Skip to contentSkip to conversation

opensm: Add initial support for optimized SLtoVLMappingTable programming

PATCH Sasha Khapyorsky 3 participants 5 messages Oct 30, 2009 – Dec 1, 2009 3 messages not in the archive InfiniBand

lore
Share & tools
lore

Sasha Khapyorsky

author
#2 lore raw
In reply to a message not in the archive
Quoted 1 line · Hal Rosenstock

Implementation description would be very useful.

What does "initial support" mean?

Quoted 54 lines
Signed-off-by Hal Rosenstock <[email protected]>

---

opensm/include/opensm/osm_subnet.h +6 -0
index 6c20de8..8443763 100644
@@ -4,6 +4,7 @@
  * Copyright (c) 1996-2003 Intel Corporation. All rights reserved.
  * Copyright (c) 2008 Xsigo Systems Inc.  All rights reserved.
  * Copyright (c) 2009 System Fabric Works, Inc. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU

@@ -204,6 +205,7 @@ typedef struct osm_subn_opt {
 	boolean_t daemon;
 	boolean_t sm_inactive;
 	boolean_t babbling_port_policy;
+	boolean_t use_optimized_slvl;
 	osm_qos_options_t qos_options;
 	osm_qos_options_t qos_ca_options;
 	osm_qos_options_t qos_sw0_options;

@@ -428,6 +430,10 @@ typedef struct osm_subn_opt {
 *	babbling_port_policy
 *		OpenSM will enforce its "babbling" port policy.
 *
+*	use_optimized_slvl
+*		Use optimized SLtoVLMappingTable programming if
+*		device indicates it supports this.
+*
 *	perfmgr
 *		Enable or disable the performance manager
 *
opensm/opensm/osm_qos.c +2 -1
index e3dfb58..592e082 100644
@@ -1,5 +1,6 @@
 /*
  * Copyright (c) 2006-2008 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU

@@ -150,7 +151,7 @@ static ib_api_status_t vlarb_update(osm_sm_t * sm, osm_physp_t * p,
 
 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,
+					  unsigned optimize, unsigned force_update,
 					  const ib_slvl_table_t * sl2vl_table)
 {
 	osm_madw_context_t context;

opensm/opensm/osm_qos.c static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p, · line 184

Quoted 7 lines
opensm/opensm/osm_qos.c +4 -0
@@ -177,10 +178,18 @@ static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p,
 	    !memcmp(p_tbl, &tbl, sizeof(tbl)))
 		return IB_SUCCESS;
 
+	/* both input port and output port wildcarded */
+	if (optimize && (in_port != 1 || out_port != 1))
+		return IB_SUCCESS;
+

This function (sl2vl_update_table()) is called for each in and out ports combination. I think that instead of calling it N * N times and filtering out the case 'in == out == 1' you just need it once somewhere at higher layer.

Quoted 33 lines
opensm/opensm/osm_qos.c +10 -3
 	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;
+	if (optimize)
+		/* both input port and output port wildcarded */
+		attr_mod = 0x30000;
+	else
+		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),

@@ -189,14 +198,17 @@ static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p,
 
 static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
 				    osm_physp_t * p, uint8_t port_num,
-				    unsigned force_update,
+				    unsigned optimize, unsigned force_update,
 				    const struct qos_config *qcfg)
 {
 	ib_api_status_t status;
 	uint8_t i, num_ports;
 	osm_physp_t *p_physp;
+	osm_node_t *p_node;
+	unsigned optimizesl2vl = 0;
 
-	if (osm_node_get_type(osm_physp_get_node_ptr(p)) == IB_NODE_TYPE_SWITCH) {
+	p_node = osm_physp_get_node_ptr(p);
+	if (osm_node_get_type(p_node) == IB_NODE_TYPE_SWITCH) {
 		if (ib_port_info_get_vl_cap(&p->port_info) == 1) {
 			/* Check port 0's capability mask */
 			p_physp = p_port->p_physp;

opensm/opensm/osm_qos.c static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port, · line 221

Quoted 6 lines
opensm/opensm/osm_qos.c +2 -1
@@ -205,7 +217,8 @@ static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
 			     capability_mask & IB_PORT_CAP_HAS_SL_MAP))
 				return IB_SUCCESS;
 		}
-		num_ports = osm_node_get_num_physp(osm_physp_get_node_ptr(p));
+		num_ports = osm_node_get_num_physp(p_node);
+		optimizesl2vl = ib_switch_info_get_opt_sl2vlmapping(&p_node->sw->switch_info) & optimize;

It should be once per switch, not for each switch port.

