static qos_mad_item_t *osm_qos_mad_create(IN osm_sm_t * sm,
IN osm_physp_t * p,
IN uint32_t data_size,
IN uint8_t * p_data,
IN ib_net16_t attr_id,
IN uint32_t attr_mod)
{
qos_mad_item_t *p_mad;
osm_madw_context_t context;
osm_madw_t *p_madw;
osm_node_t *p_node;
p_node = osm_physp_get_node_ptr(p);
switch (attr_id){
case IB_MAD_ATTR_SLVL_TABLE:
context.slvl_context.node_guid = osm_node_get_node_guid(p_node);
context.slvl_context.port_guid = osm_physp_get_port_guid(p);
context.slvl_context.set_method = TRUE;
break;
case IB_MAD_ATTR_VL_ARBITRATION:
context.vla_context.node_guid = osm_node_get_node_guid(p_node);
context.vla_context.port_guid = osm_physp_get_port_guid(p);
context.vla_context.set_method = TRUE;
break;
default:
return NULL;
}
p_mad = (qos_mad_item_t *) malloc(sizeof(*p_mad));
if (!p_mad)
return NULL;
memset(p_mad, 0, sizeof(*p_mad));
p_madw = osm_prepare_req_set(sm, osm_physp_get_dr_path_ptr(p),
p_data, data_size,
attr_id, cl_hton32(attr_mod),
CL_DISP_MSGID_NONE, &context);
if (p_madw == NULL) {
free(p_mad);
return NULL;
}
p_mad->p_madw = p_madw;
return p_mad;
}
static void mpr_rcv_build_pr(IN osm_sa_t * sa,
IN const osm_alias_guid_t * p_src_alias_guid,
IN const osm_alias_guid_t * p_dest_alias_guid,
IN uint16_t src_lid_ho, IN uint16_t dest_lid_ho,
IN uint8_t preference,
IN const osm_path_parms_t * p_parms,
OUT ib_path_rec_t * p_pr)
{
const osm_physp_t *p_src_physp, *p_dest_physp;
OSM_LOG_ENTER(sa->p_log);
p_src_physp = p_src_alias_guid->p_base_port->p_physp;
p_dest_physp = p_dest_alias_guid->p_base_port->p_physp;
p_pr->dgid.unicast.prefix = osm_physp_get_subnet_prefix(p_dest_physp);
p_pr->dgid.unicast.interface_id = osm_physp_get_port_guid(p_dest_physp);
p_pr->sgid.unicast.prefix = osm_physp_get_subnet_prefix(p_src_physp);
p_pr->sgid.unicast.interface_id = osm_physp_get_port_guid(p_src_physp);
p_pr->dlid = cl_hton16(dest_lid_ho);
p_pr->slid = cl_hton16(src_lid_ho);
p_pr->hop_flow_raw &= cl_hton32(1 << 31);
p_pr->pkey = p_parms->pkey;
ib_path_rec_set_qos_class(p_pr, 0);
ib_path_rec_set_sl(p_pr, p_parms->sl);
p_pr->mtu = (uint8_t) (p_parms->mtu | 0x80);
p_pr->rate = (uint8_t) (p_parms->rate | 0x80);
/* According to 1.2 spec definition Table 205 PacketLifeTime description,
for loopback paths, packetLifeTime shall be zero. */
if (p_src_alias_guid->p_base_port == p_dest_alias_guid->p_base_port)
p_pr->pkt_life = 0x80; /* loopback */
else
p_pr->pkt_life = (uint8_t) (p_parms->pkt_life | 0x80);
p_pr->preference = preference;
/* always return num_path = 0 so this is only the reversible component */
if (p_parms->reversible)
p_pr->num_path = 0x80;
OSM_LOG_EXIT(sa->p_log);
}
static void
osm_pi_rcv_process_set(IN osm_sm_t * sm, IN osm_node_t * const p_node,
IN const uint8_t port_num, IN osm_madw_t * const p_madw)
{
osm_physp_t *p_physp;
ib_net64_t port_guid;
ib_smp_t *p_smp;
ib_port_info_t *p_pi;
osm_pi_context_t *p_context;
osm_log_level_t level;
OSM_LOG_ENTER(sm->p_log);
p_context = osm_madw_get_pi_context_ptr(p_madw);
CL_ASSERT(p_node);
p_physp = osm_node_get_physp_ptr(p_node, port_num);
CL_ASSERT(p_physp);
port_guid = osm_physp_get_port_guid(p_physp);
p_smp = osm_madw_get_smp_ptr(p_madw);
p_pi = (ib_port_info_t *) ib_smp_get_payload_ptr(p_smp);
/* check for error */
if (cl_ntoh16(p_smp->status) & 0x7fff) {
/* If port already ACTIVE, don't treat status 7 as error */
if (p_context->active_transition &&
(cl_ntoh16(p_smp->status) & 0x7fff) == 0x1c) {
level = OSM_LOG_INFO;
OSM_LOG(sm->p_log, OSM_LOG_INFO,
"Received error status 0x%x for SetResp() during ACTIVE transition\n",
cl_ntoh16(p_smp->status) & 0x7fff);
/* Should there be a subsequent Get to validate that port is ACTIVE ? */
} else {
level = OSM_LOG_ERROR;
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 0F10: "
"Received error status for SetResp()\n");
}
osm_dump_port_info(sm->p_log,
osm_node_get_node_guid(p_node),
port_guid, port_num, p_pi, level);
}
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Received logical SetResp() for GUID 0x%" PRIx64
", port num %u"
"\n\t\t\t\tfor parent node GUID 0x%" PRIx64
" TID 0x%" PRIx64 "\n",
cl_ntoh64(port_guid), port_num,
cl_ntoh64(osm_node_get_node_guid(p_node)),
cl_ntoh64(p_smp->trans_id));
osm_physp_set_port_info(p_physp, p_pi);
OSM_LOG_EXIT(sm->p_log);
}
static void osm_lash_process_switch(lash_t * p_lash, osm_switch_t * p_sw)
{
osm_log_t *p_log = &p_lash->p_osm->log;
int i, port_count;
osm_physp_t *p_current_physp, *p_remote_physp;
unsigned switch_a_lash_id, switch_b_lash_id;
OSM_LOG_ENTER(p_log);
switch_a_lash_id = get_lash_id(p_sw);
port_count = osm_node_get_num_physp(p_sw->p_node);
/* starting at port 1, ignoring management port on switch */
for (i = 1; i < port_count; i++) {
p_current_physp = osm_node_get_physp_ptr(p_sw->p_node, i);
if (p_current_physp) {
p_remote_physp = p_current_physp->p_remote_physp;
if (p_remote_physp && p_remote_physp->p_node->sw) {
int physical_port_a_num =
osm_physp_get_port_num(p_current_physp);
int physical_port_b_num =
osm_physp_get_port_num(p_remote_physp);
switch_b_lash_id =
get_lash_id(p_remote_physp->p_node->sw);
connect_switches(p_lash, switch_a_lash_id,
switch_b_lash_id,
physical_port_a_num);
OSM_LOG(p_log, OSM_LOG_VERBOSE,
"LASH SUCCESS connected G 0x%016" PRIx64
" , lash_id(%u), P(%u) " " to G 0x%016"
PRIx64 " , lash_id(%u) , P(%u)\n",
cl_ntoh64(osm_physp_get_port_guid
(p_current_physp)),
switch_a_lash_id, physical_port_a_num,
cl_ntoh64(osm_physp_get_port_guid
(p_remote_physp)),
switch_b_lash_id, physical_port_b_num);
}
}
}
OSM_LOG_EXIT(p_log);
}
/** =========================================================================
*/
void ep_guid_to_lid_tbl_rec_init(osm_port_t *p_port,
struct ep_guid_to_lid_tbl_rec *p_rec)
{
p_rec->guid = osm_physp_get_port_guid(p_port->p_physp);
p_rec->lid = osm_physp_get_base_lid(p_port->p_physp);
p_rec->lmc = osm_physp_get_lmc(p_port->p_physp);
p_rec->is_switch = (osm_node_get_type(p_port->p_node) == IB_NODE_TYPE_SWITCH);
memset(&p_rec->pad, 0, sizeof(p_rec->pad));
}
/** ===========================================================================
*/
static void extract_guid2lid(osm_port_t *p_port, uint64_t *p_offset,
struct ssa_db_extract *p_ssa_db)
{
struct ep_map_rec *p_map_rec;
#ifdef SSA_PLUGIN_VERBOSE_LOGGING
uint8_t is_fdr10_active;
ssa_log(SSA_LOG_VERBOSE, "Port GUID 0x%" PRIx64 " LID %u Port state %d"
"(%s)\n", ntohll(osm_physp_get_port_guid(p_port->p_physp)),
ntohs(osm_port_get_base_lid(p_port)),
osm_physp_get_port_state(p_port->p_physp),
(osm_physp_get_port_state(p_port->p_physp) < 5 ?
port_state_str[osm_physp_get_port_state(p_port->p_physp)] :
"???"));
is_fdr10_active =
p_port->p_physp->ext_port_info.link_speed_active & FDR10;
ssa_log(SSA_LOG_VERBOSE, "FDR10 %s active\n",
is_fdr10_active ? "" : "not");
#endif
/* check for valid LID first */
if ((ntohs(osm_port_get_base_lid(p_port)) < IB_LID_UCAST_START_HO) ||
(ntohs(osm_port_get_base_lid(p_port)) > IB_LID_UCAST_END_HO)) {
ssa_log(SSA_LOG_VERBOSE, "Port GUID 0x%" PRIx64
" has invalid LID %u\n",
ntohll(osm_physp_get_port_guid(p_port->p_physp)),
ntohs(osm_port_get_base_lid(p_port)));
}
smdb_guid2lid_init(p_port, &p_ssa_db->p_guid_to_lid_tbl[*p_offset]);
p_map_rec = ep_map_rec_init(*p_offset);
if (!p_map_rec) {
/* add memory allocation failure handling */
ssa_log(SSA_LOG_VERBOSE, "Quick MAP rec memory allocation failed\n");
}
cl_qmap_insert(&p_ssa_db->ep_guid_to_lid_tbl,
osm_physp_get_port_guid(p_port->p_physp),
&p_map_rec->map_item);
*p_offset = *p_offset + 1;
}
static void ni_rcv_get_port_info(IN osm_sm_t * sm, IN osm_node_t * node,
IN const osm_madw_t * madw)
{
osm_madw_context_t context;
osm_physp_t *physp;
ib_node_info_t *ni;
unsigned port;
ib_api_status_t status;
int mlnx_epi_supported = 0;
ni = ib_smp_get_payload_ptr(osm_madw_get_smp_ptr(madw));
port = ib_node_info_get_local_port_num(ni);
if (sm->p_subn->opt.fdr10)
mlnx_epi_supported = is_mlnx_ext_port_info_supported(ni->device_id);
physp = osm_node_get_physp_ptr(node, port);
if (!physp) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR OD1E: "
"Failed to find physp for port %d of Node GUID 0x%"
PRIx64 "\n", port,
cl_ntoh64(osm_node_get_node_guid(node)));
return;
}
context.pi_context.node_guid = osm_node_get_node_guid(node);
context.pi_context.port_guid = osm_physp_get_port_guid(physp);
context.pi_context.set_method = FALSE;
context.pi_context.light_sweep = FALSE;
context.pi_context.active_transition = FALSE;
context.pi_context.client_rereg = FALSE;
status = osm_req_get(sm, osm_physp_get_dr_path_ptr(physp),
IB_MAD_ATTR_PORT_INFO, cl_hton32(port),
TRUE, 0, CL_DISP_MSGID_NONE, &context);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR OD02: "
"Failure initiating PortInfo request (%s)\n",
ib_get_err_str(status));
if (mlnx_epi_supported) {
status = osm_req_get(sm,
osm_physp_get_dr_path_ptr(physp),
IB_MAD_ATTR_MLNX_EXTENDED_PORT_INFO,
cl_hton32(port),
TRUE, 0, CL_DISP_MSGID_NONE, &context);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 0D0B: "
"Failure initiating MLNX ExtPortInfo request (%s)\n",
ib_get_err_str(status));
}
}
/**********************************************************************
The plock must be held before calling this function.
