Skip to contentSkip to conversation

opensm/osm_console.c: Support portstatus output for incorrect link speeds

PATCH In review Albert Chu 2 participants 3 messages Feb 21, 2013 – Feb 22, 2013 2 messages not in the archive InfiniBand

lore
Share & tools
lore

Albert Chu

author
#0 lore raw

PATCH 1/1 opensm/osm_console.c: Support portstatus output for incorrect link speeds

Usually coinciding with use of the force_link_speed configuration option, some link speeds can be observed to be incorrect. For example, a link may be active at QDR even if SDR & DDR speeds are the only ones enabled.

Under these scenarios, the portstatus may incorrectly output port status, confusing users.

A new port output status of "incorrect speed" differentiates ports in this state from other states, so as to not confuse users.

---
 opensm/osm_console.c |   18 ++++++++++++++++--
 1 files changed, 16 insertions(+), 2 deletions(-)
opensm/osm_console.c +16 -2
index 600007c..06c9360 100644
@@ -712,6 +712,8 @@ typedef struct {
 	uint64_t ports_fdr;
 	uint64_t ports_edr;
 	uint64_t ports_unknown_speed;
+	uint64_t ports_incorrect_speed;
+	port_report_t *incorrect_speed_ports;
 	uint64_t ports_reduced_speed;
 	port_report_t *reduced_speed_ports;
 } fabric_stats_t;
@@ -773,7 +775,14 @@ static void __get_stats(cl_map_item_t * const p_map_item, void *context)
 			fs->ports_reduced_width++;
 		}
 
-		if ((enabled_speed ^ active_speed) > active_speed) {
+		/* incorrect speed usually due to problems with force_link_speed */
+		if (!(active_speed & enabled_speed)) {
+			__tag_port_report(&(fs->incorrect_speed_ports),
+					  cl_ntoh64(node->node_info.node_guid),
+					  port, node->print_desc);
+			fs->ports_incorrect_speed++;
+		}
+		else if ((enabled_speed ^ active_speed) > active_speed) {
 			__tag_port_report(&(fs->reduced_speed_ports),
 					  cl_ntoh64(node->node_info.node_guid),
 					  port, node->print_desc);
@@ -933,13 +942,18 @@ static void portstatus_parse(char **p_last, osm_opensm_t * p_osm, FILE * out)
 		fprintf(out, "   %" PRIu64 " at 25.78125 Gbps\n", fs.ports_edr);
 
 	if (fs.ports_disabled + fs.ports_reduced_speed + fs.ports_reduced_width
-	    > 0) {
+	    + fs.ports_incorrect_speed > 0) {
 		fprintf(out, "\nPossible issues:\n");
 	}
 	if (fs.ports_disabled) {
 		fprintf(out, "   %" PRIu64 " disabled\n", fs.ports_disabled);
 		__print_port_report(out, fs.disabled_ports);
 	}
+	if (fs.ports_incorrect_speed) {
+		fprintf(out, "   %" PRIu64 " with incorrect speed\n",
+			fs.ports_incorrect_speed);
+		__print_port_report(out, fs.incorrect_speed_ports);
+	}
 	if (fs.ports_reduced_speed) {
 		fprintf(out, "   %" PRIu64 " with reduced speed\n",
 			fs.ports_reduced_speed);
Signature · 8 lines
1.7.1



--
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): 1361489963.24405.68.camel-akkeaxHeDKRliZ7u+bvwcg@public.gmane.org

#2 lore raw

Hi Al,

Quoted 3 lines · Albert Chu

Usually coinciding with use of the force_link_speed configuration option, some link speeds can be observed to be incorrect. For example, a link may be active at QDR even if SDR & DDR speeds are the only ones enabled.

Are you referring to the fact that after setting the enabled speed, the IB spec does not mandate a link renegotiation and hence one may not have occurred so the enabled speed can be out of whack with the active speed ?

Also, isn't there a similar issue with extended speeds (and force_link_speed_ext) ? If so, should that also be part of this patch ?

Same issue also is there for link width as it works the same way as link speed. Maybe that too is yet another patch.

What do you think ?

-- Hal

Quoted 7 lines

Under these scenarios, the portstatus may incorrectly output port status, confusing users.

A new port output status of "incorrect speed" differentiates ports in this state from other states, so as to not confuse users.

<snip...>
--
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): 5127C824.2090108-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org

author
#4 lore raw
Quoted 10 lines · Hal Rosenstock

Hi Al,

Quoted 3 lines · Albert Chu

Usually coinciding with use of the force_link_speed configuration option, some link speeds can be observed to be incorrect. For example, a link may be active at QDR even if SDR & DDR speeds are the only ones enabled.

Are you referring to the fact that after setting the enabled speed, the IB spec does not mandate a link renegotiation and hence one may not have occurred so the enabled speed can be out of whack with the active speed ?

Now that I re-read my patch comment, it wasn't the clearest.

What you describe above very well could have occurred or maybe it was just bad hardware. The key was that ports were running at speeds (e.g. QDR) that were not enabled (e.g. only SDR & DDR) and the current opensm console was reporting invalid/incorrect error messages.

Quoted 7 lines

Also, isn't there a similar issue with extended speeds (and force_link_speed_ext) ? If so, should that also be part of this patch ?

Same issue also is there for link width as it works the same way as link speed. Maybe that too is yet another patch.

What do you think ?

Yeah, I should have done those as well. I was only solving the one particular bug found. I'll rework the patch.

Al

Quoted 11 lines

-- Hal

Quoted 7 lines

Under these scenarios, the portstatus may incorrectly output port status, confusing users.

A new port output status of "incorrect speed" differentiates ports in this state from other states, so as to not confuse users.

<snip...>

Signature · 11 lines
Albert Chu
[email protected]
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory


--
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