Quoted 94 lines
opensm/opensm/osm_qos.c +8 -5
 	} else {
 		if (!(p->port_info.capability_mask & IB_PORT_CAP_HAS_SL_MAP))
 			return IB_SUCCESS;

@@ -213,8 +226,8 @@ static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
 	}
 
 	for (i = 0; i < num_ports; i++) {
-		status = sl2vl_update_table(sm, p, i, port_num, force_update,
-					    &qcfg->sl2vl);
+		status = sl2vl_update_table(sm, p, i, port_num, optimizesl2vl,
+					    force_update, &qcfg->sl2vl);
 		if (status != IB_SUCCESS)
 			return status;
 	}

@@ -224,7 +237,8 @@ static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
 
 static int qos_physp_setup(osm_log_t * p_log, osm_sm_t * sm,
 			   osm_port_t * p_port, osm_physp_t * p,
-			   uint8_t port_num, unsigned force_update,
+			   uint8_t port_num, unsigned optimize,
+			   unsigned force_update,
 			   const struct qos_config *qcfg)
 {
 	ib_api_status_t status;

@@ -245,7 +259,8 @@ static int qos_physp_setup(osm_log_t * p_log, osm_sm_t * sm,
 	}
 
 	/* setup SL2VL tables */
-	status = sl2vl_update(sm, p_port, p, port_num, force_update, qcfg);
+	status = sl2vl_update(sm, p_port, p, port_num, optimize, force_update,
+			      qcfg);
 	if (status != IB_SUCCESS) {
 		OSM_LOG(p_log, OSM_LOG_ERROR, "ERR 6203 : "
 			"failed to update SL2VLMapping tables "

@@ -307,6 +322,7 @@ int osm_qos_setup(osm_opensm_t * p_osm)
 				    p_osm->subn.need_update;
 				if (qos_physp_setup(&p_osm->log, &p_osm->sm,
 						    p_port, p_physp, i,
+						    p_osm->subn.opt.use_optimized_slvl,
 						    force_update, &swe_config))
 					ret = -1;
 			}

@@ -327,7 +343,7 @@ int osm_qos_setup(osm_opensm_t * p_osm)
 
 		force_update = p_physp->need_update || p_osm->subn.need_update;
 		if (qos_physp_setup(&p_osm->log, &p_osm->sm, p_port, p_physp,
-				    0, force_update, cfg))
+				    0, 0, force_update, cfg))
 			ret = -1;
 	}
opensm/opensm/osm_slvl_map_rcv.c +8 -2
index 9c37442..67c71bd 100644
@@ -2,6 +2,7 @@
  * Copyright (c) 2004-2008 Voltaire, Inc. All rights reserved.
  * Copyright (c) 2002-2005 Mellanox Technologies LTD. All rights reserved.
  * Copyright (c) 1996-2003 Intel Corporation. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU

@@ -72,7 +73,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
 	osm_slvl_context_t *p_context;
 	ib_net64_t port_guid;
 	ib_net64_t node_guid;
-	uint8_t out_port_num, in_port_num;
+	uint32_t attr_mod;
+	uint8_t out_port_num, in_port_num, startinport, startoutport,
+		endinport, endoutport;
 
 	CL_ASSERT(sm);
 

@@ -111,6 +114,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
 		    (uint8_t) cl_ntoh32(p_smp->attr_mod & 0xFF000000);
 		in_port_num =
 		    (uint8_t) cl_ntoh32((p_smp->attr_mod & 0x00FF0000) << 8);
+		attr_mod = cl_ntoh32(p_smp->attr_mod);
+		if (attr_mod & 0x30000)
+			goto opt_sl2vl;
 		p_physp = osm_node_get_physp_ptr(p_node, out_port_num);
 	} else {
 		p_physp = p_port->p_physp;

@@ -123,7 +129,7 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
 	   all we want is to update the subnet.
 	 */
 	OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
-		"Got SLtoVL get response in_port_num %u out_port_num %u with "
+		"Received SLtoVL GetResp in_port_num %u out_port_num %u with "
 		"GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64 ", TID 0x%"
 		PRIx64 "\n", in_port_num, out_port_num, cl_ntoh64(port_guid),
 		cl_ntoh64(node_guid), cl_ntoh64(p_smp->trans_id));

opensm/opensm/osm_slvl_map_rcv.c void osm_slvl_rcv_process(IN void *context, IN void *p_data) · line 153

Quoted 6 lines
opensm/opensm/osm_slvl_map_rcv.c +3 -0
@@ -142,6 +148,39 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
 				out_port_num, p_slvl_tbl, OSM_LOG_DEBUG);
 
 	osm_physp_set_slvl_tbl(p_physp, p_slvl_tbl, in_port_num);
+	goto Exit;
+
+opt_sl2vl:

I think that you need to use functions here.