**********************************************************************/
static void pi_rcv_process_switch_port0(IN osm_sm_t * sm,
IN osm_node_t * p_node,
IN osm_physp_t * p_physp,
IN ib_port_info_t * p_pi)
{
ib_api_status_t status;
osm_madw_context_t context;
uint8_t port, num_ports;
OSM_LOG_ENTER(sm->p_log);
if (p_physp->need_update)
sm->p_subn->ignore_existing_lfts = TRUE;
pi_rcv_check_and_fix_lid(sm->p_log, p_pi, p_physp);
/* Update the PortInfo attribute */
osm_physp_set_port_info(p_physp, p_pi, sm);
/* Determine if base switch port 0 */
if (p_node->sw &&
!ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info))
/* PortState is not used on BSP0 but just in case it is DOWN */
p_physp->port_info = *p_pi;
/* Now, query PortInfo for the switch external ports */
num_ports = osm_node_get_num_physp(p_node);
context.pi_context.node_guid = osm_node_get_node_guid(p_node);
context.pi_context.port_guid = osm_physp_get_port_guid(p_physp);
context.pi_context.set_method = FALSE;
context.pi_context.light_sweep = FALSE;
context.pi_context.active_transition = FALSE;
context.pi_context.client_rereg = FALSE;
for (port = 1; port < num_ports; port++) {
status = osm_req_get(sm, osm_physp_get_dr_path_ptr(p_physp),
IB_MAD_ATTR_PORT_INFO, cl_hton32(port),
FALSE,
ib_port_info_get_m_key(&p_physp->port_info),
CL_DISP_MSGID_NONE, &context);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 0F16: "
"Failure initiating PortInfo request (%s)\n",
ib_get_err_str(status));
}
pi_rcv_process_endport(sm, p_physp, p_pi);
OSM_LOG_EXIT(sm->p_log);
}
static void sa_pkey_create(IN osm_sa_t * sa, IN osm_physp_t * p_physp,
IN osm_pkey_search_ctxt_t * p_ctxt,
IN uint16_t block)
{
osm_pkey_item_t *p_rec_item;
uint16_t lid;
ib_pkey_table_t *tbl;
OSM_LOG_ENTER(sa->p_log);
p_rec_item = malloc(sizeof(*p_rec_item));
if (p_rec_item == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4602: "
"rec_item alloc failed\n");
goto Exit;
}
if (p_physp->p_node->node_info.node_type != IB_NODE_TYPE_SWITCH)
lid = p_physp->port_info.base_lid;
else
lid = osm_node_get_base_lid(p_physp->p_node, 0);
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"New P_Key table for: port 0x%016" PRIx64
", lid %u, port %u Block:%u\n",
cl_ntoh64(osm_physp_get_port_guid(p_physp)),
cl_ntoh16(lid), osm_physp_get_port_num(p_physp), block);
memset(p_rec_item, 0, sizeof(*p_rec_item));
p_rec_item->rec.lid = lid;
p_rec_item->rec.block_num = block;
p_rec_item->rec.port_num = osm_physp_get_port_num(p_physp);
/* FIXME: There are ninf.PartitionCap or swinf.PartitionEnforcementCap
pkey entries so everything in that range is a valid block number
even if opensm is not using it. Return 0. However things outside
that range should return no entries.. Not sure how to figure that
here? The range of pkey_tbl can be less than the cap, so
this falsely triggers. */
tbl = osm_pkey_tbl_block_get(osm_physp_get_pkey_tbl(p_physp), block);
if (tbl)
p_rec_item->rec.pkey_tbl = *tbl;
cl_qlist_insert_tail(p_ctxt->p_list, &p_rec_item->list_item);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
static void
__osm_sa_pkey_create(IN osm_sa_t * sa,
IN osm_physp_t * const p_physp,
IN osm_pkey_search_ctxt_t * const p_ctxt,
IN uint16_t block)
{
osm_pkey_item_t *p_rec_item;
uint16_t lid;
ib_api_status_t status = IB_SUCCESS;
OSM_LOG_ENTER(sa->p_log);
p_rec_item = malloc(sizeof(*p_rec_item));
if (p_rec_item == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4602: "
"rec_item alloc failed\n");
status = IB_INSUFFICIENT_RESOURCES;
goto Exit;
}
if (p_physp->p_node->node_info.node_type != IB_NODE_TYPE_SWITCH)
lid = p_physp->port_info.base_lid;
else
lid = osm_node_get_base_lid(p_physp->p_node, 0);
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"New P_Key table for: port 0x%016" PRIx64
", lid %u, port %u Block:%u\n",
cl_ntoh64(osm_physp_get_port_guid(p_physp)),
cl_ntoh16(lid), osm_physp_get_port_num(p_physp), block);
memset(p_rec_item, 0, sizeof(*p_rec_item));
p_rec_item->rec.lid = lid;
p_rec_item->rec.block_num = block;
p_rec_item->rec.port_num = osm_physp_get_port_num(p_physp);
p_rec_item->rec.pkey_tbl =
*(osm_pkey_tbl_block_get(osm_physp_get_pkey_tbl(p_physp), block));
cl_qlist_insert_tail(p_ctxt->p_list, &p_rec_item->list_item);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
static void
__osm_sa_slvl_create(IN osm_sa_t * sa,
IN const osm_physp_t * const p_physp,
IN osm_slvl_search_ctxt_t * const p_ctxt,
IN uint8_t in_port_idx)
{
osm_slvl_item_t *p_rec_item;
uint16_t lid;
ib_api_status_t status = IB_SUCCESS;
OSM_LOG_ENTER(sa->p_log);
p_rec_item = malloc(sizeof(*p_rec_item));
if (p_rec_item == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2602: "
"rec_item alloc failed\n");
status = IB_INSUFFICIENT_RESOURCES;
goto Exit;
}
if (p_physp->p_node->node_info.node_type != IB_NODE_TYPE_SWITCH)
lid = p_physp->port_info.base_lid;
else
lid = osm_node_get_base_lid(p_physp->p_node, 0);
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"New SLtoVL Map for: OUT port 0x%016" PRIx64
", lid 0x%X, port %u to In Port:%u\n",
cl_ntoh64(osm_physp_get_port_guid(p_physp)),
cl_ntoh16(lid), osm_physp_get_port_num(p_physp), in_port_idx);
memset(p_rec_item, 0, sizeof(*p_rec_item));
p_rec_item->rec.lid = lid;
p_rec_item->rec.out_port_num = osm_physp_get_port_num(p_physp);
p_rec_item->rec.in_port_num = in_port_idx;
p_rec_item->rec.slvl_tbl =
*(osm_physp_get_slvl_tbl(p_physp, in_port_idx));
cl_qlist_insert_tail(p_ctxt->p_list, &p_rec_item->list_item);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
static ib_api_status_t sl2vl_update_table(osm_sm_t * sm,
osm_physp_t * p, uint8_t in_port,
uint8_t out_port,
unsigned force_update,
const ib_slvl_table_t * sl2vl_table)
{
osm_madw_context_t context;
ib_slvl_table_t tbl, *p_tbl;
osm_node_t *p_node = osm_physp_get_node_ptr(p);
uint32_t attr_mod;
unsigned vl_mask;
uint8_t vl1, vl2;
int i;
vl_mask = (1 << (ib_port_info_get_op_vls(&p->port_info) - 1)) - 1;
for (i = 0; i < IB_MAX_NUM_VLS / 2; i++) {
vl1 = sl2vl_table->raw_vl_by_sl[i] >> 4;
vl2 = sl2vl_table->raw_vl_by_sl[i] & 0xf;
if (vl1 != 15)
vl1 &= vl_mask;
if (vl2 != 15)
vl2 &= vl_mask;
tbl.raw_vl_by_sl[i] = (vl1 << 4) | vl2;
}
if (!force_update && (p_tbl = osm_physp_get_slvl_tbl(p, in_port)) &&
!memcmp(p_tbl, &tbl, sizeof(tbl)))
return IB_SUCCESS;
context.slvl_context.node_guid = osm_node_get_node_guid(p_node);
context.slvl_context.port_guid = osm_physp_get_port_guid(p);
context.slvl_context.set_method = TRUE;
attr_mod = in_port << 8 | out_port;
return osm_req_set(sm, osm_physp_get_dr_path_ptr(p),
(uint8_t *) & tbl, sizeof(tbl),
IB_MAD_ATTR_SLVL_TABLE,
cl_hton32(attr_mod), CL_DISP_MSGID_NONE, &context);
}
/*
* QoS primitives
*/
static ib_api_status_t vlarb_update_table_block(osm_sm_t * sm,
osm_physp_t * p,
uint8_t port_num,
unsigned force_update,
const ib_vl_arb_table_t *
table_block,
unsigned block_length,
unsigned block_num)
{
ib_vl_arb_table_t block;
osm_madw_context_t context;
uint32_t attr_mod;
unsigned vl_mask, i;
vl_mask = (1 << (ib_port_info_get_op_vls(&p->port_info) - 1)) - 1;
memset(&block, 0, sizeof(block));
memcpy(&block, table_block, block_length * sizeof(block.vl_entry[0]));
for (i = 0; i < block_length; i++)
block.vl_entry[i].vl &= vl_mask;
if (!force_update &&
!memcmp(&p->vl_arb[block_num], &block,
block_length * sizeof(block.vl_entry[0])))
return IB_SUCCESS;
context.vla_context.node_guid =
osm_node_get_node_guid(osm_physp_get_node_ptr(p));
context.vla_context.port_guid = osm_physp_get_port_guid(p);
context.vla_context.set_method = TRUE;
attr_mod = ((block_num + 1) << 16) | port_num;
return osm_req_set(sm, osm_physp_get_dr_path_ptr(p),
(uint8_t *) & block, sizeof(block),
IB_MAD_ATTR_VL_ARBITRATION,
cl_hton32(attr_mod), CL_DISP_MSGID_NONE, &context);
}
static void print_node_report(cl_map_item_t * item, FILE * file, void *cxt)
{
osm_node_t *p_node = (osm_node_t *) item;
osm_opensm_t *osm = cxt;
const osm_physp_t *p_physp, *p_remote_physp;
const ib_port_info_t *p_pi;
uint8_t port_num;
uint32_t num_ports;
uint8_t node_type;
node_type = osm_node_get_type(p_node);
num_ports = osm_node_get_num_physp(p_node);
port_num = node_type == IB_NODE_TYPE_SWITCH ? 0 : 1;
for (; port_num < num_ports; port_num++) {
p_physp = osm_node_get_physp_ptr(p_node, port_num);
if (!p_physp)
continue;
fprintf(file, "%-11s : %s : %02X :",
osm_get_manufacturer_str(cl_ntoh64
(osm_node_get_node_guid
(p_node))),
osm_get_node_type_str_fixed_width(node_type), port_num);
p_pi = &p_physp->port_info;
/*
* Port state is not defined for base switch port 0
*/
if (port_num == 0 &&
ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) == FALSE)
fprintf(file, " :");
else
fprintf(file, " %s :",
osm_get_port_state_str_fixed_width
(ib_port_info_get_port_state(p_pi)));
/*
* LID values are only meaningful in select cases.
*/
if (ib_port_info_get_port_state(p_pi) != IB_LINK_DOWN
&& ((node_type == IB_NODE_TYPE_SWITCH && port_num == 0)
|| node_type != IB_NODE_TYPE_SWITCH))
fprintf(file, " %04X : %01X :",
cl_ntoh16(p_pi->base_lid),
ib_port_info_get_lmc(p_pi));
else
fprintf(file, " : :");
if (port_num == 0 &&
ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) == FALSE)
fprintf(file, " : : ");
else
fprintf(file, " %s : %s : %s ",
osm_get_mtu_str
(ib_port_info_get_neighbor_mtu(p_pi)),
osm_get_lwa_str(p_pi->link_width_active),
osm_get_lsa_str
(ib_port_info_get_link_speed_active(p_pi),
ib_port_info_get_link_speed_ext_active(p_pi),
ib_port_info_get_port_state(p_pi),
p_physp->ext_port_info.link_speed_active & FDR10));
if (osm_physp_get_port_guid(p_physp) == osm->subn.sm_port_guid)
fprintf(file, "* %016" PRIx64 " *",
cl_ntoh64(osm_physp_get_port_guid(p_physp)));
else
fprintf(file, ": %016" PRIx64 " :",
cl_ntoh64(osm_physp_get_port_guid(p_physp)));
if (port_num
&& (ib_port_info_get_port_state(p_pi) != IB_LINK_DOWN)) {
p_remote_physp = osm_physp_get_remote(p_physp);
if (p_remote_physp)
fprintf(file, " %016" PRIx64 " (%02X)",
cl_ntoh64(osm_physp_get_port_guid
(p_remote_physp)),
osm_physp_get_port_num(p_remote_physp));
else
fprintf(file, " UNKNOWN");
}
fprintf(file, "\n");
}
fprintf(file, "------------------------------------------------------"
"------------------------------------------------\n");
}
static void ucast_cache_validate(osm_ucast_mgr_t * p_mgr)
{
cache_switch_t *p_cache_sw;
cache_switch_t *p_remote_cache_sw;
unsigned port_num;
unsigned max_ports;
uint8_t remote_node_type;
uint16_t lid_ho;
uint16_t remote_lid_ho;
osm_switch_t *p_sw;
osm_switch_t *p_remote_sw;
osm_node_t *p_node;
osm_physp_t *p_physp;
osm_physp_t *p_remote_physp;
osm_port_t *p_remote_port;
cl_qmap_t *p_sw_tbl;
OSM_LOG_ENTER(p_mgr->p_log);
if (!p_mgr->cache_valid)
goto Exit;
/* If there are no switches in the subnet, we are done */
p_sw_tbl = &p_mgr->p_subn->sw_guid_tbl;
if (cl_qmap_count(p_sw_tbl) == 0) {
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
/*
* Scan all the physical switch ports in the subnet.
* If the port need_update flag is on, check whether
* it's just some node/port reset or a cached topology
* change. Otherwise the cache is invalid.
*/
for (p_sw = (osm_switch_t *) cl_qmap_head(p_sw_tbl);
p_sw != (osm_switch_t *) cl_qmap_end(p_sw_tbl);
p_sw = (osm_switch_t *) cl_qmap_next(&p_sw->map_item)) {
p_node = p_sw->p_node;
lid_ho = cl_ntoh16(osm_node_get_base_lid(p_node, 0));
p_cache_sw = cache_get_sw(p_mgr, lid_ho);
max_ports = osm_node_get_num_physp(p_node);
/* skip port 0 */
for (port_num = 1; port_num < max_ports; port_num++) {
p_physp = osm_node_get_physp_ptr(p_node, port_num);
if (!p_physp || !p_physp->p_remote_physp ||
!osm_physp_link_exists(p_physp,
p_physp->p_remote_physp))
/* no valid link */
continue;
/*
* While scanning all the physical ports in the subnet,
* mark corresponding leaf switches in the cache.
*/
if (p_cache_sw &&
!p_cache_sw->dropped &&
!cache_sw_is_leaf(p_cache_sw) &&
p_physp->p_remote_physp->p_node &&
osm_node_get_type(p_physp->p_remote_physp->
p_node) != IB_NODE_TYPE_SWITCH)
cache_sw_set_leaf(p_cache_sw);
if (!p_physp->need_update)
continue;
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Checking switch lid %u, port %u\n",
lid_ho, port_num);
p_remote_physp = osm_physp_get_remote(p_physp);
remote_node_type =
osm_node_get_type(p_remote_physp->p_node);
if (remote_node_type == IB_NODE_TYPE_SWITCH)
remote_lid_ho =
cl_ntoh16(osm_node_get_base_lid
(p_remote_physp->p_node, 0));
else
remote_lid_ho =
cl_ntoh16(osm_node_get_base_lid
(p_remote_physp->p_node,
osm_physp_get_port_num
(p_remote_physp)));
if (!p_cache_sw ||
port_num >= p_cache_sw->num_ports ||
!p_cache_sw->ports[port_num].remote_lid_ho) {
/*
* There is some uncached change on the port.
* In general, the reasons might be as follows:
* - switch reset
* - port reset (or port down/up)
* - quick connection location change
* - new link (or new switch)
*
* First two reasons allow cache usage, while
* the last two reasons should invalidate cache.
*
* In case of quick connection location change,
* cache would have been invalidated by
* osm_ucast_cache_check_new_link() function.
*
* In case of new link between two known nodes,
* cache also would have been invalidated by
* osm_ucast_cache_check_new_link() function.
*
* Another reason is cached link between two
* known switches went back. In this case the
* osm_ucast_cache_check_new_link() function would
* clear both sides of the link from the cache
* during the discovery process, so effectively
* this would be equivalent to port reset.
*
* So three possible reasons remain:
* - switch reset
* - port reset (or port down/up)
* - link of a new switch
*
* To validate cache, we need to check only the
* third reason - link of a new node/switch:
* - If this is the local switch that is new,
* then it should have (p_sw->need_update == 2).
* - If the remote node is switch and it's new,
* then it also should have
* (p_sw->need_update == 2).
* - If the remote node is CA/RTR and it's new,
* then its port should have is_new flag on.
*/
if (p_sw->need_update == 2) {
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"New switch found (lid %u)\n",
lid_ho);
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
if (remote_node_type == IB_NODE_TYPE_SWITCH) {
p_remote_sw =
p_remote_physp->p_node->sw;
if (p_remote_sw->need_update == 2) {
/* this could also be case of
switch coming back with an
additional link that it
didn't have before */
OSM_LOG(p_mgr->p_log,
OSM_LOG_DEBUG,
"New switch/link found (lid %u)\n",
remote_lid_ho);
osm_ucast_cache_invalidate
(p_mgr);
goto Exit;
}
} else {
/*
* Remote node is CA/RTR.
* Get p_port of the remote node and
* check its p_port->is_new flag.
*/
p_remote_port =
osm_get_port_by_guid(p_mgr->p_subn,
osm_physp_get_port_guid
(p_remote_physp));
if (p_remote_port->is_new) {
OSM_LOG(p_mgr->p_log,
OSM_LOG_DEBUG,
"New CA/RTR found (lid %u)\n",
remote_lid_ho);
osm_ucast_cache_invalidate
(p_mgr);
goto Exit;
}
}
} else {
/*
* The change on the port is cached.
* In general, the reasons might be as follows:
* - link between two known nodes went back
* - one or more nodes went back, causing all
* the links to reappear
*
* If it was link that went back, then this case
* would have been taken care of during the
* discovery by osm_ucast_cache_check_new_link(),
* so it's some node that went back.
*/
if ((p_cache_sw->ports[port_num].is_leaf &&
remote_node_type == IB_NODE_TYPE_SWITCH) ||
(!p_cache_sw->ports[port_num].is_leaf &&
remote_node_type != IB_NODE_TYPE_SWITCH)) {
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Remote node type change on switch lid %u, port %u\n",
lid_ho, port_num);
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
if (p_cache_sw->ports[port_num].remote_lid_ho !=
remote_lid_ho) {
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Remote lid change on switch lid %u, port %u"
"(was %u, now %u)\n",
lid_ho, port_num,
p_cache_sw->ports[port_num].
remote_lid_ho, remote_lid_ho);
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
/*
* We don't care who is the node that has
* reappeared in the subnet (local or remote).
* What's important that the cached link matches
* the real fabrics link.
* Just clean it from cache.
*/
p_cache_sw->ports[port_num].remote_lid_ho = 0;
p_cache_sw->ports[port_num].is_leaf = FALSE;
if (p_cache_sw->dropped) {
cache_restore_ucast_info(p_mgr,
p_cache_sw,
p_sw);
p_cache_sw->dropped = FALSE;
}
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Restored link from cache: lid %u, port %u to lid %u\n",
lid_ho, port_num, remote_lid_ho);
}
}
}
/* Remove all the cached switches that
have all their ports restored */
cache_cleanup_switches(p_mgr);
/*
* Done scanning all the physical switch ports in the subnet.
