASTER) {
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Responding to master SM with real sm_key\n");
p_smi->sm_key = sm->p_subn->opt.sm_key;
} else {
/* The requester is not authenticated as master - set sm_key to zero */
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Responding to SM not master with zero sm_key\n");
p_smi->sm_key = 0;
}
CL_PLOCK_ACQUIRE(sm->p_lock);
status = osm_req_set(sm, osm_physp_get_dr_path_ptr(p_port->p_physp),
payload, sizeof(payload), IB_MAD_ATTR_SM_INFO,
IB_SMINFO_ATTR_MOD_HANDOVER, CL_DISP_MSGID_NONE,
&context);
CL_PLOCK_RELEASE(sm->p_lock);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3317: "
"Failure requesting SMInfo (%s)\n",
ib_get_err_str(status));
Exit:
OSM_LOG_EXIT(sm->p_log);
}
/**********************************************************************
Sweeps the node 1 hop away.
This sets off a "chain reaction" that causes discovery of the subnet.
Used when there is suspicion that something on the subnet has changed.
**********************************************************************/
static ib_api_status_t state_mgr_sweep_hop_1(IN osm_sm_t * sm)
{
ib_api_status_t status = IB_SUCCESS;
osm_madw_context_t context;
osm_node_t *p_node;
osm_port_t *p_port;
osm_dr_path_t hop_1_path;
ib_net64_t port_guid;
uint8_t port_num;
uint8_t path_array[IB_SUBNET_PATH_HOPS_MAX];
uint8_t num_ports;
osm_physp_t *p_ext_physp;
OSM_LOG_ENTER(sm->p_log);
/*
* First, get our own port and node objects.
*/
port_guid = sm->p_subn->sm_port_guid;
CL_ASSERT(port_guid);
/* Set the in_sweep_hop_0 flag in subn to be FALSE.
* This will indicate the sweeping to continue beyond the
* the current node.
* This is relevant for the case of SM on switch, since in the
* switch info we need to signal that the sweeping should
* continue through the switch. */
sm->p_subn->in_sweep_hop_0 = FALSE;
p_port = osm_get_port_by_guid(sm->p_subn, port_guid);
if (!p_port) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3310: "
"No SM port object\n");
status = IB_ERROR;
goto Exit;
}
p_node = p_port->p_node;
CL_ASSERT(p_node);
port_num = ib_node_info_get_local_port_num(&p_node->node_info);
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Probing hop 1 on local port %u\n", port_num);
memset(path_array, 0, sizeof(path_array));
/* the hop_1 operations depend on the type of our node.
* Currently - legal nodes that can host SM are SW and CA */
switch (osm_node_get_type(p_node)) {
case IB_NODE_TYPE_CA:
case IB_NODE_TYPE_ROUTER:
memset(&context, 0, sizeof(context));
context.ni_context.node_guid = osm_node_get_node_guid(p_node);
context.ni_context.port_num = port_num;
path_array[1] = port_num;
osm_dr_path_init(&hop_1_path, 1, path_array);
CL_PLOCK_ACQUIRE(sm->p_lock);
status = osm_req_get(sm, &hop_1_path, IB_MAD_ATTR_NODE_INFO, 0,
CL_DISP_MSGID_NONE, &context);
CL_PLOCK_RELEASE(sm->p_lock);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3311: "
"Request for NodeInfo failed (%s)\n",
ib_get_err_str(status));
break;
case IB_NODE_TYPE_SWITCH:
/* Need to go over all the ports of the switch, and send a
* node_info from them. This doesn't include the port 0 of the
* switch, which hosts the SM.
* Note: We'll send another switchInfo on port 0, since if no
* ports are connected, we still want to get some response, and
* have the subnet come up.
*/
num_ports = osm_node_get_num_physp(p_node);
for (port_num = 1; port_num < num_ports; port_num++) {
/* go through the port only if the port is not DOWN */
p_ext_physp = osm_node_get_physp_ptr(p_node, port_num);
if (p_ext_physp && ib_port_info_get_port_state
(&(p_ext_physp->port_info)) > IB_LINK_DOWN) {
memset(&context, 0, sizeof(context));
context.ni_context.node_guid =
osm_node_get_node_guid(p_node);
context.ni_context.port_num = port_num;
path_array[1] = port_num;
osm_dr_path_init(&hop_1_path, 1, path_array);
CL_PLOCK_ACQUIRE(sm->p_lock);
status = osm_req_get(sm, &hop_1_path,
IB_MAD_ATTR_NODE_INFO, 0,
CL_DISP_MSGID_NONE,
&context);
CL_PLOCK_RELEASE(sm->p_lock);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR,
"ERR 3312: "
"Request for NodeInfo failed (%s)\n",
ib_get_err_str(status));
}
}
break;
default:
OSM_LOG(sm->p_log, OSM_LOG_ERROR,
"ERR 3313: Unknown node type %d (%s)\n",
osm_node_get_type(p_node), p_node->print_desc);
}
Exit:
OSM_LOG_EXIT(sm->p_log);
return status;
}