Quoted 15 lines
opensm/opensm/osm_slvl_map_rcv.c +15 -0
+	OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
+		"Got optimized SLtoVL get response in_port_num %u out_port_num "
+		"%u with GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64
+		", TID 0x%" PRIx64 "\n", in_port_num, out_port_num,
+		cl_ntoh64(port_guid), cl_ntoh64(node_guid),
+		cl_ntoh64(p_smp->trans_id));
+
+	osm_dump_slvl_map_table(sm->p_log, port_guid, in_port_num,
+				out_port_num, p_slvl_tbl, OSM_LOG_DEBUG);
+
+	if (attr_mod & 0x10000) {
+		startoutport = ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) ? 0 : 1;
+		endoutport = osm_node_get_num_physp(p_node);
+	} else
+		endoutport = startoutport = out_port_num;

opensm/opensm/osm_slvl_map_rcv.c void osm_slvl_rcv_process(IN void *context, IN void *p_data) · line 178

Quoted 10 lines
opensm/opensm/osm_slvl_map_rcv.c +10 -0
+	if (attr_mod & 0x20000) {
+		startinport = ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) ? 0 : 1;
+		endinport = osm_node_get_num_physp(p_node);
+	} else
+		endinport = startinport = in_port_num;
+
+	for (out_port_num = startoutport; out_port_num < endoutport;
+	     out_port_num++) {
+		p_physp = osm_node_get_physp_ptr(p_node, out_port_num);
+		if (!p_physp)

When could this be NULL?

Sasha

Quoted 43 lines
opensm/opensm/osm_slvl_map_rcv.c +5 -0
+			continue;
+		for (in_port_num = startinport; in_port_num < endinport;
+		     in_port_num++)
+			osm_physp_set_slvl_tbl(p_physp, p_slvl_tbl, in_port_num);
+	}
 
 Exit:
 	cl_plock_release(sm->p_lock);
opensm/opensm/osm_subnet.c +4 -1
index 0d11811..540165a 100644
@@ -4,6 +4,7 @@
  * Copyright (c) 1996-2003 Intel Corporation. All rights reserved.
  * Copyright (c) 2008 Xsigo Systems Inc.  All rights reserved.
  * Copyright (c) 2009 System Fabric Works, Inc. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU

@@ -352,6 +353,7 @@ static const opt_rec_t opt_tbl[] = {
 	{ "daemon", OPT_OFFSET(daemon), opts_parse_boolean, NULL, 0 },
 	{ "sm_inactive", OPT_OFFSET(sm_inactive), opts_parse_boolean, NULL, 1 },
 	{ "babbling_port_policy", OPT_OFFSET(babbling_port_policy), opts_parse_boolean, NULL, 1 },
+	{"use_optimized_slvl", OPT_OFFSET(use_optimized_slvl), opts_parse_boolean, NULL, 1 },
 #ifdef ENABLE_OSM_PERF_MGR
 	{ "perfmgr", OPT_OFFSET(perfmgr), opts_parse_boolean, NULL, 0 },
 	{ "perfmgr_redir", OPT_OFFSET(perfmgr_redir), opts_parse_boolean, NULL, 0 },

@@ -715,6 +717,7 @@ void osm_subn_set_default_opt(IN osm_subn_opt_t * const p_opt)
 	p_opt->daemon = FALSE;
 	p_opt->sm_inactive = FALSE;
 	p_opt->babbling_port_policy = FALSE;
+	p_opt->use_optimized_slvl = FALSE;
 #ifdef ENABLE_OSM_PERF_MGR
 	p_opt->perfmgr = FALSE;
 	p_opt->perfmgr_redir = TRUE;

@@ -1501,10 +1504,13 @@ int osm_subn_output_conf(FILE *out, IN osm_subn_opt_t *const p_opts)
 		"# SM Inactive\n"
 		"sm_inactive %s\n\n"
 		"# Babbling Port Policy\n"
-		"babbling_port_policy %s\n\n",
+		"babbling_port_policy %s\n\n"

opensm/opensm/osm_subnet.c int osm_subn_output_conf(FILE *out, IN osm_subn_opt_t *const p_opts) · line 1516

Quoted 10 lines
opensm/opensm/osm_subnet.c +4 -1
+		"# Use Optimized SLtoVLMapping programming if supported by device\n"
+		"use_optimized_slvl %s\n\n",
 		p_opts->daemon ? "TRUE" : "FALSE",
 		p_opts->sm_inactive ? "TRUE" : "FALSE",
-		p_opts->babbling_port_policy ? "TRUE" : "FALSE");
+		p_opts->babbling_port_policy ? "TRUE" : "FALSE",
+		p_opts->use_optimized_slvl ? "TRUE" : "FALSE");
 
 #ifdef ENABLE_OSM_PERF_MGR
 	fprintf(out,

--
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

#3 lore raw
Quoted 6 lines · Sasha Khapyorsky
Quoted 1 line · Hal Rosenstock

Implementation description would be very useful.

What does "initial support" mean?

It means there's more to come in terms of using OptimizedSLtoVLMappingProgramming. This is the simplest use/introduction of this optional feature.