* Now we need to check the other side:
* Scan all the cached switches and their ports:
* - If the cached switch is missing in the subnet
* (dropped flag is on), check that it's a leaf switch.
* If it's not a leaf, the cache is invalid, because
* cache can tolerate only leaf switch removal.
* - If the cached switch exists in fabric, check all
* its cached ports. These cached ports represent
* missing link in the fabric.
* The missing links that can be tolerated are:
* + link to missing CA/RTR
* + link to missing leaf switch
*/
for (p_cache_sw = (cache_switch_t *) cl_qmap_head(&p_mgr->cache_sw_tbl);
p_cache_sw != (cache_switch_t *) cl_qmap_end(&p_mgr->cache_sw_tbl);
p_cache_sw =
(cache_switch_t *) cl_qmap_next(&p_cache_sw->map_item)) {
if (p_cache_sw->dropped) {
if (!cache_sw_is_leaf(p_cache_sw)) {
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Missing non-leaf switch (lid %u)\n",
cache_sw_get_base_lid_ho(p_cache_sw));
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Missing leaf switch (lid %u) - "
"continuing validation\n",
cache_sw_get_base_lid_ho(p_cache_sw));
continue;
}
for (port_num = 1; port_num < p_cache_sw->num_ports; port_num++) {
if (!p_cache_sw->ports[port_num].remote_lid_ho)
continue;
if (p_cache_sw->ports[port_num].is_leaf) {
CL_ASSERT(cache_sw_is_leaf(p_cache_sw));
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Switch lid %u, port %u: missing link to CA/RTR - "
"continuing validation\n",
cache_sw_get_base_lid_ho(p_cache_sw),
port_num);
continue;
}
p_remote_cache_sw = cache_get_sw(p_mgr,
p_cache_sw->
ports[port_num].
remote_lid_ho);
if (!p_remote_cache_sw || !p_remote_cache_sw->dropped) {
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Switch lid %u, port %u: missing link to existing switch\n",
cache_sw_get_base_lid_ho(p_cache_sw),
port_num);
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
if (!cache_sw_is_leaf(p_remote_cache_sw)) {
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Switch lid %u, port %u: missing link to non-leaf switch\n",
cache_sw_get_base_lid_ho(p_cache_sw),
port_num);
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
/*
* At this point we know that the missing link is to
* a leaf switch. However, one case deserves a special
* treatment. If there was a link between two leaf
* switches, then missing leaf switch might break
* routing. It is possible that there are routes
* that use leaf switches to get from switch to switch
* and not just to get to the CAs behind the leaf switch.
*/
if (cache_sw_is_leaf(p_cache_sw) &&
cache_sw_is_leaf(p_remote_cache_sw)) {
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Switch lid %u, port %u: missing leaf-2-leaf link\n",
cache_sw_get_base_lid_ho(p_cache_sw),
port_num);
osm_ucast_cache_invalidate(p_mgr);
goto Exit;
}
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG,
"Switch lid %u, port %u: missing remote leaf switch - "
"continuing validation\n",
cache_sw_get_base_lid_ho(p_cache_sw),
port_num);
}
}
OSM_LOG(p_mgr->p_log, OSM_LOG_DEBUG, "Unicast cache is valid\n");
ucast_cache_dump(p_mgr);
Exit:
OSM_LOG_EXIT(p_mgr->p_log);
} /* osm_ucast_cache_validate() */
void osm_lr_rcv_process(IN void *context, IN void *data)
{
osm_sa_t *sa = context;
osm_madw_t *p_madw = data;
const ib_link_record_t *p_lr;
const ib_sa_mad_t *p_sa_mad;
const osm_port_t *p_src_port;
const osm_port_t *p_dest_port;
cl_qlist_t lr_list;
ib_net16_t status;
osm_physp_t *p_req_physp;
OSM_LOG_ENTER(sa->p_log);
CL_ASSERT(p_madw);
p_sa_mad = osm_madw_get_sa_mad_ptr(p_madw);
p_lr = ib_sa_mad_get_payload_ptr(p_sa_mad);
CL_ASSERT(p_sa_mad->attr_id == IB_MAD_ATTR_LINK_RECORD);
/* we only support SubnAdmGet and SubnAdmGetTable methods */
if (p_sa_mad->method != IB_MAD_METHOD_GET &&
p_sa_mad->method != IB_MAD_METHOD_GETTABLE) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 1804: "
"Unsupported Method (%s)\n",
ib_get_sa_method_str(p_sa_mad->method));
osm_sa_send_error(sa, p_madw, IB_MAD_STATUS_UNSUP_METHOD_ATTR);
goto Exit;
}
/* update the requester physical port */
p_req_physp = osm_get_physp_by_mad_addr(sa->p_log, sa->p_subn,
osm_madw_get_mad_addr_ptr
(p_madw));
if (p_req_physp == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 1805: "
"Cannot find requester physical port\n");
goto Exit;
}
if (OSM_LOG_IS_ACTIVE_V2(sa->p_log, OSM_LOG_DEBUG)) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Requester port GUID 0x%" PRIx64 "\n",
cl_ntoh64(osm_physp_get_port_guid(p_req_physp)));
osm_dump_link_record_v2(sa->p_log, p_lr, FILE_ID, OSM_LOG_DEBUG);
}
cl_qlist_init(&lr_list);
/*
Most SA functions (including this one) are read-only on the
subnet object, so we grab the lock non-exclusively.
*/
cl_plock_acquire(sa->p_lock);
status = lr_rcv_get_end_points(sa, p_madw, &p_src_port, &p_dest_port);
if (status == IB_SA_MAD_STATUS_SUCCESS)
lr_rcv_get_port_links(sa, p_lr, p_src_port, p_dest_port,
p_sa_mad->comp_mask, &lr_list,
p_req_physp);
cl_plock_release(sa->p_lock);
osm_sa_respond(sa, p_madw, sizeof(ib_link_record_t), &lr_list);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
/**********************************************************************
The plock must be held before calling this function.
**********************************************************************/
static void ni_rcv_set_links(IN osm_sm_t * sm, osm_node_t * p_node,
const uint8_t port_num,
const osm_ni_context_t * p_ni_context)
{
osm_node_t *p_neighbor_node;
osm_physp_t *p_physp, *p_remote_physp;
OSM_LOG_ENTER(sm->p_log);
/*
A special case exists in which the node we're trying to
link is our own node. In this case, the guid value in
the ni_context will be zero.
*/
if (p_ni_context->node_guid == 0) {
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Nothing to link for our own node 0x%" PRIx64 "\n",
cl_ntoh64(osm_node_get_node_guid(p_node)));
goto _exit;
}
p_neighbor_node = osm_get_node_by_guid(sm->p_subn,
p_ni_context->node_guid);
if (PF(!p_neighbor_node)) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 0D10: "
"Unexpected removal of neighbor node 0x%" PRIx64 "\n",
cl_ntoh64(p_ni_context->node_guid));
goto _exit;
}
/* When setting the link, ports on both
sides of the link should be initialized */
CL_ASSERT(osm_node_link_has_valid_ports(p_node, port_num,
p_neighbor_node,
p_ni_context->port_num));
if (osm_node_link_exists(p_node, port_num,
p_neighbor_node, p_ni_context->port_num)) {
OSM_LOG(sm->p_log, OSM_LOG_DEBUG, "Link already exists\n");
goto _exit;
}
p_physp = osm_node_get_physp_ptr(p_node, port_num);
if (!p_physp) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR OD0E: "
"Failed to find physp for port %d of Node GUID 0x%"
PRIx64 "\n", port_num,
cl_ntoh64(osm_node_get_node_guid(p_node)));
goto _exit;
}
/*
* If the link went UP, after we already discovered it, we shouldn't
* set the link between the ports and resweep.
*/
if (osm_physp_get_port_state(p_physp) == IB_LINK_DOWN &&
p_node->physp_discovered[port_num]) {
/* Link down on another side. Don't create a link*/
p_node->physp_discovered[port_num] = 0;
sm->p_subn->force_heavy_sweep = TRUE;
goto _exit;
}
if (osm_node_has_any_link(p_node, port_num) &&
sm->p_subn->force_heavy_sweep == FALSE &&
(!p_ni_context->dup_count ||
(p_ni_context->dup_node_guid == osm_node_get_node_guid(p_node) &&
p_ni_context->dup_port_num == port_num))) {
/*
Uh oh...
This could be reconnected ports, but also duplicated GUID
(2 nodes have the same guid) or a 12x link with lane reversal
that is not configured correctly.
We will try to recover by querying NodeInfo again.
In order to catch even fast port moving to new location(s)
and back we will count up to 5.
Some crazy reconnections (newly created switch loop right
before targeted CA) will not be catched this way. So in worst
case - report GUID duplication and request new discovery.
When switch node is targeted NodeInfo querying will be done
in opposite order, this is much stronger check, unfortunately
it is impossible with CAs.
*/
p_physp = osm_node_get_physp_ptr(p_node, port_num);
if (!p_physp) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR OD0F: "
"Failed to find physp for port %d of Node GUID 0x%"
PRIx64 "\n", port_num,
cl_ntoh64(osm_node_get_node_guid(p_node)));
goto _exit;
}
if (p_ni_context->dup_count > 5) {
report_duplicated_guid(sm, p_physp, p_neighbor_node,
p_ni_context->port_num);
sm->p_subn->force_heavy_sweep = TRUE;
} else if (p_node->sw)
requery_dup_node_info(sm, p_physp->p_remote_physp,
p_ni_context->dup_count + 1);
else
requery_dup_node_info(sm, p_physp,
p_ni_context->dup_count + 1);
}
/*
When there are only two nodes with exact same guids (connected back
to back) - the previous check for duplicated guid will not catch
them. But the link will be from the port to itself...
Enhanced Port 0 is an exception to this
*/
if (osm_node_get_node_guid(p_node) == p_ni_context->node_guid &&
port_num == p_ni_context->port_num &&
port_num != 0 && cl_qmap_count(&sm->p_subn->sw_guid_tbl) == 0) {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Duplicate GUID found by link from a port to itself:"
"node 0x%" PRIx64 ", port number %u\n",
cl_ntoh64(osm_node_get_node_guid(p_node)), port_num);
p_physp = osm_node_get_physp_ptr(p_node, port_num);
if (!p_physp) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR OD1D: "
"Failed to find physp for port %d of Node GUID 0x%"
PRIx64 "\n", port_num,
cl_ntoh64(osm_node_get_node_guid(p_node)));
goto _exit;
}
osm_dump_dr_path_v2(sm->p_log, osm_physp_get_dr_path_ptr(p_physp),
FILE_ID, OSM_LOG_VERBOSE);
if (sm->p_subn->opt.exit_on_fatal == TRUE) {
osm_log_v2(sm->p_log, OSM_LOG_SYS, FILE_ID,
"Errors on subnet. Duplicate GUID found "
"by link from a port to itself. "
"See verbose opensm.log for more details\n");
exit(1);
}
}
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Creating new link between:\n\t\t\t\tnode 0x%" PRIx64
", port number %u and\n\t\t\t\tnode 0x%" PRIx64
", port number %u\n",
cl_ntoh64(osm_node_get_node_guid(p_node)), port_num,
cl_ntoh64(p_ni_context->node_guid), p_ni_context->port_num);
if (sm->ucast_mgr.cache_valid)
osm_ucast_cache_check_new_link(&sm->ucast_mgr, p_node, port_num,
p_neighbor_node,
p_ni_context->port_num);
p_physp = osm_node_get_physp_ptr(p_node, port_num);
p_remote_physp = osm_node_get_physp_ptr(p_neighbor_node,
p_ni_context->port_num);
if (!p_physp || !p_remote_physp)
goto _exit;
osm_node_link(p_node, port_num, p_neighbor_node, p_ni_context->port_num);
osm_db_neighbor_set(sm->p_subn->p_neighbor,
cl_ntoh64(osm_physp_get_port_guid(p_physp)),
port_num,
cl_ntoh64(osm_physp_get_port_guid(p_remote_physp)),
p_ni_context->port_num);
osm_db_neighbor_set(sm->p_subn->p_neighbor,
cl_ntoh64(osm_physp_get_port_guid(p_remote_physp)),
p_ni_context->port_num,
cl_ntoh64(osm_physp_get_port_guid(p_physp)),
port_num);
_exit:
OSM_LOG_EXIT(sm->p_log);
}
static boolean_t
__osm_link_mgr_set_physp_pi(osm_sm_t * sm,
IN osm_physp_t * const p_physp,
IN uint8_t const port_state)
{
uint8_t payload[IB_SMP_DATA_SIZE];
ib_port_info_t *const p_pi = (ib_port_info_t *) payload;
const ib_port_info_t *p_old_pi;
osm_madw_context_t context;
osm_node_t *p_node;
ib_api_status_t status;
uint8_t port_num;
uint8_t mtu;
uint8_t op_vls;
boolean_t esp0 = FALSE;
boolean_t send_set = FALSE;
osm_physp_t *p_remote_physp;
OSM_LOG_ENTER(sm->p_log);
p_node = osm_physp_get_node_ptr(p_physp);
port_num = osm_physp_get_port_num(p_physp);
if (port_num == 0) {
/*
CAs don't have a port 0, and for switch port 0,
we need to check if this is enhanced or base port 0.
For base port 0 the following parameters are not valid (p822, table 145).
*/
if (!p_node->sw) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 4201: "
"Cannot find switch by guid: 0x%" PRIx64 "\n",
cl_ntoh64(p_node->node_info.node_guid));
goto Exit;
}
if (ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info)
== FALSE) {
/* This means the switch doesn't support enhanced port 0.
Can skip it. */
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Skipping port 0, GUID 0x%016" PRIx64 "\n",
cl_ntoh64(osm_physp_get_port_guid(p_physp)));
goto Exit;
}
esp0 = TRUE;
}
/*
PAST THIS POINT WE ARE HANDLING EITHER A NON PORT 0 OR ENHANCED PORT 0
*/
p_old_pi = &p_physp->port_info;
memset(payload, 0, IB_SMP_DATA_SIZE);
memcpy(payload, p_old_pi, sizeof(ib_port_info_t));
/*
Should never write back a value that is bigger then 3 in
the PortPhysicalState field - so can not simply copy!
Actually we want to write there:
port physical state - no change,
link down default state = polling
port state - as requested.
