@@ -246,7 +246,7 @@ void osm_perfmgr_clear_counters(osm_perfmgr_t * p_perfmgr);
void osm_perfmgr_dump_counters(osm_perfmgr_t * p_perfmgr,
perfmgr_db_dump_t dump_type);
void osm_perfmgr_print_counters(osm_perfmgr_t *pm, char *nodename, FILE *fp,
- char *port);
+ char *port, int err_only);
ib_api_status_t osm_perfmgr_bind(osm_perfmgr_t * p_perfmgr,
ib_net64_t port_guid);
@@ -190,11 +190,11 @@ perfmgr_db_err_t perfmgr_db_mark_active(perfmgr_db_t *db, uint64_t guid,
void perfmgr_db_clear_counters(perfmgr_db_t * db);
perfmgr_db_err_t perfmgr_db_dump(perfmgr_db_t * db, char *file,
perfmgr_db_dump_t dump_type);
-void perfmgr_db_print_all(perfmgr_db_t * db, FILE *fp);
+void perfmgr_db_print_all(perfmgr_db_t * db, FILE *fp, int err_only);
void perfmgr_db_print_by_name(perfmgr_db_t * db, char *nodename, FILE *fp,
- char *port);
+ char *port, int err_only);
void perfmgr_db_print_by_guid(perfmgr_db_t * db, uint64_t guid, FILE *fp,
- char *port);
+ char *port, int err_only);
/** =========================================================================
* helper functions to fill in the various db objects from wire objects
@@ -240,7 +240,7 @@ static void help_perfmgr(FILE * out, int detail)
{
fprintf(out,
"perfmgr(pm) [enable|disable\n"
- " |clear_counters|dump_counters|print_counters(pc)\n"
+ " |clear_counters|dump_counters|print_counters(pc)|print_errors(pe)\n"
" |set_rm_nodes|clear_rm_nodes|clear_inactive\n"
" |dump_redir|clear_redir|sweep_time[seconds]]\n");
if (detail) {
@@ -260,6 +260,11 @@ static void help_perfmgr(FILE * out, int detail)
fprintf(out,
" [pc [<nodename|nodeguid>][:<port>]] -- same as print_counters\n");
fprintf(out,
+ " [print_errors [<nodename|nodeguid>]] -- print only ports with errors\n"
+ " Optionaly limit output by name or guid\n");
+ fprintf(out,
+ " [pe [<nodename|nodeguid>]] -- same as print_errors\n");
+ fprintf(out,
" [dump_redir [<nodename|nodeguid>]] -- dump the redirection table\n");
fprintf(out,
" [clear_redir [<nodename|nodeguid>]] -- clear the redirection table\n");
@@ -1487,7 +1492,12 @@ static void perfmgr_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
}
}
osm_perfmgr_print_counters(&p_osm->perfmgr, p_cmd,
- out, port);
+ out, port, 0);
+ } else if (strcmp(p_cmd, "print_errors") == 0 ||
+ strcmp(p_cmd, "pe") == 0) {
+ p_cmd = name_token(p_last);
+ osm_perfmgr_print_counters(&p_osm->perfmgr, p_cmd,
+ out, NULL, 1);
} else if (strcmp(p_cmd, "dump_redir") == 0) {
p_cmd = name_token(p_last);
dump_redir(p_osm, p_cmd, out);
@@ -1437,16 +1437,18 @@ void osm_perfmgr_dump_counters(osm_perfmgr_t * pm, perfmgr_db_dump_t dump_type)
* Print the DB information to the fp specified
*******************************************************************/
void osm_perfmgr_print_counters(osm_perfmgr_t * pm, char *nodename, FILE * fp,
- char *port)
+ char *port, int err_only)
{
if (nodename) {
char *end = NULL;
uint64_t guid = strtoull(nodename, &end, 0);
if (nodename + strlen(nodename) != end)
- perfmgr_db_print_by_name(pm->db, nodename, fp, port);
+ perfmgr_db_print_by_name(pm->db, nodename, fp, port,
+ err_only);
else
- perfmgr_db_print_by_guid(pm->db, guid, fp, port);
+ perfmgr_db_print_by_guid(pm->db, guid, fp, port,
+ err_only);
} else
- perfmgr_db_print_all(pm->db, fp);
+ perfmgr_db_print_all(pm->db, fp, err_only);
}
#endif /* ENABLE_OSM_PERF_MGR */
