Skip to contentSkip to conversation

opensm/osm_state_mgr.c: Don't rely on PortInfo:PortState for base SP0

PATCH In review Hal Rosenstock 2 participants 6 messages Apr 18, 2011 – Apr 27, 2011 4 messages not in the archive InfiniBand

lore
Share & tools
lore

Hal Rosenstock

author
#0 lore raw

PATCH 1/2 opensm/osm_state_mgr.c: Don't rely on PortInfo:PortState for base SP0

For base SP0, PortState in SM PortInfo attribute is not used and base SP0 is always "active".

Signed-off-by Hal Rosenstock <hal-VPRAkNaXOzVWk0Htik3J/[email protected]>

---

opensm/osm_state_mgr.c +5 -1
index 2e1ef94..dd308f2 100644
@@ -352,7 +352,11 @@ static boolean_t state_mgr_is_sm_port_down(IN osm_sm_t * sm)
 
 	CL_ASSERT(p_physp);
 
-	state = osm_physp_get_port_state(p_physp);
+	if (p_port->p_node->sw &&
+	    !ib_switch_info_is_enhanced_port0(&p_port->p_node->sw->switch_info))
+		state = IB_LINK_ACTIVE;	/* base SP0 */
+	else
+		state = osm_physp_get_port_state(p_physp);
 	CL_PLOCK_RELEASE(sm->p_lock);
 
 Exit:

--
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): 4DAC3CCC.4090101-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org

#2 lore raw

Hi Hal,

Quoted 3 lines · Hal Rosenstock

For base SP0, PortState in SM PortInfo attribute is not used and base SP0 is always "active".

How SM can be attached to base SP0?
Moreover, if during discovery we found that SM is attached to base SP0, don't we want to act like the port is down?

Quoted 21 lines
Signed-off-by Hal Rosenstock <hal-VPRAkNaXOzVWk0Htik3J/[email protected]>

---

opensm/osm_state_mgr.c +5 -2
index 2e1ef94..dd308f2 100644
@@ -352,7 +352,11 @@ static boolean_t state_mgr_is_sm_port_down(IN osm_sm_t * sm)
 
 	CL_ASSERT(p_physp);
 
-	state = osm_physp_get_port_state(p_physp);
+	if (p_port->p_node->sw &&
+	    !ib_switch_info_is_enhanced_port0(&p_port->p_node->sw->switch_info))
+		state = IB_LINK_ACTIVE;	/* base SP0 */
+	else
+		state = osm_physp_get_port_state(p_physp);
 	CL_PLOCK_RELEASE(sm->p_lock);
 
 Exit:
--
Quoted 3 lines

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

-- Alex
--
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): 20110420141713.GD28761-iQai9MGU/dyyaiaB+Ve85laTQe2KTcn/@public.gmane.org

author
#4 lore raw

Hi Alex,

Quoted 8 lines · Alex Netes

Hi Hal,

Quoted 3 lines · Hal Rosenstock

For base SP0, PortState in SM PortInfo attribute is not used and base SP0 is always "active".

How SM can be attached to base SP0?

What precludes this ? AFAIK SM can run on any switch port 0 whether enhanced or base.

Quoted 2 lines

Moreover, if during discovery we found that SM is attached to base SP0, don't we want to act like the port is down?

For base SP0, PortState and PortPhysicalState are "not used" in the SM PortInfo attribute. That means they're "vendor" specific and they should not be relied upon to be valid.

The point of this patch is that base SP0 is always up regardless of what PortInfo for that port says.

If SM wasn't allowed to run on base SP0, it would be a different patch:
to prohibit such a port from being used at all as the SM port so the failure would be earlier in the code flow when the SM tries to bind to such a port.

-- Hal

Quoted 19 lines
Quoted 19 lines
Signed-off-by Hal Rosenstock <hal-VPRAkNaXOzVWk0Htik3J/[email protected]>

---

opensm/osm_state_mgr.c +5 -1
index 2e1ef94..dd308f2 100644
@@ -352,7 +352,11 @@ static boolean_t state_mgr_is_sm_port_down(IN osm_sm_t * sm)
 
 	CL_ASSERT(p_physp);
 
-	state = osm_physp_get_port_state(p_physp);
+	if (p_port->p_node->sw &&
+	    !ib_switch_info_is_enhanced_port0(&p_port->p_node->sw->switch_info))
+		state = IB_LINK_ACTIVE;	/* base SP0 */
+	else
+		state = osm_physp_get_port_state(p_physp);
 	CL_PLOCK_RELEASE(sm->p_lock);
Quoted 5 lines

opensm/osm_state_mgr.c static boolean_t state_mgr_is_sm_port_down(IN osm_sm_t * sm)

Quoted 2 lines
opensm/osm_state_mgr.c +0 -1
 Exit:
--

-- Alex

--
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): 4DAEF103.1000706-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org

#6 lore raw
Quoted 7 lines · Hal Rosenstock

Hi Alex,

Quoted 4 lines · Alex Netes

Hi Hal,

Quoted 1 line · Hal Rosenstock
Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

For base SP0, PortState in SM PortInfo attribute is not used and base SP0 is always "active".

How SM can be attached to base SP0?

What precludes this ? AFAIK SM can run on any switch port 0 whether enhanced or base.

I'm probably missing something. Base Switch Port 0 - A Switch Port 0 which is not an Endport (IBTA p71).
If you look further, Endport defined as a Port which can be a destination of LID-routed communication within the same Subnet as the sender.

How can SM be attached to a port which isn't an Endport?

-- Alex
--
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): 20110421122243.GA2137-iQai9MGU/dyyaiaB+Ve85laTQe2KTcn/@public.gmane.org

author
#8 lore raw
Quoted 11 lines · Alex Netes
Quoted 10 lines · Hal Rosenstock

Hi Alex,

Quoted 7 lines · Alex Netes

Hi Hal,

Quoted 3 lines · Hal Rosenstock

For base SP0, PortState in SM PortInfo attribute is not used and base SP0 is always "active".

Quoted 8 lines
Quoted 5 lines
Quoted 1 line

How SM can be attached to base SP0?

What precludes this ? AFAIK SM can run on any switch port 0 whether enhanced or base.

I'm probably missing something. Base Switch Port 0 - A Switch Port 0 which is not an Endport (IBTA p71).

The glossary is only informative (and not compliance) and has a number of such issues.

You can see that the Endport definition on p.73 contradicts this:
"All Channel Adapter ports on the subnet are
endports of that subnet, as is Port 0 of each Switch in the subnet. Switch ports other than Port 0 may not be endports."

A spec comment (RefID 4352) was entered quite some time ago on this to change this to something like:

A <Switch Port 0> with a set of basic capabilities (defined in this specification). That set is less than the capabilities of an <Enhanced Switch Port 0>, and also less than the minimum set of capabilities required of a <Target Channel Adapter>.

Quoted 2 lines

If you look further, Endport defined as a Port which can be a destination of LID-routed communication within the same Subnet as the sender.

Actually this is imprecise as well as the phrase "within the same Subnet as the sender" should be eliminated.

Quoted 1 line

How can SM be attached to a port which isn't an Endport?

Base SP0 does have a LID (look at PortInfo:LID Used By column is X'd).

-- Hal

Quoted 6 lines

-- Alex
--
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

--
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