*/
p_pi->state_info2 = 0x02;
ib_port_info_set_port_state(p_pi, port_state);
if (ib_port_info_get_link_down_def_state(p_pi) !=
ib_port_info_get_link_down_def_state(p_old_pi))
send_set = TRUE;
/* didn't get PortInfo before */
if (!ib_port_info_get_port_state(p_old_pi))
send_set = TRUE;
/* we only change port fields if we do not change state */
if (port_state == IB_LINK_NO_CHANGE) {
/* The following fields are relevant only for CA port, router, or Enh. SP0 */
if (osm_node_get_type(p_node) != IB_NODE_TYPE_SWITCH ||
port_num == 0) {
p_pi->m_key = sm->p_subn->opt.m_key;
if (memcmp(&p_pi->m_key, &p_old_pi->m_key,
sizeof(p_pi->m_key)))
send_set = TRUE;
p_pi->subnet_prefix = sm->p_subn->opt.subnet_prefix;
if (memcmp(&p_pi->subnet_prefix,
&p_old_pi->subnet_prefix,
sizeof(p_pi->subnet_prefix)))
send_set = TRUE;
p_pi->base_lid = osm_physp_get_base_lid(p_physp);
if (memcmp(&p_pi->base_lid, &p_old_pi->base_lid,
sizeof(p_pi->base_lid)))
send_set = TRUE;
/* we are initializing the ports with our local sm_base_lid */
p_pi->master_sm_base_lid = sm->p_subn->sm_base_lid;
if (memcmp(&p_pi->master_sm_base_lid,
&p_old_pi->master_sm_base_lid,
sizeof(p_pi->master_sm_base_lid)))
send_set = TRUE;
p_pi->m_key_lease_period =
sm->p_subn->opt.m_key_lease_period;
if (memcmp(&p_pi->m_key_lease_period,
&p_old_pi->m_key_lease_period,
sizeof(p_pi->m_key_lease_period)))
send_set = TRUE;
if (esp0 == FALSE)
p_pi->mkey_lmc = sm->p_subn->opt.lmc;
else {
if (sm->p_subn->opt.lmc_esp0)
p_pi->mkey_lmc = sm->p_subn->opt.lmc;
else
p_pi->mkey_lmc = 0;
}
if (memcmp(&p_pi->mkey_lmc, &p_old_pi->mkey_lmc,
sizeof(p_pi->mkey_lmc)))
send_set = TRUE;
ib_port_info_set_timeout(p_pi,
sm->p_subn->opt.
subnet_timeout);
if (ib_port_info_get_timeout(p_pi) !=
ib_port_info_get_timeout(p_old_pi))
send_set = TRUE;
}
/*
Several timeout mechanisms:
*/
p_remote_physp = osm_physp_get_remote(p_physp);
if (port_num != 0 && p_remote_physp) {
if (osm_node_get_type(osm_physp_get_node_ptr(p_physp))
== IB_NODE_TYPE_ROUTER) {
ib_port_info_set_hoq_lifetime(p_pi,
sm->p_subn->
opt.
leaf_head_of_queue_lifetime);
} else
if (osm_node_get_type
(osm_physp_get_node_ptr(p_physp)) ==
IB_NODE_TYPE_SWITCH) {
/* Is remote end CA or router (a leaf port) ? */
if (osm_node_get_type
(osm_physp_get_node_ptr(p_remote_physp)) !=
IB_NODE_TYPE_SWITCH) {
ib_port_info_set_hoq_lifetime(p_pi,
sm->
p_subn->
opt.
leaf_head_of_queue_lifetime);
ib_port_info_set_vl_stall_count(p_pi,
sm->
p_subn->
opt.
leaf_vl_stall_count);
} else {
ib_port_info_set_hoq_lifetime(p_pi,
sm->
p_subn->
opt.
head_of_queue_lifetime);
ib_port_info_set_vl_stall_count(p_pi,
sm->
p_subn->
opt.
vl_stall_count);
}
}
if (ib_port_info_get_hoq_lifetime(p_pi) !=
ib_port_info_get_hoq_lifetime(p_old_pi) ||
ib_port_info_get_vl_stall_count(p_pi) !=
ib_port_info_get_vl_stall_count(p_old_pi))
send_set = TRUE;
}
ib_port_info_set_phy_and_overrun_err_thd(p_pi,
sm->p_subn->opt.
local_phy_errors_threshold,
sm->p_subn->opt.
overrun_errors_threshold);
if (memcmp(&p_pi->error_threshold, &p_old_pi->error_threshold,
sizeof(p_pi->error_threshold)))
send_set = TRUE;
/*
Set the easy common parameters for all port types,
then determine the neighbor MTU.
*/
p_pi->link_width_enabled = p_old_pi->link_width_supported;
if (memcmp(&p_pi->link_width_enabled,
&p_old_pi->link_width_enabled,
sizeof(p_pi->link_width_enabled)))
send_set = TRUE;
if (sm->p_subn->opt.force_link_speed &&
(sm->p_subn->opt.force_link_speed != 15 ||
ib_port_info_get_link_speed_enabled(p_pi) !=
ib_port_info_get_link_speed_sup(p_pi))) {
ib_port_info_set_link_speed_enabled(p_pi,
sm->p_subn->opt.
force_link_speed);
if (memcmp(&p_pi->link_speed, &p_old_pi->link_speed,
sizeof(p_pi->link_speed)))
send_set = TRUE;
}
/* calc new op_vls and mtu */
op_vls =
osm_physp_calc_link_op_vls(sm->p_log, sm->p_subn, p_physp);
mtu = osm_physp_calc_link_mtu(sm->p_log, p_physp);
ib_port_info_set_neighbor_mtu(p_pi, mtu);
if (ib_port_info_get_neighbor_mtu(p_pi) !=
ib_port_info_get_neighbor_mtu(p_old_pi))
send_set = TRUE;
ib_port_info_set_op_vls(p_pi, op_vls);
if (ib_port_info_get_op_vls(p_pi) !=
ib_port_info_get_op_vls(p_old_pi))
send_set = TRUE;
/* provide the vl_high_limit from the qos mgr */
if (sm->p_subn->opt.qos &&
p_physp->vl_high_limit != p_old_pi->vl_high_limit) {
send_set = TRUE;
p_pi->vl_high_limit = p_physp->vl_high_limit;
}
}
if (port_state != IB_LINK_NO_CHANGE &&
port_state != ib_port_info_get_port_state(p_old_pi)) {
send_set = TRUE;
if (port_state == IB_LINK_ACTIVE)
context.pi_context.active_transition = TRUE;
else
context.pi_context.active_transition = FALSE;
}
context.pi_context.node_guid = osm_node_get_node_guid(p_node);
context.pi_context.port_guid = osm_physp_get_port_guid(p_physp);
context.pi_context.set_method = TRUE;
context.pi_context.light_sweep = FALSE;
/* We need to send the PortInfoSet request with the new sm_lid
in the following cases:
1. There is a change in the values (send_set == TRUE)
2. This is a switch external port (so it wasn't handled yet by
osm_lid_mgr) and first_time_master_sweep flag on the subnet is TRUE,
which means the SM just became master, and it then needs to send at
PortInfoSet to every port.
*/
if (osm_node_get_type(p_node) == IB_NODE_TYPE_SWITCH && port_num
&& sm->p_subn->first_time_master_sweep == TRUE)
send_set = TRUE;
if (send_set)
status = osm_req_set(sm, osm_physp_get_dr_path_ptr(p_physp),
payload, sizeof(payload),
IB_MAD_ATTR_PORT_INFO,
cl_hton32(port_num),
CL_DISP_MSGID_NONE, &context);
Exit:
OSM_LOG_EXIT(sm->p_log);
return send_set;
}
static int link_mgr_set_physp_pi(osm_sm_t * sm, IN osm_physp_t * p_physp,
IN uint8_t port_state)
{
uint8_t payload[IB_SMP_DATA_SIZE], payload2[IB_SMP_DATA_SIZE];
ib_port_info_t *p_pi = (ib_port_info_t *) payload;
ib_mlnx_ext_port_info_t *p_epi = (ib_mlnx_ext_port_info_t *) payload2;
const ib_port_info_t *p_old_pi;
const ib_mlnx_ext_port_info_t *p_old_epi;
osm_madw_context_t context;
osm_node_t *p_node;
ib_api_status_t status;
uint8_t port_num, mtu, op_vls, smsl = OSM_DEFAULT_SL;
boolean_t esp0 = FALSE, send_set = FALSE, send_set2 = FALSE;
osm_physp_t *p_remote_physp, *physp0;
int qdr_change = 0, fdr10_change = 0;
int ret = 0;
ib_net32_t attr_mod, cap_mask;
OSM_LOG_ENTER(sm->p_log);
p_node = osm_physp_get_node_ptr(p_physp);
p_old_pi = &p_physp->port_info;
port_num = osm_physp_get_port_num(p_physp);
if (port_num == 0) {
/*
CAs don't have a port 0, and for switch port 0,
we need to check if this is enhanced or base port 0.
For base port 0 the following parameters are not valid
(IBA 1.2.1 p.830 table 146).
*/
if (!p_node->sw) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 4201: "
"Cannot find switch by guid: 0x%" PRIx64 "\n",
cl_ntoh64(p_node->node_info.node_guid));
goto Exit;
}
if (ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info)
== FALSE) {
/* Even for base port 0 we might have to set smsl
(if we are using lash routing) */
smsl = link_mgr_get_smsl(sm, p_physp);
if (smsl != ib_port_info_get_master_smsl(p_old_pi)) {
send_set = TRUE;
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Setting SMSL to %d on port 0 GUID 0x%016"
PRIx64 "\n", smsl,
cl_ntoh64(osm_physp_get_port_guid
(p_physp)));
} else {
/* This means the switch doesn't support
enhanced port 0 and we don't need to
change SMSL. Can skip it. */
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Skipping port 0, GUID 0x%016" PRIx64
"\n",
cl_ntoh64(osm_physp_get_port_guid
(p_physp)));
goto Exit;
}
} else
esp0 = TRUE;
}
memcpy(payload, p_old_pi, sizeof(ib_port_info_t));
/*
Should never write back a value that is bigger then 3 in
the PortPhysicalState field - so can not simply copy!
Actually we want to write there:
port physical state - no change,
link down default state = polling
port state - as requested.
*/
p_pi->state_info2 = 0x02;
ib_port_info_set_port_state(p_pi, port_state);
/* Check whether this is base port0 smsl handling only */
if (port_num == 0 && esp0 == FALSE) {
ib_port_info_set_master_smsl(p_pi, smsl);
goto Send;
}
/*
PAST THIS POINT WE ARE HANDLING EITHER A NON PORT 0 OR ENHANCED PORT 0
*/
if (ib_port_info_get_link_down_def_state(p_pi) !=
ib_port_info_get_link_down_def_state(p_old_pi))
send_set = TRUE;
/* didn't get PortInfo before */
if (!ib_port_info_get_port_state(p_old_pi))
send_set = TRUE;
/* we only change port fields if we do not change state */
if (port_state == IB_LINK_NO_CHANGE) {
/* The following fields are relevant only for CA port, router, or Enh. SP0 */
if (osm_node_get_type(p_node) != IB_NODE_TYPE_SWITCH ||
port_num == 0) {
p_pi->m_key = sm->p_subn->opt.m_key;
if (memcmp(&p_pi->m_key, &p_old_pi->m_key,
sizeof(p_pi->m_key)))
send_set = TRUE;
p_pi->subnet_prefix = sm->p_subn->opt.subnet_prefix;
if (memcmp(&p_pi->subnet_prefix,
&p_old_pi->subnet_prefix,
sizeof(p_pi->subnet_prefix)))
send_set = TRUE;
p_pi->base_lid = osm_physp_get_base_lid(p_physp);
if (memcmp(&p_pi->base_lid, &p_old_pi->base_lid,
sizeof(p_pi->base_lid)))
send_set = TRUE;
/* we are initializing the ports with our local sm_base_lid */
p_pi->master_sm_base_lid = sm->p_subn->sm_base_lid;
if (memcmp(&p_pi->master_sm_base_lid,
&p_old_pi->master_sm_base_lid,
sizeof(p_pi->master_sm_base_lid)))
send_set = TRUE;
smsl = link_mgr_get_smsl(sm, p_physp);
if (smsl != ib_port_info_get_master_smsl(p_old_pi)) {
ib_port_info_set_master_smsl(p_pi, smsl);
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Setting SMSL to %d on GUID 0x%016"
PRIx64 ", port %d\n", smsl,
cl_ntoh64(osm_physp_get_port_guid
(p_physp)), port_num);
send_set = TRUE;
}
p_pi->m_key_lease_period =
sm->p_subn->opt.m_key_lease_period;
if (memcmp(&p_pi->m_key_lease_period,
&p_old_pi->m_key_lease_period,
sizeof(p_pi->m_key_lease_period)))
send_set = TRUE;
/* M_KeyProtectBits are currently always zero */
p_pi->mkey_lmc = 0;
if (esp0 == FALSE || sm->p_subn->opt.lmc_esp0)
ib_port_info_set_lmc(p_pi, sm->p_subn->opt.lmc);
if (ib_port_info_get_lmc(p_old_pi) !=
ib_port_info_get_lmc(p_pi) ||
ib_port_info_get_mpb(p_old_pi) !=
ib_port_info_get_mpb(p_pi))
send_set = TRUE;
ib_port_info_set_timeout(p_pi,
sm->p_subn->opt.
subnet_timeout);
if (ib_port_info_get_timeout(p_pi) !=
ib_port_info_get_timeout(p_old_pi))
send_set = TRUE;
}
/*
Several timeout mechanisms:
*/
p_remote_physp = osm_physp_get_remote(p_physp);
if (port_num != 0 && p_remote_physp) {
if (osm_node_get_type(osm_physp_get_node_ptr(p_physp))
== IB_NODE_TYPE_ROUTER) {
ib_port_info_set_hoq_lifetime(p_pi,
sm->p_subn->
opt.
leaf_head_of_queue_lifetime);
} else
if (osm_node_get_type
(osm_physp_get_node_ptr(p_physp)) ==
IB_NODE_TYPE_SWITCH) {
/* Is remote end CA or router (a leaf port) ? */
if (osm_node_get_type
(osm_physp_get_node_ptr(p_remote_physp)) !=
IB_NODE_TYPE_SWITCH) {
ib_port_info_set_hoq_lifetime(p_pi,
sm->
p_subn->
opt.
leaf_head_of_queue_lifetime);
ib_port_info_set_vl_stall_count(p_pi,
sm->
p_subn->
opt.
leaf_vl_stall_count);
} else {
ib_port_info_set_hoq_lifetime(p_pi,
sm->
p_subn->
opt.
head_of_queue_lifetime);
ib_port_info_set_vl_stall_count(p_pi,
sm->
p_subn->
opt.
vl_stall_count);
}
}
if (ib_port_info_get_hoq_lifetime(p_pi) !=
ib_port_info_get_hoq_lifetime(p_old_pi) ||
ib_port_info_get_vl_stall_count(p_pi) !=
ib_port_info_get_vl_stall_count(p_old_pi))
send_set = TRUE;
}
ib_port_info_set_phy_and_overrun_err_thd(p_pi,
sm->p_subn->opt.
local_phy_errors_threshold,
sm->p_subn->opt.
overrun_errors_threshold);
if (memcmp(&p_pi->error_threshold, &p_old_pi->error_threshold,
sizeof(p_pi->error_threshold)))
send_set = TRUE;
/*
Set the easy common parameters for all port types,
then determine the neighbor MTU.
