opensm: Add update_desc command to opensm console

PATCH Superseded Eli Dorfman (Voltaire) 2 participants 5 messages Jan 12, 2010 – Jan 13, 2010 2 messages not in the archive InfiniBand

lore

Share & tools

lore

Eli Dorfman (Voltaire)

author

2010-01-12 08:32 UTC #0 lore raw

PATCH 1/1 opensm: Add update_desc command to opensm console

Add update_desc command to opensm console

Sometimes nodes are discovered by the opensm with their default name before real host name was resolves and updated in the HCA. Since Mellanox ConnectX HCA do not support NodeDescription change trap (as part of trap 144) we need an option to refresh NodeDescription from console.

Signed-off-by Eli Dorfman <elid-smomgflXvOZWk0Htik3J/[email protected]>

---
 opensm/opensm/osm_console.c   |   16 +++++++++++++
 opensm/opensm/osm_state_mgr.c |   48 +++++++++++++++++++++++++++++-----------
 2 files changed, 51 insertions(+), 13 deletions(-)

opensm/opensm/osm_console.c +16 -0

index 206e7f7..e90cb99 100644
@@ -56,6 +56,8 @@
 #include <opensm/osm_perfmgr.h>
 #include <opensm/osm_subnet.h>
 
+extern void osm_update_node_desc(IN osm_sm_t *sm);
+
 struct command {
 	char *name;
 	void (*help_function) (FILE * out, int detail);

@@ -207,6 +209,14 @@ static void help_dump_conf(FILE *out, int detail)
 	}
 }
 
+static void help_update_desc(FILE *out, int detail)
+{
+	fprintf(out, "update_desc\n");
+	if (detail) {
+		fprintf(out, "update node description for all nodes\n");
+	}
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void help_perfmgr(FILE * out, int detail)
 {

@@ -1134,6 +1144,11 @@ static void dump_conf_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 	osm_subn_output_conf(out, &p_osm->subn.opt);
 }
 
+static void update_desc_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
+{
+	osm_update_node_desc(&p_osm->sm);
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void perfmgr_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 {

@@ -1326,6 +1341,7 @@ static const struct command console_cmds[] = {
 	{"switchbalance", &help_switchbalance, &switchbalance_parse},
 	{"lidbalance", &help_lidbalance, &lidbalance_parse},
 	{"dump_conf", &help_dump_conf, &dump_conf_parse},
+	{"update_desc", &help_update_desc, &update_desc_parse},
 	{"version", &help_version, &version_parse},
 #ifdef ENABLE_OSM_PERF_MGR
 	{"perfmgr", &help_perfmgr, &perfmgr_parse},

opensm/opensm/osm_state_mgr.c +35 ~~-13~~

66 lines not on this page · raw message

Signature · 6 lines

1.5.5

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

A message that is not in the archive (sent off-list or not yet received): [email protected]

Sasha Khapyorsky

2010-01-12 12:14 UTC #2 lore raw

Hi Eli,

Quoted 9 lines · Eli Dorfman (Voltaire)

Add update_desc command to opensm console
Sometimes nodes are discovered by the opensm with their default name before real host name was resolves and updated in the HCA. Since Mellanox ConnectX HCA do not support NodeDescription change trap (as part of trap 144) we need an option to refresh NodeDescription from console.
Signed-off-by Eli Dorfman <elid-smomgflXvOZWk0Htik3J/[email protected]>

I understand the reason for the patch. Two questions are below.

Quoted 88 lines

---
 opensm/opensm/osm_console.c   |   16 +++++++++++++
 opensm/opensm/osm_state_mgr.c |   48 +++++++++++++++++++++++++++++-----------
 2 files changed, 51 insertions(+), 13 deletions(-)

opensm/opensm/osm_console.c +16 -0

index 206e7f7..e90cb99 100644
@@ -56,6 +56,8 @@
 #include <opensm/osm_perfmgr.h>
 #include <opensm/osm_subnet.h>
 
+extern void osm_update_node_desc(IN osm_sm_t *sm);
+
 struct command {
 	char *name;
 	void (*help_function) (FILE * out, int detail);

@@ -207,6 +209,14 @@ static void help_dump_conf(FILE *out, int detail)
 	}
 }
 
+static void help_update_desc(FILE *out, int detail)
+{
+	fprintf(out, "update_desc\n");
+	if (detail) {
+		fprintf(out, "update node description for all nodes\n");
+	}
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void help_perfmgr(FILE * out, int detail)
 {

@@ -1134,6 +1144,11 @@ static void dump_conf_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 	osm_subn_output_conf(out, &p_osm->subn.opt);
 }
 