Quoted 57 lines
Quoted 57 lines
Signed-off-by Hal Rosenstock <[email protected]>

---

opensm/include/opensm/osm_subnet.h +6 -0
index 6c20de8..8443763 100644
@@ -4,6 +4,7 @@
   * Copyright (c) 1996-2003 Intel Corporation. All rights reserved.
   * Copyright (c) 2008 Xsigo Systems Inc.  All rights reserved.
   * Copyright (c) 2009 System Fabric Works, Inc. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
   *
   * This software is available to you under a choice of one of two
   * licenses.  You may choose to be licensed under the terms of the GNU

@@ -204,6 +205,7 @@ typedef struct osm_subn_opt {
       boolean_t daemon;
       boolean_t sm_inactive;
       boolean_t babbling_port_policy;
+     boolean_t use_optimized_slvl;
       osm_qos_options_t qos_options;
       osm_qos_options_t qos_ca_options;
       osm_qos_options_t qos_sw0_options;

@@ -428,6 +430,10 @@ typedef struct osm_subn_opt {
  *    babbling_port_policy
  *            OpenSM will enforce its "babbling" port policy.
  *
+*    use_optimized_slvl
+*            Use optimized SLtoVLMappingTable programming if
+*            device indicates it supports this.
+*
  *    perfmgr
  *            Enable or disable the performance manager
  *
opensm/opensm/osm_qos.c +2 -1
index e3dfb58..592e082 100644
@@ -1,5 +1,6 @@
  /*
   * Copyright (c) 2006-2008 Voltaire, Inc. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
   *
   * This software is available to you under a choice of one of two
   * licenses.  You may choose to be licensed under the terms of the GNU

@@ -150,7 +151,7 @@ static ib_api_status_t vlarb_update(osm_sm_t * sm, osm_physp_t * p,
 
  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,
+                                       unsigned optimize, unsigned force_update,
                                         const ib_slvl_table_t * sl2vl_table)
  {
       osm_madw_context_t context;

@@ -177,10 +178,18 @@ static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p,
           !memcmp(p_tbl, &tbl, sizeof(tbl)))
               return IB_SUCCESS;
Quoted 9 lines

opensm/opensm/osm_qos.c static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p, · line 184

Quoted 4 lines
opensm/opensm/osm_qos.c +4 -0
+     /* both input port and output port wildcarded */
+     if (optimize && (in_port != 1 || out_port != 1))
+             return IB_SUCCESS;
+

This function (sl2vl_update_table()) is called for each in and out ports combination. I think that instead of calling it N * N times and filtering out the case 'in == out == 1' you just need it once somewhere at higher layer.

I was trying to minimize the changes to the current code flow. It can be done the way you propose but that will cause larger changes. Are you sure about this direction ?

Quoted 26 lines
Quoted 26 lines
opensm/opensm/osm_qos.c +8 -2
       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;
+     if (optimize)
+             /* both input port and output port wildcarded */
+             attr_mod = 0x30000;
+     else
+             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),

@@ -189,14 +198,17 @@ static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p,
 
  static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
                                   osm_physp_t * p, uint8_t port_num,
-                                 unsigned force_update,
+                                 unsigned optimize, unsigned force_update,
                                   const struct qos_config *qcfg)
  {
       ib_api_status_t status;
       uint8_t i, num_ports;
       osm_physp_t *p_physp;
+     osm_node_t *p_node;
+     unsigned optimizesl2vl = 0;
Quoted 15 lines

opensm/opensm/osm_qos.c static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port, · line 221

Quoted 13 lines
opensm/opensm/osm_qos.c +4 -2
-     if (osm_node_get_type(osm_physp_get_node_ptr(p)) == IB_NODE_TYPE_SWITCH) {
+     p_node = osm_physp_get_node_ptr(p);
+     if (osm_node_get_type(p_node) == IB_NODE_TYPE_SWITCH) {
               if (ib_port_info_get_vl_cap(&p->port_info) == 1) {
                       /* Check port 0's capability mask */
                       p_physp = p_port->p_physp;

@@ -205,7 +217,8 @@ static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
                            capability_mask & IB_PORT_CAP_HAS_SL_MAP))
                               return IB_SUCCESS;
               }
-             num_ports = osm_node_get_num_physp(osm_physp_get_node_ptr(p));
+             num_ports = osm_node_get_num_physp(p_node);
+             optimizesl2vl = ib_switch_info_get_opt_sl2vlmapping(&p_node->sw->switch_info) & optimize;

It should be once per switch, not for each switch port.

Sure; if (still) needed here (might depend on the approach) is a tradeoff of a simple local operation v. an additional passed in parameter.