*/
p_pi->link_width_enabled = p_old_pi->link_width_supported;
if (memcmp(&p_pi->link_width_enabled,
&p_old_pi->link_width_enabled,
sizeof(p_pi->link_width_enabled)))
send_set = TRUE;
if (sm->p_subn->opt.force_link_speed &&
(sm->p_subn->opt.force_link_speed != 15 ||
ib_port_info_get_link_speed_enabled(p_pi) !=
ib_port_info_get_link_speed_sup(p_pi))) {
ib_port_info_set_link_speed_enabled(p_pi,
sm->p_subn->opt.
force_link_speed);
if (memcmp(&p_pi->link_speed, &p_old_pi->link_speed,
sizeof(p_pi->link_speed))) {
send_set = TRUE;
/* Determine whether QDR in LSE is being changed */
if ((ib_port_info_get_link_speed_enabled(p_pi) &
IB_LINK_SPEED_ACTIVE_10 &&
!(ib_port_info_get_link_speed_enabled(p_old_pi) &
IB_LINK_SPEED_ACTIVE_10)) ||
((!(ib_port_info_get_link_speed_enabled(p_pi) &
IB_LINK_SPEED_ACTIVE_10) &&
ib_port_info_get_link_speed_enabled(p_old_pi) &
IB_LINK_SPEED_ACTIVE_10)))
qdr_change = 1;
}
}
if (sm->p_subn->opt.fdr10 &&
p_physp->ext_port_info.link_speed_supported & FDR10) {
if (sm->p_subn->opt.fdr10 == 1) { /* enable */
if (!(p_physp->ext_port_info.link_speed_enabled & FDR10))
fdr10_change = 1;
} else { /* disable */
if (p_physp->ext_port_info.link_speed_enabled & FDR10)
fdr10_change = 1;
}
if (fdr10_change) {
p_old_epi = &p_physp->ext_port_info;
memcpy(payload2, p_old_epi,
sizeof(ib_mlnx_ext_port_info_t));
p_epi->state_change_enable = 0x01;
if (sm->p_subn->opt.fdr10 == 1)
p_epi->link_speed_enabled = FDR10;
else
p_epi->link_speed_enabled = 0;
send_set2 = TRUE;
}
}
if (osm_node_get_type(p_physp->p_node) == IB_NODE_TYPE_SWITCH) {
physp0 = osm_node_get_physp_ptr(p_physp->p_node, 0);
cap_mask = physp0->port_info.capability_mask;
} else
cap_mask = p_pi->capability_mask;
if (!(cap_mask & IB_PORT_CAP_HAS_EXT_SPEEDS))
qdr_change = 0;
/* Do peer ports support extended link speeds ? */
if (port_num != 0 && p_remote_physp) {
osm_physp_t *rphysp0;
ib_net32_t rem_cap_mask;
if (osm_node_get_type(p_remote_physp->p_node) ==
IB_NODE_TYPE_SWITCH) {
rphysp0 = osm_node_get_physp_ptr(p_remote_physp->p_node, 0);
rem_cap_mask = rphysp0->port_info.capability_mask;
} else
rem_cap_mask = p_remote_physp->port_info.capability_mask;
if (cap_mask & IB_PORT_CAP_HAS_EXT_SPEEDS &&
rem_cap_mask & IB_PORT_CAP_HAS_EXT_SPEEDS) {
if (sm->p_subn->opt.force_link_speed_ext &&
(sm->p_subn->opt.force_link_speed_ext != IB_LINK_SPEED_EXT_SET_LSES ||
p_pi->link_speed_ext_enabled !=
ib_port_info_get_link_speed_sup(p_pi))) {
p_pi->link_speed_ext_enabled = sm->p_subn->opt.force_link_speed_ext;
if (memcmp(&p_pi->link_speed_ext_enabled,
&p_old_pi->link_speed_ext_enabled,
sizeof(p_pi->link_speed_ext_enabled)))
send_set = TRUE;
}
}
}
/* calc new op_vls and mtu */
op_vls =
osm_physp_calc_link_op_vls(sm->p_log, sm->p_subn, p_physp);
mtu = osm_physp_calc_link_mtu(sm->p_log, p_physp);
ib_port_info_set_neighbor_mtu(p_pi, mtu);
if (ib_port_info_get_neighbor_mtu(p_pi) !=
ib_port_info_get_neighbor_mtu(p_old_pi))
send_set = TRUE;
ib_port_info_set_op_vls(p_pi, op_vls);
if (ib_port_info_get_op_vls(p_pi) !=
ib_port_info_get_op_vls(p_old_pi))
send_set = TRUE;
/* provide the vl_high_limit from the qos mgr */
if (sm->p_subn->opt.qos &&
p_physp->vl_high_limit != p_old_pi->vl_high_limit) {
send_set = TRUE;
p_pi->vl_high_limit = p_physp->vl_high_limit;
}
}
Send:
if (port_state != IB_LINK_NO_CHANGE &&
port_state != ib_port_info_get_port_state(p_old_pi)) {
send_set = TRUE;
if (port_state == IB_LINK_ACTIVE)
context.pi_context.active_transition = TRUE;
else
context.pi_context.active_transition = FALSE;
}
context.pi_context.node_guid = osm_node_get_node_guid(p_node);
context.pi_context.port_guid = osm_physp_get_port_guid(p_physp);
context.pi_context.set_method = TRUE;
context.pi_context.light_sweep = FALSE;
/* We need to send the PortInfoSet request with the new sm_lid
in the following cases:
1. There is a change in the values (send_set == TRUE)
2. This is a switch external port (so it wasn't handled yet by
osm_lid_mgr) and first_time_master_sweep flag on the subnet is TRUE,
which means the SM just became master, and it then needs to send at
PortInfoSet to every port.
*/
if (osm_node_get_type(p_node) == IB_NODE_TYPE_SWITCH && port_num
&& sm->p_subn->first_time_master_sweep == TRUE)
send_set = TRUE;
if (!send_set)
goto Exit;
attr_mod = cl_hton32(port_num);
if (qdr_change)
attr_mod |= cl_hton32(1 << 31); /* AM SMSupportExtendedSpeeds */
status = osm_req_set(sm, osm_physp_get_dr_path_ptr(p_physp),
payload, sizeof(payload), IB_MAD_ATTR_PORT_INFO,
attr_mod, CL_DISP_MSGID_NONE, &context);
if (status)
ret = -1;
if (send_set2) {
status = osm_req_set(sm, osm_physp_get_dr_path_ptr(p_physp),
payload2, sizeof(payload2),
IB_MAD_ATTR_MLNX_EXTENDED_PORT_INFO,
cl_hton32(port_num),
CL_DISP_MSGID_NONE, &context);
if (status)
ret = -1;
}
Exit:
OSM_LOG_EXIT(sm->p_log);
return ret;
}
void osm_slvl_rec_rcv_process(IN void *ctx, IN void *data)
{
osm_sa_t *sa = ctx;
osm_madw_t *p_madw = data;
const ib_sa_mad_t *p_rcvd_mad;
const ib_slvl_table_record_t *p_rcvd_rec;
const osm_port_t *p_port = NULL;
cl_qlist_t rec_list;
osm_slvl_search_ctxt_t context;
ib_api_status_t status = IB_SUCCESS;
ib_net64_t comp_mask;
osm_physp_t *p_req_physp;
CL_ASSERT(sa);
OSM_LOG_ENTER(sa->p_log);
CL_ASSERT(p_madw);
p_rcvd_mad = osm_madw_get_sa_mad_ptr(p_madw);
p_rcvd_rec =
(ib_slvl_table_record_t *) ib_sa_mad_get_payload_ptr(p_rcvd_mad);
comp_mask = p_rcvd_mad->comp_mask;
CL_ASSERT(p_rcvd_mad->attr_id == IB_MAD_ATTR_SLVL_RECORD);
/* we only support SubnAdmGet and SubnAdmGetTable methods */
if (p_rcvd_mad->method != IB_MAD_METHOD_GET &&
p_rcvd_mad->method != IB_MAD_METHOD_GETTABLE) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2604: "
"Unsupported Method (%s)\n",
ib_get_sa_method_str(p_rcvd_mad->method));
osm_sa_send_error(sa, p_madw, IB_MAD_STATUS_UNSUP_METHOD_ATTR);
goto Exit;
}
/* update the requester physical port */
p_req_physp = osm_get_physp_by_mad_addr(sa->p_log, sa->p_subn,
osm_madw_get_mad_addr_ptr
(p_madw));
if (p_req_physp == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2603: "
"Cannot find requester physical port\n");
goto Exit;
}
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Requester port GUID 0x%" PRIx64 "\n",
cl_ntoh64(osm_physp_get_port_guid(p_req_physp)));
cl_qlist_init(&rec_list);
context.p_rcvd_rec = p_rcvd_rec;
context.p_list = &rec_list;
context.comp_mask = p_rcvd_mad->comp_mask;
context.sa = sa;
context.in_port_num = p_rcvd_rec->in_port_num;
context.p_req_physp = p_req_physp;
cl_plock_acquire(sa->p_lock);
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Got Query Lid:%u(%02X), In-Port:0x%02X(%02X), Out-Port:0x%02X(%02X)\n",
cl_ntoh16(p_rcvd_rec->lid),
(comp_mask & IB_SLVL_COMPMASK_LID) != 0,
p_rcvd_rec->in_port_num,
(comp_mask & IB_SLVL_COMPMASK_IN_PORT) != 0,
p_rcvd_rec->out_port_num,
(comp_mask & IB_SLVL_COMPMASK_OUT_PORT) != 0);
/*
If the user specified a LID, it obviously narrows our
work load, since we don't have to search every port
*/
if (comp_mask & IB_SLVL_COMPMASK_LID) {
p_port = osm_get_port_by_lid(sa->p_subn, p_rcvd_rec->lid);
if (!p_port) {
status = IB_NOT_FOUND;
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2608: "
"No port found with LID %u\n",
cl_ntoh16(p_rcvd_rec->lid));
}
}
if (status == IB_SUCCESS) {
/* if we have a unique port - no need for a port search */
if (p_port)
/* this does the loop on all the port phys ports */
sa_slvl_by_comp_mask(sa, p_port, &context);
else
cl_qmap_apply_func(&sa->p_subn->port_guid_tbl,
sa_slvl_by_comp_mask_cb, &context);
}
cl_plock_release(sa->p_lock);
osm_sa_respond(sa, p_madw, sizeof(ib_slvl_table_record_t), &rec_list);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
static void lr_rcv_get_physp_link(IN osm_sa_t * sa,
IN const ib_link_record_t * p_lr,
IN const osm_physp_t * p_src_physp,
IN const osm_physp_t * p_dest_physp,
IN const ib_net64_t comp_mask,
IN cl_qlist_t * p_list,
IN const osm_physp_t * p_req_physp)
{
uint8_t src_port_num;
uint8_t dest_port_num;
ib_net16_t from_base_lid;
ib_net16_t to_base_lid;
ib_net16_t lmc_mask;
OSM_LOG_ENTER(sa->p_log);
/*
If only one end of the link is specified, determine
the other side.
*/
if (p_src_physp) {
if (p_dest_physp) {
/*
Ensure the two physp's are actually connected.
If not, bail out.
*/
if (osm_physp_get_remote(p_src_physp) != p_dest_physp)
goto Exit;
} else {
p_dest_physp = osm_physp_get_remote(p_src_physp);
if (p_dest_physp == NULL)
goto Exit;
}
} else {
if (p_dest_physp) {
p_src_physp = osm_physp_get_remote(p_dest_physp);
if (p_src_physp == NULL)
goto Exit;
} else
goto Exit; /* no physp's, so nothing to do */
}
/* Check that the p_src_physp, p_dest_physp and p_req_physp
all share a pkey (doesn't have to be the same p_key). */
if (!osm_physp_share_pkey(sa->p_log, p_src_physp, p_dest_physp,
sa->p_subn->opt.allow_both_pkeys)) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Source and Dest PhysPorts do not share PKey\n");
goto Exit;
}
if (!osm_physp_share_pkey(sa->p_log, p_src_physp, p_req_physp,
sa->p_subn->opt.allow_both_pkeys)) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Source and Requester PhysPorts do not share PKey\n");
goto Exit;
}
if (!osm_physp_share_pkey(sa->p_log, p_req_physp, p_dest_physp,
sa->p_subn->opt.allow_both_pkeys)) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Requester and Dest PhysPorts do not share PKey\n");
goto Exit;
}
src_port_num = osm_physp_get_port_num(p_src_physp);
dest_port_num = osm_physp_get_port_num(p_dest_physp);
if (comp_mask & IB_LR_COMPMASK_FROM_PORT)
if (src_port_num != p_lr->from_port_num)
goto Exit;
if (comp_mask & IB_LR_COMPMASK_TO_PORT)
if (dest_port_num != p_lr->to_port_num)
goto Exit;
from_base_lid = get_base_lid(p_src_physp);
to_base_lid = get_base_lid(p_dest_physp);
lmc_mask = ~((1 << sa->p_subn->opt.lmc) - 1);
lmc_mask = cl_hton16(lmc_mask);
if (comp_mask & IB_LR_COMPMASK_FROM_LID)
if (from_base_lid != (p_lr->from_lid & lmc_mask))
goto Exit;
if (comp_mask & IB_LR_COMPMASK_TO_LID)
if (to_base_lid != (p_lr->to_lid & lmc_mask))
goto Exit;
OSM_LOG(sa->p_log, OSM_LOG_DEBUG, "Acquiring link record\n"
"\t\t\t\tsrc port 0x%" PRIx64 " (port %u)"
", dest port 0x%" PRIx64 " (port %u)\n",
cl_ntoh64(osm_physp_get_port_guid(p_src_physp)), src_port_num,
cl_ntoh64(osm_physp_get_port_guid(p_dest_physp)),
dest_port_num);
lr_rcv_build_physp_link(sa, from_base_lid, to_base_lid, src_port_num,
dest_port_num, p_list);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
static void get_pkey_table(IN osm_log_t * p_log, IN osm_sm_t * sm,
IN osm_node_t * p_node, IN osm_physp_t * p_physp)
{
osm_madw_context_t context;
ib_api_status_t status;
osm_dr_path_t path;
osm_physp_t *physp0;
ib_net64_t m_key;
uint8_t port_num;
uint16_t block_num, max_blocks;
uint32_t attr_mod_ho;
OSM_LOG_ENTER(p_log);
path = *osm_physp_get_dr_path_ptr(p_physp);
context.pkey_context.node_guid = osm_node_get_node_guid(p_node);
context.pkey_context.port_guid = osm_physp_get_port_guid(p_physp);
context.pkey_context.set_method = FALSE;
port_num = p_physp->port_num;
if (!p_node->sw || port_num == 0)
/* The maximum blocks is defined by the node info partition cap
for CA, router, and switch management ports. */
max_blocks =
(cl_ntoh16(p_node->node_info.partition_cap) +
IB_NUM_PKEY_ELEMENTS_IN_BLOCK - 1)
/ IB_NUM_PKEY_ELEMENTS_IN_BLOCK;
else {
/* This is a switch, and not a management port. The maximum blocks
is defined in the switch info partition enforcement cap. */
/* Check for IBM eHCA firmware defect in reporting partition enforcement cap */
if (cl_ntoh32(ib_node_info_get_vendor_id(&p_node->node_info)) ==
IBM_VENDOR_ID)
p_node->sw->switch_info.enforce_cap = 0;
/* Bail out if this is a switch with no partition enforcement capability */
if (cl_ntoh16(p_node->sw->switch_info.enforce_cap) == 0)
goto Exit;
max_blocks = (cl_ntoh16(p_node->sw->switch_info.enforce_cap) +
IB_NUM_PKEY_ELEMENTS_IN_BLOCK -
1) / IB_NUM_PKEY_ELEMENTS_IN_BLOCK;
}
p_physp->pkeys.rcv_blocks_cnt = max_blocks;
for (block_num = 0; block_num < max_blocks; block_num++) {
if (osm_node_get_type(p_node) != IB_NODE_TYPE_SWITCH ||
osm_physp_get_port_num(p_physp) == 0) {
attr_mod_ho = block_num;
m_key = ib_port_info_get_m_key(&p_physp->port_info);
} else {
attr_mod_ho = block_num | (port_num << 16);
physp0 = osm_node_get_physp_ptr(p_node, 0);
m_key = ib_port_info_get_m_key(&physp0->port_info);
}
status = osm_req_get(sm, &path, IB_MAD_ATTR_P_KEY_TABLE,
cl_hton32(attr_mod_ho), FALSE,
m_key, CL_DISP_MSGID_NONE, &context);
if (status != IB_SUCCESS) {
OSM_LOG(p_log, OSM_LOG_ERROR, "ERR 0F12: "
"Failure initiating PKeyTable request (%s)\n",
ib_get_err_str(status));
goto Exit;
}
}
Exit:
OSM_LOG_EXIT(p_log);
}
static ib_api_status_t mpr_rcv_get_path_parms(IN osm_sa_t * sa,
IN const ib_multipath_rec_t *
p_mpr,
IN const osm_alias_guid_t * p_src_alias_guid,
IN const osm_alias_guid_t * p_dest_alias_guid,
IN const uint16_t dest_lid_ho,
IN const ib_net64_t comp_mask,
OUT osm_path_parms_t * p_parms)
{
const osm_node_t *p_node;
const osm_physp_t *p_physp, *p_physp0;
const osm_physp_t *p_src_physp;
const osm_physp_t *p_dest_physp;
const osm_prtn_t *p_prtn = NULL;
const ib_port_info_t *p_pi, *p_pi0;
ib_slvl_table_t *p_slvl_tbl;
ib_api_status_t status = IB_SUCCESS;
uint8_t mtu;
uint8_t rate;
uint8_t pkt_life;
uint8_t required_mtu;
uint8_t required_rate;
ib_net16_t required_pkey;
uint8_t required_sl;
uint8_t required_pkt_life;
ib_net16_t dest_lid;
int hops = 0;
int in_port_num = 0;
uint8_t i;
osm_qos_level_t *p_qos_level = NULL;
uint16_t valid_sl_mask = 0xffff;
OSM_LOG_ENTER(sa->p_log);
dest_lid = cl_hton16(dest_lid_ho);
p_dest_physp = p_dest_alias_guid->p_base_port->p_physp;
p_physp = p_src_alias_guid->p_base_port->p_physp;
p_src_physp = p_physp;
p_pi = &p_physp->port_info;
mtu = ib_port_info_get_mtu_cap(p_pi);
rate = ib_port_info_compute_rate(p_pi,
p_pi->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS);
/*
Mellanox Tavor device performance is better using 1K MTU.