+static void update_desc_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
+{
+	osm_update_node_desc(&p_osm->sm);
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void perfmgr_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 {

@@ -1326,6 +1341,7 @@ static const struct command console_cmds[] = {
 	{"switchbalance", &help_switchbalance, &switchbalance_parse},
 	{"lidbalance", &help_lidbalance, &lidbalance_parse},
 	{"dump_conf", &help_dump_conf, &dump_conf_parse},
+	{"update_desc", &help_update_desc, &update_desc_parse},
 	{"version", &help_version, &version_parse},
 #ifdef ENABLE_OSM_PERF_MGR
 	{"perfmgr", &help_perfmgr, &perfmgr_parse},

opensm/opensm/osm_state_mgr.c +3 ~~-13~~

index 6296f21..01dfba6 100644
@@ -510,11 +510,7 @@ static void query_sm_info(cl_map_item_t * item, void *cxt)
 			ib_get_err_str(ret));
 }
 
-/**********************************************************************
- During a light sweep, check each node to see if the node description
- is valid and if not issue a ND query.
-**********************************************************************/
-static void state_mgr_get_node_desc(IN cl_map_item_t * obj, IN void *context)
+static void state_mgr_update_node_desc(IN cl_map_item_t * obj, IN void *context)
 {
 	osm_madw_context_t mad_context;
 	osm_node_t *p_node = (osm_node_t *) obj;

@@ -527,14 +523,8 @@ static void state_mgr_get_node_desc(IN cl_map_item_t * obj, IN void *context)
 
 	CL_ASSERT(p_node);
 
-	if (p_node->print_desc
-	    && strcmp(p_node->print_desc, OSM_NODE_DESC_UNKNOWN))
-		/* if ND is valid, do nothing */
-		goto exit;
-
-	OSM_LOG(sm->p_log, OSM_LOG_ERROR,
-		"ERR 3319: Unknown node description for node GUID "
-		"0x%016" PRIx64 ".  Reissuing ND query\n",
+	OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
+		"Updating NodeDesc for 0x%016" PRIx64 "\n",
 		cl_ntoh64(osm_node_get_node_guid(p_node)));
 
 	/* get a physp to request from. */

opensm/opensm/osm_state_mgr.c exit: · line 558

Quoted 6 lines

opensm/opensm/osm_state_mgr.c +3 -0

@@ -563,6 +553,38 @@ exit:
 	OSM_LOG_EXIT(sm->p_log);
 }
 
+void osm_update_node_desc(IN osm_sm_t *sm)
+{
+	CL_PLOCK_ACQUIRE(sm->p_lock);

Wouldn't you need exclusive locking here (CL_PLOCK_EXCL_ACQUIRE())?

Quoted 13 lines

opensm/opensm/osm_state_mgr.c +13 -0

+	cl_qmap_apply_func(&sm->p_subn->node_guid_tbl, state_mgr_update_node_desc,
+			   sm);
+	CL_PLOCK_RELEASE(sm->p_lock);
+}
+
+/**********************************************************************
+ During a light sweep, check each node to see if the node description
+ is valid and if not issue a ND query.
+**********************************************************************/
+static void state_mgr_get_node_desc(IN cl_map_item_t * obj, IN void *context)
+{
+	osm_node_t *p_node = (osm_node_t *) obj;
+	osm_sm_t *sm = context;

opensm/opensm/osm_state_mgr.c exit: · line 581

Quoted 10 lines

opensm/opensm/osm_state_mgr.c +10 -0

+
+	OSM_LOG_ENTER(sm->p_log);
+
+	CL_ASSERT(p_node);
+
+	if (p_node->print_desc
+	    && strcmp(p_node->print_desc, OSM_NODE_DESC_UNKNOWN))
+		/* if ND is valid, do nothing */
+		goto exit;
+

Seems that "ERR 3319:" error message was dropped forever. Any reason for doing this?

Sasha

Quoted 12 lines

opensm/opensm/osm_state_mgr.c +6 -0

+	state_mgr_update_node_desc(obj, context);
+
+exit:
+	OSM_LOG_EXIT(sm->p_log);
+}
+
 /**********************************************************************
  Initiates a lightweight sweep of the subnet.
  Used during normal sweeps after the subnet is up.