Quoted 96 lines
Quoted 96 lines
opensm/opensm/osm_qos.c +6 -4
       } else {
               if (!(p->port_info.capability_mask & IB_PORT_CAP_HAS_SL_MAP))
                       return IB_SUCCESS;

@@ -213,8 +226,8 @@ static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
       }
 
       for (i = 0; i < num_ports; i++) {
-             status = sl2vl_update_table(sm, p, i, port_num, force_update,
-                                         &qcfg->sl2vl);
+             status = sl2vl_update_table(sm, p, i, port_num, optimizesl2vl,
+                                         force_update, &qcfg->sl2vl);
               if (status != IB_SUCCESS)
                       return status;
       }

@@ -224,7 +237,8 @@ static ib_api_status_t sl2vl_update(osm_sm_t * sm, osm_port_t * p_port,
 
  static int qos_physp_setup(osm_log_t * p_log, osm_sm_t * sm,
                          osm_port_t * p_port, osm_physp_t * p,
-                        uint8_t port_num, unsigned force_update,
+                        uint8_t port_num, unsigned optimize,
+                        unsigned force_update,
                          const struct qos_config *qcfg)
  {
       ib_api_status_t status;

@@ -245,7 +259,8 @@ static int qos_physp_setup(osm_log_t * p_log, osm_sm_t * sm,
       }
 
       /* setup SL2VL tables */
-     status = sl2vl_update(sm, p_port, p, port_num, force_update, qcfg);
+     status = sl2vl_update(sm, p_port, p, port_num, optimize, force_update,
+                           qcfg);
       if (status != IB_SUCCESS) {

              OSM_LOG(p_log, OSM_LOG_ERROR, "ERR 6203 : " "failed to update SL2VLMapping tables "

opensm/opensm/osm_qos.c +2 -1
@@ -307,6 +322,7 @@ int osm_qos_setup(osm_opensm_t * p_osm)
                                   p_osm->subn.need_update;
                               if (qos_physp_setup(&p_osm->log, &p_osm->sm,
                                                   p_port, p_physp, i,
+                                                 p_osm->subn.opt.use_optimized_slvl,
                                                   force_update, &swe_config))
                                       ret = -1;
                       }

@@ -327,7 +343,7 @@ int osm_qos_setup(osm_opensm_t * p_osm)
 
               force_update = p_physp->need_update || p_osm->subn.need_update;
               if (qos_physp_setup(&p_osm->log, &p_osm->sm, p_port, p_physp,
-                                 0, force_update, cfg))
+                                 0, 0, force_update, cfg))
                       ret = -1;
       }
opensm/opensm/osm_slvl_map_rcv.c +4 -1
index 9c37442..67c71bd 100644
@@ -2,6 +2,7 @@
   * Copyright (c) 2004-2008 Voltaire, Inc. All rights reserved.
   * Copyright (c) 2002-2005 Mellanox Technologies LTD. All rights reserved.
   * Copyright (c) 1996-2003 Intel Corporation. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
   *
   * This software is available to you under a choice of one of two
   * licenses.  You may choose to be licensed under the terms of the GNU

@@ -72,7 +73,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
       osm_slvl_context_t *p_context;
       ib_net64_t port_guid;
       ib_net64_t node_guid;
-     uint8_t out_port_num, in_port_num;
+     uint32_t attr_mod;
+     uint8_t out_port_num, in_port_num, startinport, startoutport,
+             endinport, endoutport;
 
       CL_ASSERT(sm);
 

@@ -111,6 +114,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
                   (uint8_t) cl_ntoh32(p_smp->attr_mod & 0xFF000000);

              in_port_num =
                  (uint8_t) cl_ntoh32((p_smp->attr_mod & 0x00FF0000) << 8);
+             attr_mod = cl_ntoh32(p_smp->attr_mod);
+             if (attr_mod & 0x30000)
+                     goto opt_sl2vl;
              p_physp = osm_node_get_physp_ptr(p_node, out_port_num); } else {
              p_physp = p_port->p_physp;

opensm/opensm/osm_slvl_map_rcv.c +0 -0
@@ -123,7 +129,7 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)

         all we want is to update the subnet.
       */
      OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
-             "Got SLtoVL get response in_port_num %u out_port_num %u with "
+             "Received SLtoVL GetResp in_port_num %u out_port_num %u with " "GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64 ", TID 0x%" PRIx64 "\n", in_port_num, out_port_num, cl_ntoh64(port_guid), cl_ntoh64(node_guid), cl_ntoh64(p_smp->trans_id));

opensm/opensm/osm_slvl_map_rcv.c +0 -0
@@ -142,6 +148,39 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
                               out_port_num, p_slvl_tbl, OSM_LOG_DEBUG);
Quoted 6 lines

opensm/opensm/osm_slvl_map_rcv.c void osm_slvl_rcv_process(IN void *context, IN void *p_data) · line 153

Quoted 4 lines
opensm/opensm/osm_slvl_map_rcv.c +3 -0
       osm_physp_set_slvl_tbl(p_physp, p_slvl_tbl, in_port_num);
+     goto Exit;
+
+opt_sl2vl:

I think that you need to use functions here.