If required MTU and MTU selector are such that 1K is OK
and at least one end of the path is Tavor we override the
port MTU with 1K.
*/
if (sa->p_subn->opt.enable_quirks &&
sa_multipath_rec_apply_tavor_mtu_limit(p_mpr,
p_src_alias_guid->p_base_port,
p_dest_alias_guid->p_base_port,
comp_mask))
if (mtu > IB_MTU_LEN_1024) {
mtu = IB_MTU_LEN_1024;
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Optimized Path MTU to 1K for Mellanox Tavor device\n");
}
/*
Walk the subnet object from source to destination,
tracking the most restrictive rate and mtu values along the way...
If source port node is a switch, then p_physp should
point to the port that routes the destination lid
*/
p_node = osm_physp_get_node_ptr(p_physp);
if (p_node->sw) {
/*
* Source node is a switch.
* Make sure that p_physp points to the out port of the
* switch that routes to the destination lid (dest_lid_ho)
*/
p_physp = osm_switch_get_route_by_lid(p_node->sw, dest_lid);
if (p_physp == 0) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4514: "
"Can't find routing to LID %u on switch %s "
"(GUID 0x%016"PRIx64")\n", dest_lid_ho,
p_node->print_desc,
cl_ntoh64(osm_node_get_node_guid(p_node)));
status = IB_NOT_FOUND;
goto Exit;
}
}
if (sa->p_subn->opt.qos) {
/*
* Whether this node is switch or CA, the IN port for
* the sl2vl table is 0, because this is a source node.
*/
p_slvl_tbl = osm_physp_get_slvl_tbl(p_physp, 0);
/* update valid SLs that still exist on this route */
for (i = 0; i < IB_MAX_NUM_VLS; i++) {
if (valid_sl_mask & (1 << i) &&
ib_slvl_table_get(p_slvl_tbl, i) == IB_DROP_VL)
valid_sl_mask &= ~(1 << i);
}
if (!valid_sl_mask) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"All the SLs lead to VL15 on this path\n");
status = IB_NOT_FOUND;
goto Exit;
}
}
/*
* Same as above
*/
p_node = osm_physp_get_node_ptr(p_dest_physp);
if (p_node->sw) {
/*
* if destination is switch, we want p_dest_physp to point to port 0
*/
p_dest_physp =
osm_switch_get_route_by_lid(p_node->sw, dest_lid);
if (p_dest_physp == 0) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4515: "
"Can't find routing to LID %u on switch %s "
"(GUID 0x%016"PRIx64")\n", dest_lid_ho,
p_node->print_desc,
cl_ntoh64(osm_node_get_node_guid(p_node)));
status = IB_NOT_FOUND;
goto Exit;
}
}
/*
* Now go through the path step by step
*/
while (p_physp != p_dest_physp) {
int tmp_pnum = p_physp->port_num;
p_node = osm_physp_get_node_ptr(p_physp);
p_physp = osm_physp_get_remote(p_physp);
if (p_physp == 0) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4505: "
"Can't find remote phys port of %s (GUID "
"0x%016"PRIx64") port %d "
"while routing to LID %u",
p_node->print_desc,
cl_ntoh64(osm_node_get_node_guid(p_node)),
tmp_pnum,
dest_lid_ho);
status = IB_ERROR;
goto Exit;
}
/* update number of hops traversed */
hops++;
if (hops > MAX_HOPS) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4520: "
"Path from GUID 0x%016" PRIx64 " (%s) to"
" lid %u GUID 0x%016" PRIx64 " (%s) needs"
" more than %d hops, max %d hops allowed\n",
cl_ntoh64(osm_physp_get_port_guid(p_src_physp)),
p_src_physp->p_node->print_desc, dest_lid_ho,
cl_ntoh64(osm_physp_get_port_guid
(p_dest_physp)),
p_dest_physp->p_node->print_desc, hops,
MAX_HOPS);
status = IB_NOT_FOUND;
goto Exit;
}
in_port_num = osm_physp_get_port_num(p_physp);
/*
This is point to point case (no switch in between)
*/
if (p_physp == p_dest_physp)
break;
p_node = osm_physp_get_node_ptr(p_physp);
if (!p_node->sw) {
/*
There is some sort of problem in the subnet object!
If this isn't a switch, we should have reached
the destination by now!
*/
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4503: "
"Internal error, bad path while routing "
"from %s (GUID: 0x%016"PRIx64") port %d "
"to %s (GUID: 0x%016"PRIx64") port %d; "
"ended at %s port %d\n",
p_src_alias_guid->p_base_port->p_node->print_desc,
cl_ntoh64(p_src_alias_guid->p_base_port->p_node->node_info.node_guid),
p_src_alias_guid->p_base_port->p_physp->port_num,
p_dest_alias_guid->p_base_port->p_node->print_desc,
cl_ntoh64(p_dest_alias_guid->p_base_port->p_node->node_info.node_guid),
p_dest_alias_guid->p_base_port->p_physp->port_num,
p_node->print_desc,
p_physp->port_num);
status = IB_ERROR;
goto Exit;
}
/*
Check parameters for the ingress port in this switch.
*/
p_pi = &p_physp->port_info;
if (mtu > ib_port_info_get_mtu_cap(p_pi))
mtu = ib_port_info_get_mtu_cap(p_pi);
p_physp0 = osm_node_get_physp_ptr((osm_node_t *)p_node, 0);
p_pi0 = &p_physp0->port_info;
if (ib_path_compare_rates(rate,
ib_port_info_compute_rate(p_pi,
p_pi0->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS)) > 0)
rate = ib_port_info_compute_rate(p_pi,
p_pi0->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS);
/*
Continue with the egress port on this switch.
*/
p_physp = osm_switch_get_route_by_lid(p_node->sw, dest_lid);
if (p_physp == 0) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4516: "
"Dead end path on switch "
"%s (GUID: 0x%016"PRIx64") to LID %u\n",
p_node->print_desc,
cl_ntoh64(osm_node_get_node_guid(p_node)),
dest_lid_ho);
status = IB_ERROR;
goto Exit;
}
p_pi = &p_physp->port_info;
if (mtu > ib_port_info_get_mtu_cap(p_pi))
mtu = ib_port_info_get_mtu_cap(p_pi);
p_physp0 = osm_node_get_physp_ptr((osm_node_t *)p_node, 0);
p_pi0 = &p_physp0->port_info;
if (ib_path_compare_rates(rate,
ib_port_info_compute_rate(p_pi,
p_pi0->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS)) > 0)
rate = ib_port_info_compute_rate(p_pi,
p_pi0->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS);
if (sa->p_subn->opt.qos) {
/*
* Check SL2VL table of the switch and update valid SLs
*/
p_slvl_tbl =
osm_physp_get_slvl_tbl(p_physp, in_port_num);
for (i = 0; i < IB_MAX_NUM_VLS; i++) {
if (valid_sl_mask & (1 << i) &&
ib_slvl_table_get(p_slvl_tbl,
i) == IB_DROP_VL)
valid_sl_mask &= ~(1 << i);
}
if (!valid_sl_mask) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"All the SLs lead to VL15 "
"on this path\n");
status = IB_NOT_FOUND;
goto Exit;
}
}
}
/*
p_physp now points to the destination
*/
p_pi = &p_physp->port_info;
if (mtu > ib_port_info_get_mtu_cap(p_pi))
mtu = ib_port_info_get_mtu_cap(p_pi);
if (ib_path_compare_rates(rate,
ib_port_info_compute_rate(p_pi,
p_pi->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS)) > 0)
rate = ib_port_info_compute_rate(p_pi,
p_pi->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS);
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Path min MTU = %u, min rate = %u\n", mtu, rate);
/*
* Get QoS Level object according to the MultiPath request
* and adjust MultiPath parameters according to QoS settings
*/
if (sa->p_subn->opt.qos && sa->p_subn->p_qos_policy &&
(p_qos_level =
osm_qos_policy_get_qos_level_by_mpr(sa->p_subn->p_qos_policy,
p_mpr, p_src_physp,
p_dest_physp, comp_mask))) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"MultiPathRecord request matches QoS Level '%s' (%s)\n",
p_qos_level->name,
p_qos_level->use ? p_qos_level->use : "no description");
if (p_qos_level->mtu_limit_set
&& (mtu > p_qos_level->mtu_limit))
mtu = p_qos_level->mtu_limit;
if (p_qos_level->rate_limit_set
&& (ib_path_compare_rates(rate, p_qos_level->rate_limit) > 0))
rate = p_qos_level->rate_limit;
if (p_qos_level->sl_set) {
required_sl = p_qos_level->sl;
if (!(valid_sl_mask & (1 << required_sl))) {
status = IB_NOT_FOUND;
goto Exit;
}
}
}
/*
Determine if these values meet the user criteria
*/
/* we silently ignore cases where only the MTU selector is defined */
if ((comp_mask & IB_MPR_COMPMASK_MTUSELEC) &&
(comp_mask & IB_MPR_COMPMASK_MTU)) {
required_mtu = ib_multipath_rec_mtu(p_mpr);
switch (ib_multipath_rec_mtu_sel(p_mpr)) {
case 0: /* must be greater than */
if (mtu <= required_mtu)
status = IB_NOT_FOUND;
break;
case 1: /* must be less than */
if (mtu >= required_mtu) {
/* adjust to use the highest mtu
lower then the required one */
if (required_mtu > 1)
mtu = required_mtu - 1;
else
status = IB_NOT_FOUND;
}
break;
case 2: /* exact match */
if (mtu < required_mtu)
status = IB_NOT_FOUND;
else
mtu = required_mtu;
break;
case 3: /* largest available */
/* can't be disqualified by this one */
break;
default:
/* if we're here, there's a bug in ib_multipath_rec_mtu_sel() */
CL_ASSERT(FALSE);
status = IB_ERROR;
break;
}
}
if (status != IB_SUCCESS)
goto Exit;
/* we silently ignore cases where only the Rate selector is defined */
if ((comp_mask & IB_MPR_COMPMASK_RATESELEC) &&
(comp_mask & IB_MPR_COMPMASK_RATE)) {
required_rate = ib_multipath_rec_rate(p_mpr);
switch (ib_multipath_rec_rate_sel(p_mpr)) {
case 0: /* must be greater than */
if (ib_path_compare_rates(rate, required_rate) <= 0)
status = IB_NOT_FOUND;
break;
case 1: /* must be less than */
if (ib_path_compare_rates(rate, required_rate) >= 0) {
/* adjust the rate to use the highest rate
lower then the required one */
rate = ib_path_rate_get_prev(required_rate);
if (!rate)
status = IB_NOT_FOUND;
}
break;
case 2: /* exact match */
if (ib_path_compare_rates(rate, required_rate))
status = IB_NOT_FOUND;
else
rate = required_rate;
break;
case 3: /* largest available */
/* can't be disqualified by this one */
break;
default:
/* if we're here, there's a bug in ib_multipath_rec_mtu_sel() */
CL_ASSERT(FALSE);
status = IB_ERROR;
break;
}
}
if (status != IB_SUCCESS)
goto Exit;
/* Verify the pkt_life_time */
/* According to spec definition IBA 1.2 Table 205 PacketLifeTime description,
for loopback paths, packetLifeTime shall be zero. */
if (p_src_alias_guid->p_base_port == p_dest_alias_guid->p_base_port)
pkt_life = 0; /* loopback */
else if (p_qos_level && p_qos_level->pkt_life_set)
pkt_life = p_qos_level->pkt_life;
else
pkt_life = sa->p_subn->opt.subnet_timeout;
/* we silently ignore cases where only the PktLife selector is defined */
if ((comp_mask & IB_MPR_COMPMASK_PKTLIFETIMESELEC) &&
(comp_mask & IB_MPR_COMPMASK_PKTLIFETIME)) {
required_pkt_life = ib_multipath_rec_pkt_life(p_mpr);
switch (ib_multipath_rec_pkt_life_sel(p_mpr)) {
case 0: /* must be greater than */
if (pkt_life <= required_pkt_life)
status = IB_NOT_FOUND;
break;
case 1: /* must be less than */
if (pkt_life >= required_pkt_life) {
/* adjust the lifetime to use the highest possible
lower then the required one */
if (required_pkt_life > 1)
pkt_life = required_pkt_life - 1;
else
status = IB_NOT_FOUND;
}
break;
case 2: /* exact match */
if (pkt_life < required_pkt_life)
status = IB_NOT_FOUND;
else
pkt_life = required_pkt_life;
break;
case 3: /* smallest available */
/* can't be disqualified by this one */
break;
default:
/* if we're here, there's a bug in ib_path_rec_pkt_life_sel() */
CL_ASSERT(FALSE);
status = IB_ERROR;
break;
}
}
if (status != IB_SUCCESS)
goto Exit;
/*
* set Pkey for this MultiPath record request
*/
if (comp_mask & IB_MPR_COMPMASK_RAWTRAFFIC &&
cl_ntoh32(p_mpr->hop_flow_raw) & (1 << 31))
required_pkey =
osm_physp_find_common_pkey(p_src_physp, p_dest_physp,
sa->p_subn->opt.allow_both_pkeys);
else if (comp_mask & IB_MPR_COMPMASK_PKEY) {
/*
* MPR request has a specific pkey:
* Check that source and destination share this pkey.
* If QoS level has pkeys, check that this pkey exists
* in the QoS level pkeys.
* MPR returned pkey is the requested pkey.
*/
required_pkey = p_mpr->pkey;
if (!osm_physp_share_this_pkey
(p_src_physp, p_dest_physp, required_pkey,
sa->p_subn->opt.allow_both_pkeys)) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4518: "
"Ports src 0x%016"PRIx64" (%s port %d) "
"and dst 0x%016"PRIx64" (%s port %d) "
"do not share the specified PKey 0x%04x\n",
cl_ntoh64(osm_physp_get_port_guid(p_src_physp)),
p_src_physp->p_node->print_desc,
p_src_physp->port_num,
cl_ntoh64(osm_physp_get_port_guid
(p_dest_physp)),
p_dest_physp->p_node->print_desc,
p_dest_physp->port_num,
cl_ntoh16(required_pkey));
status = IB_NOT_FOUND;
goto Exit;
}
if (p_qos_level && p_qos_level->pkey_range_len &&
!osm_qos_level_has_pkey(p_qos_level, required_pkey)) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 451C: "
"Ports src 0x%016"PRIx64" (%s port %d) "
"and dst 0x%016"PRIx64" (%s port %d) "
"do not share specified PKey (0x%04x) as "
"defined by QoS level \"%s\"\n",
cl_ntoh64(osm_physp_get_port_guid(p_src_physp)),
p_src_physp->p_node->print_desc,
p_src_physp->port_num,
cl_ntoh64(osm_physp_get_port_guid
(p_dest_physp)),
p_dest_physp->p_node->print_desc,
p_dest_physp->port_num,
cl_ntoh16(required_pkey),
p_qos_level->name);
status = IB_NOT_FOUND;
goto Exit;
}
} else if (p_qos_level && p_qos_level->pkey_range_len) {
/*
* MPR request doesn't have a specific pkey, but QoS level
* has pkeys - get shared pkey from QoS level pkeys
*/
required_pkey = osm_qos_level_get_shared_pkey(p_qos_level,
p_src_physp,
p_dest_physp,
sa->p_subn->opt.allow_both_pkeys);
if (!required_pkey) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 451D: "
"Ports src 0x%016"PRIx64" (%s port %d) "
"and dst 0x%016"PRIx64" (%s port %d) "
"do not share a PKey as defined by QoS "
"level \"%s\"\n",
cl_ntoh64(osm_physp_get_port_guid(p_src_physp)),
p_src_physp->p_node->print_desc,
p_src_physp->port_num,
cl_ntoh64(osm_physp_get_port_guid
(p_dest_physp)),
p_dest_physp->p_node->print_desc,
p_dest_physp->port_num,
p_qos_level->name);
status = IB_NOT_FOUND;
goto Exit;
}
} else {
/*
* Neither MPR request nor QoS level have pkey.