Signature · 1 line

1.5.5

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

Eli Dorfman (Voltaire)

author

2010-01-13 15:18 UTC #3 lore raw

Quoted 106 lines · Sasha Khapyorsky

Hi Eli,

Quoted 9 lines · Eli Dorfman (Voltaire)

Add update_desc command to opensm console
Sometimes nodes are discovered by the opensm with their default name before real host name was resolves and updated in the HCA. Since Mellanox ConnectX HCA do not support NodeDescription change trap (as part of trap 144) we need an option to refresh NodeDescription from console.
Signed-off-by Eli Dorfman <elid-smomgflXvOZWk0Htik3J/[email protected]>

I understand the reason for the patch. Two questions are below.

Quoted 91 lines

---
 opensm/opensm/osm_console.c   |   16 +++++++++++++
 opensm/opensm/osm_state_mgr.c |   48 +++++++++++++++++++++++++++++-----------
 2 files changed, 51 insertions(+), 13 deletions(-)

opensm/opensm/osm_console.c +16 -0

index 206e7f7..e90cb99 100644
@@ -56,6 +56,8 @@
 #include <opensm/osm_perfmgr.h>
 #include <opensm/osm_subnet.h>
 
+extern void osm_update_node_desc(IN osm_sm_t *sm);
+
 struct command {
 	char *name;
 	void (*help_function) (FILE * out, int detail);

@@ -207,6 +209,14 @@ static void help_dump_conf(FILE *out, int detail)
 	}
 }
 
+static void help_update_desc(FILE *out, int detail)
+{
+	fprintf(out, "update_desc\n");
+	if (detail) {
+		fprintf(out, "update node description for all nodes\n");
+	}
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void help_perfmgr(FILE * out, int detail)
 {

@@ -1134,6 +1144,11 @@ static void dump_conf_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 	osm_subn_output_conf(out, &p_osm->subn.opt);
 }
 
+static void update_desc_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
+{
+	osm_update_node_desc(&p_osm->sm);
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void perfmgr_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 {

@@ -1326,6 +1341,7 @@ static const struct command console_cmds[] = {
 	{"switchbalance", &help_switchbalance, &switchbalance_parse},
 	{"lidbalance", &help_lidbalance, &lidbalance_parse},
 	{"dump_conf", &help_dump_conf, &dump_conf_parse},
+	{"update_desc", &help_update_desc, &update_desc_parse},
 	{"version", &help_version, &version_parse},
 #ifdef ENABLE_OSM_PERF_MGR
 	{"perfmgr", &help_perfmgr, &perfmgr_parse},

opensm/opensm/osm_state_mgr.c +3 ~~-13~~

index 6296f21..01dfba6 100644
@@ -510,11 +510,7 @@ static void query_sm_info(cl_map_item_t * item, void *cxt)
 			ib_get_err_str(ret));
 }
 
-/**********************************************************************
- During a light sweep, check each node to see if the node description
- is valid and if not issue a ND query.
-**********************************************************************/
-static void state_mgr_get_node_desc(IN cl_map_item_t * obj, IN void *context)
+static void state_mgr_update_node_desc(IN cl_map_item_t * obj, IN void *context)
 {
 	osm_madw_context_t mad_context;
 	osm_node_t *p_node = (osm_node_t *) obj;

@@ -527,14 +523,8 @@ static void state_mgr_get_node_desc(IN cl_map_item_t * obj, IN void *context)
 
 	CL_ASSERT(p_node);
 
-	if (p_node->print_desc
-	    && strcmp(p_node->print_desc, OSM_NODE_DESC_UNKNOWN))
-		/* if ND is valid, do nothing */
-		goto exit;
-
-	OSM_LOG(sm->p_log, OSM_LOG_ERROR,
-		"ERR 3319: Unknown node description for node GUID "
-		"0x%016" PRIx64 ".  Reissuing ND query\n",
+	OSM_LOG(sm->p_log, OSM_LOG_DEBUG,
+		"Updating NodeDesc for 0x%016" PRIx64 "\n",
 		cl_ntoh64(osm_node_get_node_guid(p_node)));
 
 	/* get a physp to request from. */

@@ -563,6 +553,38 @@ exit:
 	OSM_LOG_EXIT(sm->p_log);
 }

Quoted 5 lines

opensm/opensm/osm_state_mgr.c exit: · line 558
Quoted 3 lines
opensm/opensm/osm_state_mgr.c +3 -0
+void osm_update_node_desc(IN osm_sm_t *sm)
+{
+	CL_PLOCK_ACQUIRE(sm->p_lock);
Wouldn't you need exclusive locking here (CL_PLOCK_EXCL_ACQUIRE())?

no, since I'm only sending ND request. update is done when response arrives.