Where is here ? To what are you referring ?

Quoted 19 lines
Quoted 19 lines
opensm/opensm/osm_slvl_map_rcv.c +19 -0
+     OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
+             "Got optimized SLtoVL get response in_port_num %u out_port_num "
+             "%u with GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64
+             ", TID 0x%" PRIx64 "\n", in_port_num, out_port_num,
+             cl_ntoh64(port_guid), cl_ntoh64(node_guid),
+             cl_ntoh64(p_smp->trans_id));
+
+     osm_dump_slvl_map_table(sm->p_log, port_guid, in_port_num,
+                             out_port_num, p_slvl_tbl, OSM_LOG_DEBUG);
+
+     if (attr_mod & 0x10000) {
+             startoutport = ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) ? 0 : 1;
+             endoutport = osm_node_get_num_physp(p_node);
+     } else
+             endoutport = startoutport = out_port_num;
+     if (attr_mod & 0x20000) {
+             startinport = ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) ? 0 : 1;
+             endinport = osm_node_get_num_physp(p_node);
+     } else
Quoted 8 lines

opensm/opensm/osm_slvl_map_rcv.c void osm_slvl_rcv_process(IN void *context, IN void *p_data) · line 178

Quoted 6 lines
opensm/opensm/osm_slvl_map_rcv.c +6 -0
+             endinport = startinport = in_port_num;
+
+     for (out_port_num = startoutport; out_port_num < endoutport;
+          out_port_num++) {
+             p_physp = osm_node_get_physp_ptr(p_node, out_port_num);
+             if (!p_physp)

When could this be NULL?

It was just defensive. Should it be removed ?

-- Hal

Quoted 52 lines

Sasha

Quoted 50 lines
opensm/opensm/osm_slvl_map_rcv.c +5 -0
+                     continue;
+             for (in_port_num = startinport; in_port_num < endinport;
+                  in_port_num++)
+                     osm_physp_set_slvl_tbl(p_physp, p_slvl_tbl, in_port_num);
+     }

 Exit:
      cl_plock_release(sm->p_lock);

opensm/opensm/osm_subnet.c +3 -0
index 0d11811..540165a 100644
@@ -4,6 +4,7 @@
   * Copyright (c) 1996-2003 Intel Corporation. All rights reserved.
   * Copyright (c) 2008 Xsigo Systems Inc.  All rights reserved.
   * Copyright (c) 2009 System Fabric Works, Inc. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
   *
   * This software is available to you under a choice of one of two
   * licenses.  You may choose to be licensed under the terms of the GNU

@@ -352,6 +353,7 @@ static const opt_rec_t opt_tbl[] = {
       { "daemon", OPT_OFFSET(daemon), opts_parse_boolean, NULL, 0 },
       { "sm_inactive", OPT_OFFSET(sm_inactive), opts_parse_boolean, NULL, 1 },
       { "babbling_port_policy", OPT_OFFSET(babbling_port_policy), opts_parse_boolean, NULL, 1 },
+     {"use_optimized_slvl", OPT_OFFSET(use_optimized_slvl), opts_parse_boolean, NULL, 1 },
  #ifdef ENABLE_OSM_PERF_MGR
       { "perfmgr", OPT_OFFSET(perfmgr), opts_parse_boolean, NULL, 0 },
       { "perfmgr_redir", OPT_OFFSET(perfmgr_redir), opts_parse_boolean, NULL, 0 },

@@ -715,6 +717,7 @@ void osm_subn_set_default_opt(IN osm_subn_opt_t * const p_opt)
       p_opt->daemon = FALSE;
       p_opt->sm_inactive = FALSE;
       p_opt->babbling_port_policy = FALSE;
+     p_opt->use_optimized_slvl = FALSE;
  #ifdef ENABLE_OSM_PERF_MGR
       p_opt->perfmgr = FALSE;
       p_opt->perfmgr_redir = TRUE;

@@ -1501,10 +1504,13 @@ int osm_subn_output_conf(FILE *out, IN osm_subn_opt_t *const p_opts)

              "# SM Inactive\n"
              "sm_inactive %s\n\n"
              "# Babbling Port Policy\n"
-             "babbling_port_policy %s\n\n",
+             "babbling_port_policy %s\n\n"
+             "# Use Optimized SLtoVLMapping programming if supported by device\n"
+             "use_optimized_slvl %s\n\n",
              p_opts->daemon ? "TRUE" : "FALSE",
              p_opts->sm_inactive ? "TRUE" : "FALSE",
-             p_opts->babbling_port_policy ? "TRUE" : "FALSE");
+             p_opts->babbling_port_policy ? "TRUE" : "FALSE",
+             p_opts->use_optimized_slvl ? "TRUE" : "FALSE");

Quoted 8 lines
Quoted 3 lines

 #ifdef ENABLE_OSM_PERF_MGR
      fprintf(out,

--
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

#5 lore raw
In reply to a message not in the archive
Quoted 3 lines · Hal Rosenstock
Quoted 1 line · Sasha Khapyorsky

Implementation description would be very useful. What does "initial support" mean?