* Just get any shared pkey.
*/
required_pkey =
osm_physp_find_common_pkey(p_src_physp, p_dest_physp,
sa->p_subn->opt.allow_both_pkeys);
if (!required_pkey) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4519: "
"Ports src 0x%016"PRIx64" (%s port %d) "
"and dst 0x%016"PRIx64" (%s port %d) "
"do not have any shared PKeys\n",
cl_ntoh64(osm_physp_get_port_guid(p_src_physp)),
p_src_physp->p_node->print_desc,
p_src_physp->port_num,
cl_ntoh64(osm_physp_get_port_guid
(p_dest_physp)),
p_dest_physp->p_node->print_desc,
p_dest_physp->port_num);
status = IB_NOT_FOUND;
goto Exit;
}
}
if (required_pkey) {
p_prtn =
(osm_prtn_t *) cl_qmap_get(&sa->p_subn->prtn_pkey_tbl,
required_pkey &
cl_ntoh16((uint16_t) ~ 0x8000));
if (p_prtn ==
(osm_prtn_t *) cl_qmap_end(&sa->p_subn->prtn_pkey_tbl))
p_prtn = NULL;
}
/*
* Set MultiPathRecord SL.
*/
if (comp_mask & IB_MPR_COMPMASK_SL) {
/*
* Specific SL was requested
*/
required_sl = ib_multipath_rec_sl(p_mpr);
if (p_qos_level && p_qos_level->sl_set &&
p_qos_level->sl != required_sl) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 451E: "
"QoS constraints: required MultiPathRecord SL "
"(%u) doesn't match QoS policy \"%s\" SL (%u) "
"[%s port %d <-> %s port %d]\n", required_sl,
p_qos_level->name,
p_qos_level->sl,
p_src_alias_guid->p_base_port->p_node->print_desc,
p_src_alias_guid->p_base_port->p_physp->port_num,
p_dest_alias_guid->p_base_port->p_node->print_desc,
p_dest_alias_guid->p_base_port->p_physp->port_num);
status = IB_NOT_FOUND;
goto Exit;
}
} else if (p_qos_level && p_qos_level->sl_set) {
/*
* No specific SL was requested,
* but there is an SL in QoS level.
*/
required_sl = p_qos_level->sl;
if (required_pkey && p_prtn && p_prtn->sl != p_qos_level->sl)
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"QoS level SL (%u) overrides partition SL (%u)\n",
p_qos_level->sl, p_prtn->sl);
} else if (required_pkey) {
/*
* No specific SL in request or in QoS level - use partition SL
*/
p_prtn =
(osm_prtn_t *) cl_qmap_get(&sa->p_subn->prtn_pkey_tbl,
required_pkey &
cl_ntoh16((uint16_t) ~ 0x8000));
if (!p_prtn) {
required_sl = OSM_DEFAULT_SL;
/* this may be possible when pkey tables are created somehow in
previous runs or things are going wrong here */
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 451A: "
"No partition found for PKey 0x%04x - "
"using default SL %d "
"[%s port %d <-> %s port %d]\n",
cl_ntoh16(required_pkey), required_sl,
p_src_alias_guid->p_base_port->p_node->print_desc,
p_src_alias_guid->p_base_port->p_physp->port_num,
p_dest_alias_guid->p_base_port->p_node->print_desc,
p_dest_alias_guid->p_base_port->p_physp->port_num);
} else
required_sl = p_prtn->sl;
} else if (sa->p_subn->opt.qos) {
if (valid_sl_mask & (1 << OSM_DEFAULT_SL))
required_sl = OSM_DEFAULT_SL;
else {
for (i = 0; i < IB_MAX_NUM_VLS; i++)
if (valid_sl_mask & (1 << i))
break;
required_sl = i;
}
} else
required_sl = OSM_DEFAULT_SL;
if (sa->p_subn->opt.qos && !(valid_sl_mask & (1 << required_sl))) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 451F: "
"Selected SL (%u) leads to VL15 "
"[%s port %d <-> %s port %d]\n",
required_sl,
p_src_alias_guid->p_base_port->p_node->print_desc,
p_src_alias_guid->p_base_port->p_physp->port_num,
p_dest_alias_guid->p_base_port->p_node->print_desc,
p_dest_alias_guid->p_base_port->p_physp->port_num);
status = IB_NOT_FOUND;
goto Exit;
}
/* reset pkey when raw traffic */
if (comp_mask & IB_MPR_COMPMASK_RAWTRAFFIC &&
cl_ntoh32(p_mpr->hop_flow_raw) & (1 << 31))
required_pkey = 0;
p_parms->mtu = mtu;
p_parms->rate = rate;
p_parms->pkey = required_pkey;
p_parms->pkt_life = pkt_life;
p_parms->sl = required_sl;
p_parms->hops = hops;
OSM_LOG(sa->p_log, OSM_LOG_DEBUG, "MultiPath params:"
" mtu = %u, rate = %u, packet lifetime = %u,"
" pkey = 0x%04X, sl = %u, hops = %u\n", mtu, rate,
pkt_life, cl_ntoh16(required_pkey), required_sl, hops);
Exit:
OSM_LOG_EXIT(sa->p_log);
return status;
}
void osm_lftr_rcv_process(IN void *ctx, IN void *data)
{
osm_sa_t *sa = ctx;
osm_madw_t *p_madw = data;
const ib_sa_mad_t *p_rcvd_mad;
const ib_lft_record_t *p_rcvd_rec;
cl_qlist_t rec_list;
osm_lftr_search_ctxt_t context;
osm_physp_t *p_req_physp;
CL_ASSERT(sa);
OSM_LOG_ENTER(sa->p_log);
CL_ASSERT(p_madw);
p_rcvd_mad = osm_madw_get_sa_mad_ptr(p_madw);
p_rcvd_rec = (ib_lft_record_t *) ib_sa_mad_get_payload_ptr(p_rcvd_mad);
CL_ASSERT(p_rcvd_mad->attr_id == IB_MAD_ATTR_LFT_RECORD);
/* we only support SubnAdmGet and SubnAdmGetTable methods */
if (p_rcvd_mad->method != IB_MAD_METHOD_GET &&
p_rcvd_mad->method != IB_MAD_METHOD_GETTABLE) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4408: "
"Unsupported Method (%s)\n",
ib_get_sa_method_str(p_rcvd_mad->method));
osm_sa_send_error(sa, p_madw, IB_MAD_STATUS_UNSUP_METHOD_ATTR);
goto Exit;
}
/* update the requester physical port */
p_req_physp = osm_get_physp_by_mad_addr(sa->p_log, sa->p_subn,
osm_madw_get_mad_addr_ptr
(p_madw));
if (p_req_physp == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4407: "
"Cannot find requester physical port\n");
goto Exit;
}
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Requester port GUID 0x%" PRIx64 "\n",
cl_ntoh64(osm_physp_get_port_guid(p_req_physp)));
cl_qlist_init(&rec_list);
context.p_rcvd_rec = p_rcvd_rec;
context.p_list = &rec_list;
context.comp_mask = p_rcvd_mad->comp_mask;
context.sa = sa;
context.p_req_physp = p_req_physp;
cl_plock_acquire(sa->p_lock);
/* Go over all switches */
cl_qmap_apply_func(&sa->p_subn->sw_guid_tbl, lftr_rcv_by_comp_mask,
&context);
cl_plock_release(sa->p_lock);
osm_sa_respond(sa, p_madw, sizeof(ib_lft_record_t), &rec_list);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
/**********************************************************************
The plock must be held before calling this function.
**********************************************************************/
static void pi_rcv_process_switch_ext_port(IN osm_sm_t * sm,
IN osm_node_t * p_node,
IN osm_physp_t * p_physp,
IN ib_port_info_t * p_pi)
{
ib_api_status_t status = IB_SUCCESS;
osm_madw_context_t context;
osm_physp_t *p_remote_physp, *physp0;
osm_node_t *p_remote_node;
ib_net64_t m_key;
unsigned data_vls;
uint8_t port_num;
uint8_t remote_port_num;
osm_dr_path_t path;
int mlnx_epi_supported = 0;
OSM_LOG_ENTER(sm->p_log);
/*
Check the state of the physical port.
If there appears to be something on the other end of the wire,
then ask for NodeInfo. Ignore the switch management port.
*/
port_num = osm_physp_get_port_num(p_physp);
if (sm->p_subn->opt.fdr10)
mlnx_epi_supported = is_mlnx_ext_port_info_supported(
ib_node_info_get_vendor_id(&p_node->node_info),
p_node->node_info.device_id);
/* if in_sweep_hop_0 is TRUE, then this means the SM is on the switch,
and we got switchInfo of our local switch. Do not continue
probing through the switch. */
switch (ib_port_info_get_port_state(p_pi)) {
case IB_LINK_DOWN:
p_remote_physp = osm_physp_get_remote(p_physp);
if (p_remote_physp) {
p_remote_node =
osm_physp_get_node_ptr(p_remote_physp);
remote_port_num =
osm_physp_get_port_num(p_remote_physp);
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Unlinking local node 0x%" PRIx64
", port %u"
"\n\t\t\t\tand remote node 0x%" PRIx64
", port %u\n",
cl_ntoh64(osm_node_get_node_guid
(p_node)), port_num,
cl_ntoh64(osm_node_get_node_guid
(p_remote_node)),
remote_port_num);
if (sm->ucast_mgr.cache_valid)
osm_ucast_cache_add_link(&sm->ucast_mgr,
p_physp,
p_remote_physp);
osm_node_unlink(p_node, (uint8_t) port_num,
p_remote_node,
(uint8_t) remote_port_num);
}
break;
case IB_LINK_INIT:
case IB_LINK_ARMED:
case IB_LINK_ACTIVE:
physp0 = osm_node_get_physp_ptr(p_node, 0);
if (mlnx_epi_supported) {
m_key = ib_port_info_get_m_key(&physp0->port_info);
context.pi_context.node_guid = osm_node_get_node_guid(p_node);
context.pi_context.port_guid = osm_physp_get_port_guid(p_physp);
context.pi_context.set_method = FALSE;
context.pi_context.light_sweep = FALSE;
context.pi_context.active_transition = FALSE;
context.pi_context.client_rereg = FALSE;
status = osm_req_get(sm,
osm_physp_get_dr_path_ptr(p_physp),
IB_MAD_ATTR_MLNX_EXTENDED_PORT_INFO,
cl_hton32(port_num), FALSE, m_key,
CL_DISP_MSGID_NONE, &context);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 0F11: "
"Failure initiating MLNX ExtPortInfo request (%s)\n",
ib_get_err_str(status));
}
if (sm->p_subn->in_sweep_hop_0 == FALSE) {
/*
To avoid looping forever, only probe the port if it
is NOT the port that responded to the SMP.
Request node info from the other end of this link:
1) Copy the current path from the parent node.
2) Extend the path to the next hop thru this port.
3) Request node info with the new path
*/
if (p_pi->local_port_num !=
osm_physp_get_port_num(p_physp)) {
path = *osm_physp_get_dr_path_ptr(p_physp);
if (osm_dr_path_extend(&path,
osm_physp_get_port_num
(p_physp))) {
OSM_LOG(sm->p_log, OSM_LOG_ERROR,
"ERR 0F08: "
"DR path with hop count %d couldn't be extended\n",
path.hop_count);
break;
}
memset(&context, 0, sizeof(context));
context.ni_context.node_guid =
osm_node_get_node_guid(p_node);
context.ni_context.port_num =
osm_physp_get_port_num(p_physp);
status = osm_req_get(sm, &path,
IB_MAD_ATTR_NODE_INFO, 0,
TRUE, 0,
CL_DISP_MSGID_NONE,
&context);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR,
"ERR 0F02: "
"Failure initiating NodeInfo request (%s)\n",
ib_get_err_str(status));
} else
OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
"Skipping SMP responder port %u\n",
p_pi->local_port_num);
}
break;
default:
OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 0F03: "
"Unknown link state = %u, port = %u\n",
ib_port_info_get_port_state(p_pi),
p_pi->local_port_num);
break;
}
if (ib_port_info_get_port_state(p_pi) > IB_LINK_INIT && p_node->sw &&
!ib_switch_info_get_state_change(&p_node->sw->switch_info) &&
p_node->sw->need_update == 1)
p_node->sw->need_update = 0;
if (p_physp->need_update)
sm->p_subn->ignore_existing_lfts = TRUE;
/*
Update the PortInfo attribute.
*/
osm_physp_set_port_info(p_physp, p_pi, sm);
if (ib_port_info_get_port_state(p_pi) == IB_LINK_DOWN)
goto Exit;
p_remote_physp = osm_physp_get_remote(p_physp);
if (p_remote_physp) {
p_remote_node = osm_physp_get_node_ptr(p_remote_physp);
if (p_remote_node->sw) {
data_vls = 1U << (ib_port_info_get_vl_cap(p_pi) - 1);
if (data_vls > 1U << (sm->p_subn->opt.max_op_vls - 1))
data_vls = 1U << (sm->p_subn->opt.max_op_vls - 1);
if (data_vls >= IB_MAX_NUM_VLS)
data_vls = IB_MAX_NUM_VLS - 1;
if ((uint8_t)data_vls < sm->p_subn->min_sw_data_vls) {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Setting switch port minimal data VLs "
"to:%u defined by node:0x%"
PRIx64 ", port:%u\n", data_vls,
cl_ntoh64(osm_node_get_node_guid(p_node)),
port_num);
sm->p_subn->min_sw_data_vls = data_vls;
}
}
}
Exit:
OSM_LOG_EXIT(sm->p_log);
}
void osm_smir_rcv_process(IN void *ctx, IN void *data)
{
osm_sa_t *sa = ctx;
osm_madw_t *p_madw = data;
const ib_sa_mad_t *sad_mad;
const ib_sminfo_record_t *p_rcvd_rec;
const osm_port_t *p_port = NULL;
const ib_sm_info_t *p_smi;
cl_qlist_t rec_list;
osm_smir_search_ctxt_t context;
ib_api_status_t status = IB_SUCCESS;
ib_net64_t comp_mask;
ib_net64_t port_guid;
osm_physp_t *p_req_physp;
osm_port_t *local_port;
osm_remote_sm_t *p_rem_sm;
cl_qmap_t *p_sm_guid_tbl;
uint8_t pri_state;
CL_ASSERT(sa);
OSM_LOG_ENTER(sa->p_log);
CL_ASSERT(p_madw);
sad_mad = osm_madw_get_sa_mad_ptr(p_madw);
p_rcvd_rec = (ib_sminfo_record_t *) ib_sa_mad_get_payload_ptr(sad_mad);
comp_mask = sad_mad->comp_mask;
CL_ASSERT(sad_mad->attr_id == IB_MAD_ATTR_SMINFO_RECORD);
/* we only support SubnAdmGet and SubnAdmGetTable methods */
if (sad_mad->method != IB_MAD_METHOD_GET &&
sad_mad->method != IB_MAD_METHOD_GETTABLE) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2804: "
"Unsupported Method (%s)\n",
ib_get_sa_method_str(sad_mad->method));
osm_sa_send_error(sa, p_madw, IB_MAD_STATUS_UNSUP_METHOD_ATTR);
goto Exit;
}
/* update the requester physical port */
p_req_physp = osm_get_physp_by_mad_addr(sa->p_log, sa->p_subn,
osm_madw_get_mad_addr_ptr
(p_madw));
if (p_req_physp == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2803: "
"Cannot find requester physical port\n");
goto Exit;
}
if (OSM_LOG_IS_ACTIVE_V2(sa->p_log, OSM_LOG_DEBUG)) {
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Requester port GUID 0x%" PRIx64 "\n",
cl_ntoh64(osm_physp_get_port_guid(p_req_physp)));
osm_dump_sm_info_record_v2(sa->p_log, p_rcvd_rec, FILE_ID, OSM_LOG_DEBUG);
}
p_smi = &p_rcvd_rec->sm_info;
cl_qlist_init(&rec_list);
context.p_rcvd_rec = p_rcvd_rec;
context.p_list = &rec_list;
context.comp_mask = sad_mad->comp_mask;
context.sa = sa;
context.p_req_physp = p_req_physp;
cl_plock_acquire(sa->p_lock);
/*
If the user specified a LID, it obviously narrows our
work load, since we don't have to search every port
*/
if (comp_mask & IB_SMIR_COMPMASK_LID) {
p_port = osm_get_port_by_lid(sa->p_subn, p_rcvd_rec->lid);
if (!p_port) {
status = IB_NOT_FOUND;
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2806: "
"No port found with LID %u\n",
cl_ntoh16(p_rcvd_rec->lid));
}
}
if (status == IB_SUCCESS) {
/* Handle our own SM first */
local_port = osm_get_port_by_guid(sa->p_subn,
sa->p_subn->sm_port_guid);
if (!local_port) {
cl_plock_release(sa->p_lock);
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2809: "
"No port found with GUID 0x%016" PRIx64 "\n",
cl_ntoh64(sa->p_subn->sm_port_guid));
goto Exit;
}
if (!p_port || local_port == p_port) {
if (FALSE ==
osm_physp_share_pkey(sa->p_log, p_req_physp,
local_port->p_physp,
sa->p_subn->opt.allow_both_pkeys)) {
cl_plock_release(sa->p_lock);
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 2805: "
"Cannot get SMInfo record due to pkey violation\n");
goto Exit;
}
/* Check that other search components specified match */
if ((comp_mask & IB_SMIR_COMPMASK_GUID) &&
sa->p_subn->sm_port_guid != p_smi->guid)
goto Remotes;
if ((comp_mask & IB_SMIR_COMPMASK_PRIORITY) &&
sa->p_subn->opt.sm_priority !=
ib_sminfo_get_priority(p_smi))
goto Remotes;
if ((comp_mask & IB_SMIR_COMPMASK_SMSTATE) &&
sa->p_subn->sm_state != ib_sminfo_get_state(p_smi))
goto Remotes;
/* Now, add local SMInfo to list */
pri_state = sa->p_subn->sm_state & 0x0F;
pri_state |= (sa->p_subn->opt.sm_priority & 0x0F) << 4;
smir_rcv_new_smir(sa, local_port, context.p_list,
sa->p_subn->sm_port_guid,
cl_ntoh32(sa->p_subn->p_osm->stats.