@@ -751,7 +751,7 @@ static void dump_hr_dc(FILE *fp, uint64_t val64, int data)
/**********************************************************************
* Output a human readable output of the port counters
**********************************************************************/
-static void dump_node_hr(db_node_t * node, FILE * fp, char *port)
+static void dump_node_hr(db_node_t * node, FILE * fp, char *port, int err_only)
{
int i = (node->esp0) ? 0 : 1;
int num_ports = node->num_ports;
@@ -766,7 +766,6 @@ static void dump_node_hr(db_node_t * node, FILE * fp, char *port)
fprintf(fp, "Warning: \"%s\" is not a valid port\n", port);
}
}
- fprintf(fp, "\n");
for (/* set above */; i < num_ports; i++) {
char *since = ctime(&node->ports[i].last_reset);
@@ -774,37 +773,63 @@ static void dump_node_hr(db_node_t * node, FILE * fp, char *port)
continue;
since[strlen(since) - 1] = '\0'; /* remove \n */
+ perfmgr_db_err_reading_t *err = &node->ports[i].err_total;
+
+ if (err_only
+ && err->symbol_err_cnt == 0
+ && err->link_err_recover == 0
+ && err->link_downed == 0
+ && err->rcv_err == 0
+ && err->rcv_rem_phys_err == 0
+ && err->rcv_switch_relay_err == 0
+ && err->xmit_discards == 0
+ && err->xmit_constraint_err == 0
+ && err->rcv_constraint_err == 0
+ && err->link_integrity == 0
+ && err->buffer_overrun == 0
+ && err->vl15_dropped == 0)
+ continue;
- fprintf(fp, "\"%s\" 0x%" PRIx64 " active %s port %d (Since %s)\n"
- " symbol_err_cnt : %" PRIu64 "\n"
- " link_err_recover : %" PRIu64 "\n"
- " link_downed : %" PRIu64 "\n"
- " rcv_err : %" PRIu64 "\n"
- " rcv_rem_phys_err : %" PRIu64 "\n"
- " rcv_switch_relay_err : %" PRIu64 "\n"
- " xmit_discards : %" PRIu64 "\n"
- " xmit_constraint_err : %" PRIu64 "\n"
- " rcv_constraint_err : %" PRIu64 "\n"
- " link_integrity_err : %" PRIu64 "\n"
- " buf_overrun_err : %" PRIu64 "\n"
- " vl15_dropped : %" PRIu64 "\n",
- node->node_name,
- node->node_guid,
- node->active ? "TRUE":"FALSE",
- i,
- since,
- node->ports[i].err_total.symbol_err_cnt,
- node->ports[i].err_total.link_err_recover,
- node->ports[i].err_total.link_downed,
- node->ports[i].err_total.rcv_err,
- node->ports[i].err_total.rcv_rem_phys_err,
- node->ports[i].err_total.rcv_switch_relay_err,
- node->ports[i].err_total.xmit_discards,
- node->ports[i].err_total.xmit_constraint_err,
- node->ports[i].err_total.rcv_constraint_err,
- node->ports[i].err_total.link_integrity,
- node->ports[i].err_total.buffer_overrun,
- node->ports[i].err_total.vl15_dropped);
+ fprintf(fp, "\"%s\" 0x%" PRIx64 " active %s port %d (Since %s)\n",
+ node->node_name, node->node_guid,
+ node->active ? "TRUE":"FALSE", i, since);
+
+ if (!err_only || err->symbol_err_cnt != 0)
+ fprintf(fp, " symbol_err_cnt : %" PRIu64 "\n",
+ err->symbol_err_cnt);
+ if (!err_only || err->link_err_recover != 0)
+ fprintf(fp, " link_err_recover : %" PRIu64 "\n",
+ err->link_err_recover);
+ if (!err_only || err->link_downed != 0)
+ fprintf(fp, " link_downed : %" PRIu64 "\n",
+ err->link_downed);
+ if (!err_only || err->rcv_err != 0)
+ fprintf(fp, " rcv_err : %" PRIu64 "\n",
+ err->rcv_err);
+ if (!err_only || err->rcv_rem_phys_err != 0)
+ fprintf(fp, " rcv_rem_phys_err : %" PRIu64 "\n",
+ err->rcv_rem_phys_err);
+ if (!err_only || err->rcv_switch_relay_err != 0)
+ fprintf(fp, " rcv_switch_relay_err : %" PRIu64 "\n",