It means there's more to come in terms of using OptimizedSLtoVLMappingProgramming. This is the simplest use/introduction of this optional feature.

You can just send people to reads specs, your change log should explain what the patch is about, if this is a big change to opensm, maybe even RFC it will a detailed writeup

Or.
--
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

author
#6 lore raw
In reply to a message not in the archive
Quoted 23 lines · Hal Rosenstock
Quoted 13 lines
Quoted 13 lines
(quoted diff) +1 -1
@@ -150,7 +151,7 @@ static ib_api_status_t vlarb_update(osm_sm_t * sm, osm_physp_t * p,
 
  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,
+                                       unsigned optimize, unsigned force_update,
                                         const ib_slvl_table_t * sl2vl_table)
  {
       osm_madw_context_t context;

@@ -177,10 +178,18 @@ static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p,
           !memcmp(p_tbl, &tbl, sizeof(tbl)))
               return IB_SUCCESS;
Quoted 9 lines

Review static ib_api_status_t sl2vl_update_table(osm_sm_t * sm, osm_physp_t * p, · line 184

Quoted 4 lines
(quoted diff) +4 -0
+     /* both input port and output port wildcarded */
+     if (optimize && (in_port != 1 || out_port != 1))
+             return IB_SUCCESS;
+

This function (sl2vl_update_table()) is called for each in and out ports combination. I think that instead of calling it N * N times and filtering out the case 'in == out == 1' you just need it once somewhere at higher layer.

Quoted 3 lines

I was trying to minimize the changes to the current code flow. It can be done the way you propose but that will cause larger changes. Are you sure about this direction ?

I'm sure that that we don't need a bunch of empty loops. Also I think that doing this on an appropriate layer will not complicate things more.

Quoted 33 lines
Quoted 33 lines
Quoted 33 lines
(quoted diff) +3 -1
@@ -72,7 +73,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
       osm_slvl_context_t *p_context;
       ib_net64_t port_guid;
       ib_net64_t node_guid;
-     uint8_t out_port_num, in_port_num;
+     uint32_t attr_mod;
+     uint8_t out_port_num, in_port_num, startinport, startoutport,
+             endinport, endoutport;
 
       CL_ASSERT(sm);
 

@@ -111,6 +114,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
                   (uint8_t) cl_ntoh32(p_smp->attr_mod & 0xFF000000);

              in_port_num =
                  (uint8_t) cl_ntoh32((p_smp->attr_mod & 0x00FF0000) << 8);
+             attr_mod = cl_ntoh32(p_smp->attr_mod);
+             if (attr_mod & 0x30000)
+                     goto opt_sl2vl;
              p_physp = osm_node_get_physp_ptr(p_node, out_port_num); } else {
              p_physp = p_port->p_physp;

(quoted diff) +0 -0
@@ -123,7 +129,7 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)

         all we want is to update the subnet.
       */
      OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
-             "Got SLtoVL get response in_port_num %u out_port_num %u with "
+             "Received SLtoVL GetResp in_port_num %u out_port_num %u with " "GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64 ", TID 0x%" PRIx64 "\n", in_port_num, out_port_num, cl_ntoh64(port_guid), cl_ntoh64(node_guid), cl_ntoh64(p_smp->trans_id));

(quoted diff) +0 -0
@@ -142,6 +148,39 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
                               out_port_num, p_slvl_tbl, OSM_LOG_DEBUG);
Quoted 8 lines
Quoted 6 lines

Review void osm_slvl_rcv_process(IN void *context, IN void *p_data) · line 153

Quoted 4 lines
(quoted diff) +3 -0
       osm_physp_set_slvl_tbl(p_physp, p_slvl_tbl, in_port_num);
+     goto Exit;
+
+opt_sl2vl:

I think that you need to use functions here.

Where is here ? To what are you referring ?

When your function becomes to form of:

	if (cond1)
		goto do1;
	
	do_else();

	goto exit;

do1:

	do_1();

exit:

	return;

I'm suggesting to improve flows, for example by using functions (or another way if applicable and achieves the simplification goal).