Quoted 19 lines

opensm/opensm/osm_state_mgr.c +18 -0

+	cl_qmap_apply_func(&sm->p_subn->node_guid_tbl, state_mgr_update_node_desc,
+			   sm);
+	CL_PLOCK_RELEASE(sm->p_lock);
+}
+
+/**********************************************************************
+ During a light sweep, check each node to see if the node description
+ is valid and if not issue a ND query.
+**********************************************************************/
+static void state_mgr_get_node_desc(IN cl_map_item_t * obj, IN void *context)
+{
+	osm_node_t *p_node = (osm_node_t *) obj;
+	osm_sm_t *sm = context;
+
+	OSM_LOG_ENTER(sm->p_log);
+
+	CL_ASSERT(p_node);
+

Quoted 8 lines

opensm/opensm/osm_state_mgr.c exit: · line 581
Quoted 5 lines
opensm/opensm/osm_state_mgr.c +5 -0
+	if (p_node->print_desc
+	    && strcmp(p_node->print_desc, OSM_NODE_DESC_UNKNOWN))
+		/* if ND is valid, do nothing */
+		goto exit;
+
Seems that "ERR 3319:" error message was dropped forever. Any reason for doing this?

dropped by mistake. will add this to patch v2.

Eli

Quoted 15 lines

Sasha

Quoted 12 lines

opensm/opensm/osm_state_mgr.c +6 -0

+	state_mgr_update_node_desc(obj, context);
+
+exit:
+	OSM_LOG_EXIT(sm->p_log);
+}
+
 /**********************************************************************
  Initiates a lightweight sweep of the subnet.
  Used during normal sweeps after the subnet is up.

Signature · 1 line

1.5.5

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

Eli Dorfman (Voltaire)

author

2010-01-13 15:34 UTC #4 lore raw

PATCH v2 1/1 opensm: Add update_desc command to opensm console

In reply to Sasha Khapyorsky

Signed-off-by Eli Dorfman <elid-smomgflXvOZWk0Htik3J/[email protected]>

---

v2: Revert error (ERR 3319) message deletion

 opensm/opensm/osm_console.c   |   16 ++++++++++++
 opensm/opensm/osm_state_mgr.c |   53 +++++++++++++++++++++++++++++++----------
 2 files changed, 56 insertions(+), 13 deletions(-)

opensm/opensm/osm_console.c +16 -0

index 206e7f7..e90cb99 100644
@@ -56,6 +56,8 @@
 #include <opensm/osm_perfmgr.h>
 #include <opensm/osm_subnet.h>
 
+extern void osm_update_node_desc(IN osm_sm_t *sm);
+
 struct command {
 	char *name;
 	void (*help_function) (FILE * out, int detail);

@@ -207,6 +209,14 @@ static void help_dump_conf(FILE *out, int detail)
 	}
 }
 
+static void help_update_desc(FILE *out, int detail)
+{
+	fprintf(out, "update_desc\n");
+	if (detail) {
+		fprintf(out, "update node description for all nodes\n");
+	}
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void help_perfmgr(FILE * out, int detail)
 {

@@ -1134,6 +1144,11 @@ static void dump_conf_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 	osm_subn_output_conf(out, &p_osm->subn.opt);
 }
 
+static void update_desc_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
+{
+	osm_update_node_desc(&p_osm->sm);
+}
+
 #ifdef ENABLE_OSM_PERF_MGR
 static void perfmgr_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 {

@@ -1326,6 +1341,7 @@ static const struct command console_cmds[] = {
 	{"switchbalance", &help_switchbalance, &switchbalance_parse},
 	{"lidbalance", &help_lidbalance, &lidbalance_parse},
 	{"dump_conf", &help_dump_conf, &dump_conf_parse},
+	{"update_desc", &help_update_desc, &update_desc_parse},
 	{"version", &help_version, &version_parse},
 #ifdef ENABLE_OSM_PERF_MGR
 	{"perfmgr", &help_perfmgr, &perfmgr_parse},

opensm/opensm/osm_state_mgr.c +40 ~~-13~~

71 lines not on this page · raw message

Signature · 6 lines

1.5.5

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

A message that is not in the archive (sent off-list or not yet received): [email protected]

Sasha Khapyorsky

Applied

2010-01-13 17:26 UTC #6 lore raw

Quoted 8 lines · Eli Dorfman (Voltaire)

Sometimes nodes are discovered by the opensm with their default name before real host name was resolves and updated in the HCA. Since Mellanox ConnectX HCA do not support NodeDescription change trap (as part of trap 144) we need an option to refresh NodeDescription from console.
Signed-off-by Eli Dorfman <elid-smomgflXvOZWk0Htik3J/[email protected]>

Applied. Thanks.

Sasha
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html