+ err->rcv_switch_relay_err);
+ if (!err_only || err->xmit_discards != 0)
+ fprintf(fp, " xmit_discards : %" PRIu64 "\n",
+ err->xmit_discards);
+ if (!err_only || err->xmit_constraint_err != 0)
+ fprintf(fp, " xmit_constraint_err : %" PRIu64 "\n",
+ err->xmit_constraint_err);
+ if (!err_only || err->rcv_constraint_err != 0)
+ fprintf(fp, " rcv_constraint_err : %" PRIu64 "\n",
+ err->rcv_constraint_err);
+ if (!err_only || err->link_integrity != 0)
+ fprintf(fp, " link_integrity_err : %" PRIu64 "\n",
+ err->link_integrity);
+ if (!err_only || err->buffer_overrun != 0)
+ fprintf(fp, " buf_overrun_err : %" PRIu64 "\n",
+ err->buffer_overrun);
+ if (!err_only || err->vl15_dropped != 0)
+ fprintf(fp, " vl15_dropped : %" PRIu64 "\n",
+ err->vl15_dropped);
fprintf(fp, " xmit_data : %" PRIu64,
node->ports[i].dc_total.xmit_data);
@@ -852,7 +877,7 @@ static void db_dump(cl_map_item_t * const p_map_item, void *context)
break;
case PERFMGR_EVENT_DB_DUMP_HR:
default:
- dump_node_hr(node, fp, NULL);
+ dump_node_hr(node, fp, NULL, 0);
break;
}
}
@@ -861,7 +886,7 @@ static void db_dump(cl_map_item_t * const p_map_item, void *context)
* print all node data to fp
**********************************************************************/
void
-perfmgr_db_print_all(perfmgr_db_t * db, FILE *fp)
+perfmgr_db_print_all(perfmgr_db_t * db, FILE *fp, int err_only)
{
cl_map_item_t *item;
db_node_t *node;
@@ -870,7 +895,7 @@ perfmgr_db_print_all(perfmgr_db_t * db, FILE *fp)
item = cl_qmap_head(&db->pc_data);
while (item != cl_qmap_end(&db->pc_data)) {
node = (db_node_t *)item;
- dump_node_hr(node, fp, NULL);
+ dump_node_hr(node, fp, NULL, err_only);
item = cl_qmap_next(item);
}
cl_plock_release(&db->lock);
@@ -881,7 +906,7 @@ perfmgr_db_print_all(perfmgr_db_t * db, FILE *fp)
**********************************************************************/
void
perfmgr_db_print_by_name(perfmgr_db_t * db, char *nodename, FILE *fp,
- char *port)
+ char *port, int err_only)
{
cl_map_item_t *item;
db_node_t *node;
@@ -893,7 +918,7 @@ perfmgr_db_print_by_name(perfmgr_db_t * db, char *nodename, FILE *fp,
while (item != cl_qmap_end(&db->pc_data)) {
node = (db_node_t *)item;
if (strcmp(node->node_name, nodename) == 0) {
- dump_node_hr(node, fp, port);
+ dump_node_hr(node, fp, port, err_only);
goto done;
}
item = cl_qmap_next(item);
@@ -909,7 +934,7 @@ done:
**********************************************************************/
void
perfmgr_db_print_by_guid(perfmgr_db_t * db, uint64_t nodeguid, FILE *fp,
- char *port)
+ char *port, int err_only)
{
cl_map_item_t *node;
@@ -917,7 +942,7 @@ perfmgr_db_print_by_guid(perfmgr_db_t * db, uint64_t nodeguid, FILE *fp,
node = cl_qmap_get(&db->pc_data, nodeguid);
if (node != cl_qmap_end(&db->pc_data))
- dump_node_hr((db_node_t *)node, fp, port);
+ dump_node_hr((db_node_t *)node, fp, port, err_only);
else
fprintf(fp, "Node 0x%" PRIx64 " not found...\n", nodeguid);
Only print ports with error counters > 0 Changes since V1: My rebase got cross wired for the console help message. Fix this. Signed-off-by: Ira Weiny <weiny2@llnl.gov> --- include/opensm/osm_perfmgr.h | 2 +- include/opensm/osm_perfmgr_db.h | 6 +- opensm/osm_console.c | 14 +++++- opensm/osm_perfmgr.c | 10 ++-- opensm/osm_perfmgr_db.c | 103 ++++++++++++++++++++++++--------------- 5 files changed, 86 insertions(+), 49 deletions(-)