Quoted 29 lines
Quoted 20 lines
Quoted 20 lines
(quoted diff) +19 -0
+     OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
+             "Got optimized SLtoVL get response in_port_num %u out_port_num "
+             "%u with GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64
+             ", TID 0x%" PRIx64 "\n", in_port_num, out_port_num,
+             cl_ntoh64(port_guid), cl_ntoh64(node_guid),
+             cl_ntoh64(p_smp->trans_id));
+
+     osm_dump_slvl_map_table(sm->p_log, port_guid, in_port_num,
+                             out_port_num, p_slvl_tbl, OSM_LOG_DEBUG);
+
+     if (attr_mod & 0x10000) {
+             startoutport = ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) ? 0 : 1;
+             endoutport = osm_node_get_num_physp(p_node);
+     } else
+             endoutport = startoutport = out_port_num;
+     if (attr_mod & 0x20000) {
+             startinport = ib_switch_info_is_enhanced_port0(&p_node->sw->switch_info) ? 0 : 1;
+             endinport = osm_node_get_num_physp(p_node);
+     } else
Quoted 8 lines

Review void osm_slvl_rcv_process(IN void *context, IN void *p_data) · line 178

Quoted 6 lines
(quoted diff) +6 -0
+             endinport = startinport = in_port_num;
+
+     for (out_port_num = startoutport; out_port_num < endoutport;
+          out_port_num++) {
+             p_physp = osm_node_get_physp_ptr(p_node, out_port_num);
+             if (!p_physp)

When could this be NULL?

Quoted 1 line

It was just defensive. Should it be removed ?

I wasn't sure, so asked. If it is "just defensive" please remove it.

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

#7 lore raw
In reply to Sasha Khapyorsky

On Thu, Oct 29, 2009 at 9:23 PM, Sasha Khapyorsky <sashak-smomgflXvOZWk0Htik3J/[email protected]> wrote:
<snip...>

Quoted 45 lines
Quoted 45 lines
opensm/opensm/osm_slvl_map_rcv.c +4 -1
index 9c37442..67c71bd 100644
@@ -2,6 +2,7 @@
   * Copyright (c) 2004-2008 Voltaire, Inc. All rights reserved.
   * Copyright (c) 2002-2005 Mellanox Technologies LTD. All rights reserved.
   * Copyright (c) 1996-2003 Intel Corporation. All rights reserved.
+ * Copyright (c) 2009 HNR Consulting. All rights reserved.
   *
   * This software is available to you under a choice of one of two
   * licenses.  You may choose to be licensed under the terms of the GNU

@@ -72,7 +73,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
       osm_slvl_context_t *p_context;
       ib_net64_t port_guid;
       ib_net64_t node_guid;
-     uint8_t out_port_num, in_port_num;
+     uint32_t attr_mod;
+     uint8_t out_port_num, in_port_num, startinport, startoutport,
+             endinport, endoutport;
 
       CL_ASSERT(sm);
 

@@ -111,6 +114,9 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
                   (uint8_t) cl_ntoh32(p_smp->attr_mod & 0xFF000000);

              in_port_num =
                  (uint8_t) cl_ntoh32((p_smp->attr_mod & 0x00FF0000) << 8);
+             attr_mod = cl_ntoh32(p_smp->attr_mod);
+             if (attr_mod & 0x30000)
+                     goto opt_sl2vl;
              p_physp = osm_node_get_physp_ptr(p_node, out_port_num); } else {
              p_physp = p_port->p_physp;

opensm/opensm/osm_slvl_map_rcv.c +0 -0
@@ -123,7 +129,7 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)

         all we want is to update the subnet.
       */
      OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
-             "Got SLtoVL get response in_port_num %u out_port_num %u with "
+             "Received SLtoVL GetResp in_port_num %u out_port_num %u with " "GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64 ", TID 0x%" PRIx64 "\n", in_port_num, out_port_num, cl_ntoh64(port_guid), cl_ntoh64(node_guid), cl_ntoh64(p_smp->trans_id));

opensm/opensm/osm_slvl_map_rcv.c +0 -0
@@ -142,6 +148,39 @@ void osm_slvl_rcv_process(IN void *context, IN void *p_data)
                               out_port_num, p_slvl_tbl, OSM_LOG_DEBUG);
Quoted 14 lines

opensm/opensm/osm_slvl_map_rcv.c void osm_slvl_rcv_process(IN void *context, IN void *p_data) · line 153

Quoted 4 lines
opensm/opensm/osm_slvl_map_rcv.c +3 -0
       osm_physp_set_slvl_tbl(p_physp, p_slvl_tbl, in_port_num);
+     goto Exit;
+
+opt_sl2vl:

I think that you need to use functions here.

Quoted 7 lines
opensm/opensm/osm_slvl_map_rcv.c +7 -0
+     OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
+             "Got optimized SLtoVL get response in_port_num %u out_port_num "
+             "%u with GUID 0x%" PRIx64 " for parent node GUID 0x%" PRIx64
+             ", TID 0x%" PRIx64 "\n", in_port_num, out_port_num,
+             cl_ntoh64(port_guid), cl_ntoh64(node_guid),
+             cl_ntoh64(p_smp->trans_id));
+

Is this a macro/va args limit ?

Does the existing log message this was modeled after need to change too ? It is virtually the same (maybe one less arg).

-- Hal

<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