qp0_mads_sent), pri_state,
p_req_physp);
}
Remotes:
if (p_port && p_port != local_port) {
/* Find remote SM corresponding to p_port */
port_guid = osm_port_get_guid(p_port);
p_sm_guid_tbl = &sa->p_subn->sm_guid_tbl;
p_rem_sm =
(osm_remote_sm_t *) cl_qmap_get(p_sm_guid_tbl,
port_guid);
if (p_rem_sm !=
(osm_remote_sm_t *) cl_qmap_end(p_sm_guid_tbl))
sa_smir_by_comp_mask(sa, p_rem_sm, &context);
else
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 280A: "
"No remote SM for GUID 0x%016" PRIx64
"\n", cl_ntoh64(port_guid));
} else {
/* Go over all other known (remote) SMs */
cl_qmap_apply_func(&sa->p_subn->sm_guid_tbl,
sa_smir_by_comp_mask_cb, &context);
}
}
cl_plock_release(sa->p_lock);
osm_sa_respond(sa, p_madw, sizeof(ib_sminfo_record_t), &rec_list);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
void osm_pkey_rec_rcv_process(IN void *ctx, IN void *data)
{
osm_sa_t *sa = ctx;
osm_madw_t *p_madw = data;
const ib_sa_mad_t *p_rcvd_mad;
const ib_pkey_table_record_t *p_rcvd_rec;
const osm_port_t *p_port = NULL;
cl_qlist_t rec_list;
osm_pkey_search_ctxt_t context;
ib_net64_t comp_mask;
osm_physp_t *p_req_physp;
CL_ASSERT(sa);
OSM_LOG_ENTER(sa->p_log);
CL_ASSERT(p_madw);
p_rcvd_mad = osm_madw_get_sa_mad_ptr(p_madw);
p_rcvd_rec =
(ib_pkey_table_record_t *) ib_sa_mad_get_payload_ptr(p_rcvd_mad);
comp_mask = p_rcvd_mad->comp_mask;
CL_ASSERT(p_rcvd_mad->attr_id == IB_MAD_ATTR_PKEY_TBL_RECORD);
/* we only support SubnAdmGet and SubnAdmGetTable methods */
if (p_rcvd_mad->method != IB_MAD_METHOD_GET &&
p_rcvd_mad->method != IB_MAD_METHOD_GETTABLE) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4605: "
"Unsupported Method (%s)\n",
ib_get_sa_method_str(p_rcvd_mad->method));
osm_sa_send_error(sa, p_madw, IB_MAD_STATUS_UNSUP_METHOD_ATTR);
goto Exit;
}
/*
p922 - P_KeyTableRecords shall only be provided in response
to trusted requests.
Check that the requester is a trusted one.
*/
if (p_rcvd_mad->sm_key != sa->p_subn->opt.sa_key) {
/* This is not a trusted requester! */
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4608: "
"Request from non-trusted requester: "
"Given SM_Key:0x%016" PRIx64 "\n",
cl_ntoh64(p_rcvd_mad->sm_key));
osm_sa_send_error(sa, p_madw, IB_SA_MAD_STATUS_REQ_INVALID);
goto Exit;
}
/* update the requester physical port */
p_req_physp = osm_get_physp_by_mad_addr(sa->p_log, sa->p_subn,
osm_madw_get_mad_addr_ptr
(p_madw));
if (p_req_physp == NULL) {
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 4604: "
"Cannot find requester physical port\n");
goto Exit;
}
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Requester port GUID 0x%" PRIx64 "\n",
cl_ntoh64(osm_physp_get_port_guid(p_req_physp)));
cl_qlist_init(&rec_list);
context.p_rcvd_rec = p_rcvd_rec;
context.p_list = &rec_list;
context.comp_mask = p_rcvd_mad->comp_mask;
context.sa = sa;
context.block_num = cl_ntoh16(p_rcvd_rec->block_num);
context.p_req_physp = p_req_physp;
OSM_LOG(sa->p_log, OSM_LOG_DEBUG,
"Got Query Lid:%u(%02X), Block:0x%02X(%02X), Port:0x%02X(%02X)\n",
cl_ntoh16(p_rcvd_rec->lid),
(comp_mask & IB_PKEY_COMPMASK_LID) != 0, p_rcvd_rec->port_num,
(comp_mask & IB_PKEY_COMPMASK_PORT) != 0, context.block_num,
(comp_mask & IB_PKEY_COMPMASK_BLOCK) != 0);
cl_plock_acquire(sa->p_lock);
/*
If the user specified a LID, it obviously narrows our
work load, since we don't have to search every port
*/
if (comp_mask & IB_PKEY_COMPMASK_LID) {
p_port = osm_get_port_by_lid(sa->p_subn, p_rcvd_rec->lid);
if (!p_port)
OSM_LOG(sa->p_log, OSM_LOG_ERROR, "ERR 460B: "
"No port found with LID %u\n",
cl_ntoh16(p_rcvd_rec->lid));
else
sa_pkey_by_comp_mask(sa, p_port, &context);
} else
cl_qmap_apply_func(&sa->p_subn->port_guid_tbl,
sa_pkey_by_comp_mask_cb, &context);
cl_plock_release(sa->p_lock);
osm_sa_respond(sa, p_madw, sizeof(ib_pkey_table_record_t), &rec_list);
Exit:
OSM_LOG_EXIT(sa->p_log);
}
static void
__osm_pi_rcv_process_endport(IN osm_sm_t * sm,
IN osm_physp_t * const p_physp,
IN const ib_port_info_t * const p_pi)
{
osm_madw_context_t context;
ib_api_status_t status;
ib_net64_t port_guid;
uint8_t rate, mtu;
cl_qmap_t *p_sm_tbl;
osm_remote_sm_t *p_sm;
OSM_LOG_ENTER(sm->p_log);
port_guid = osm_physp_get_port_guid(p_physp);
/* HACK extended port 0 should be handled too! */
if (osm_physp_get_port_num(p_physp) != 0) {
/* track the minimal endport MTU and rate */
mtu = ib_port_info_get_mtu_cap(p_pi);
if (mtu < sm->p_subn->min_ca_mtu) {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Setting endport minimal MTU to:%u defined by port:0x%"
PRIx64 "\n", mtu, cl_ntoh64(port_guid));
sm->p_subn->min_ca_mtu = mtu;
}
rate = ib_port_info_compute_rate(p_pi);
if (rate < sm->p_subn->min_ca_rate) {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Setting endport minimal rate to:%u defined by port:0x%"
PRIx64 "\n", rate, cl_ntoh64(port_guid));
sm->p_subn->min_ca_rate = rate;
}
}
if (port_guid == sm->p_subn->sm_port_guid) {
/*
We received the PortInfo for our own port.
*/
if (!(p_pi->capability_mask & IB_PORT_CAP_IS_SM))
/*
Set the IS_SM bit to indicate our port hosts an SM.
*/
__osm_pi_rcv_set_sm(sm, p_physp);
} else {
p_sm_tbl = &sm->p_subn->sm_guid_tbl;
if (p_pi->capability_mask & IB_PORT_CAP_IS_SM) {
/*
* Before querying the SM - we want to make sure we
* clean its state, so if the querying fails we
* recognize that this SM is not active.
*/
p_sm = (osm_remote_sm_t *) cl_qmap_get(p_sm_tbl, port_guid);
if (p_sm != (osm_remote_sm_t *) cl_qmap_end(p_sm_tbl))
/* clean it up */
p_sm->smi.pri_state = 0xF0 & p_sm->smi.pri_state;
if (sm->p_subn->opt.ignore_other_sm)
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Ignoring SM on port 0x%" PRIx64 "\n",
cl_ntoh64(port_guid));
else {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Detected another SM. Requesting SMInfo"
"\n\t\t\t\tPort 0x%" PRIx64 "\n",
cl_ntoh64(port_guid));
/*
This port indicates it's an SM and
it's not our own port.
Acquire the SMInfo Attribute.
*/
memset(&context, 0, sizeof(context));
context.smi_context.set_method = FALSE;
context.smi_context.port_guid = port_guid;
status = osm_req_get(sm,
osm_physp_get_dr_path_ptr
(p_physp),
IB_MAD_ATTR_SM_INFO, 0,
CL_DISP_MSGID_NONE,
&context);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR,
"ERR 0F05: "
"Failure requesting SMInfo (%s)\n",
ib_get_err_str(status));
}
} else {
p_sm = (osm_remote_sm_t *) cl_qmap_remove(p_sm_tbl, port_guid);
if (p_sm != (osm_remote_sm_t *) cl_qmap_end(p_sm_tbl))
free(p_sm);
}
}
OSM_LOG_EXIT(sm->p_log);
}
static void pi_rcv_process_endport(IN osm_sm_t * sm, IN osm_physp_t * p_physp,
IN const ib_port_info_t * p_pi)
{
osm_madw_context_t context;
ib_api_status_t status;
ib_net64_t port_guid;
int extended;
uint8_t rate, mtu, mpb;
unsigned data_vls;
cl_qmap_t *p_sm_tbl;
osm_remote_sm_t *p_sm;
OSM_LOG_ENTER(sm->p_log);
port_guid = osm_physp_get_port_guid(p_physp);
/* HACK extended port 0 should be handled too! */
if (osm_physp_get_port_num(p_physp) != 0 &&
ib_port_info_get_port_state(p_pi) != IB_LINK_DOWN) {
/* track the minimal endport MTU, rate, and operational VLs */
mtu = ib_port_info_get_mtu_cap(p_pi);
if (mtu < sm->p_subn->min_ca_mtu) {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Setting endport minimal MTU to:%u defined by port:0x%"
PRIx64 "\n", mtu, cl_ntoh64(port_guid));
sm->p_subn->min_ca_mtu = mtu;
}
extended = p_pi->capability_mask & IB_PORT_CAP_HAS_EXT_SPEEDS;
rate = ib_port_info_compute_rate(p_pi, extended);
if (ib_path_compare_rates(rate, sm->p_subn->min_ca_rate) < 0) {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Setting endport minimal rate to:%u defined by port:0x%"
PRIx64 "\n", rate, cl_ntoh64(port_guid));
sm->p_subn->min_ca_rate = rate;
}
data_vls = 1U << (ib_port_info_get_vl_cap(p_pi) - 1);
if (data_vls > 1U << (sm->p_subn->opt.max_op_vls - 1))
data_vls = 1U << (sm->p_subn->opt.max_op_vls - 1);
if (data_vls >= IB_MAX_NUM_VLS)
data_vls = IB_MAX_NUM_VLS - 1;
if ((uint8_t)data_vls < sm->p_subn->min_data_vls) {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Setting endport minimal data VLs to:%u defined by port:0x%"
PRIx64 "\n", data_vls, cl_ntoh64(port_guid));
sm->p_subn->min_data_vls = data_vls;
}
}
/* Check M_Key vs M_Key protect, can we control the port ? */
mpb = ib_port_info_get_mpb(p_pi);
if (mpb > 0 && p_pi->m_key == 0) {
OSM_LOG(sm->p_log, OSM_LOG_INFO,
"Port 0x%" PRIx64 " has unknown M_Key, protection level %u\n",
cl_ntoh64(port_guid), mpb);
}
if (port_guid != sm->p_subn->sm_port_guid) {
p_sm_tbl = &sm->p_subn->sm_guid_tbl;
if (p_pi->capability_mask & IB_PORT_CAP_IS_SM) {
/*
* Before querying the SM - we want to make sure we
* clean its state, so if the querying fails we
* recognize that this SM is not active.
*/
p_sm =
(osm_remote_sm_t *) cl_qmap_get(p_sm_tbl,
port_guid);
if (p_sm != (osm_remote_sm_t *) cl_qmap_end(p_sm_tbl))
/* clean it up */
p_sm->smi.pri_state =
0xF0 & p_sm->smi.pri_state;
if (sm->p_subn->opt.ignore_other_sm)
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Ignoring SM on port 0x%" PRIx64 "\n",
cl_ntoh64(port_guid));
else {
OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
"Detected another SM. Requesting SMInfo "
"from port 0x%" PRIx64 "\n",
cl_ntoh64(port_guid));
/*
This port indicates it's an SM and
it's not our own port.
Acquire the SMInfo Attribute.
*/
memset(&context, 0, sizeof(context));
context.smi_context.set_method = FALSE;
context.smi_context.port_guid = port_guid;
status = osm_req_get(sm,
osm_physp_get_dr_path_ptr
(p_physp),
IB_MAD_ATTR_SM_INFO, 0,
FALSE,
ib_port_info_get_m_key(&p_physp->port_info),
CL_DISP_MSGID_NONE,
&context);
if (status != IB_SUCCESS)
OSM_LOG(sm->p_log, OSM_LOG_ERROR,
"ERR 0F05: "
"Failure requesting SMInfo (%s) "
"from port 0x%" PRIx64 "\n",
ib_get_err_str(status),
cl_ntoh64(port_guid));
}
} else {
p_sm =
(osm_remote_sm_t *) cl_qmap_remove(p_sm_tbl,
port_guid);
if (p_sm != (osm_remote_sm_t *) cl_qmap_end(p_sm_tbl))
free(p_sm);
}
}
OSM_LOG_EXIT(sm->p_log);
}
