[ofa-general] Re: [PATCH v2] opensm: set IS_SM bit during opensm init

[Sasha Khapyorsky]

On 16:15 Thu 12 Mar     , Eli Dorfman (Voltaire) wrote:
> 
> diff --git a/opensm/opensm/osm_opensm.c b/opensm/opensm/osm_opensm.c
> index 7de2e5b..cfe6474 100644
> --- a/opensm/opensm/osm_opensm.c
> +++ b/opensm/opensm/osm_opensm.c
> @@ -475,6 +475,11 @@ osm_opensm_bind(IN osm_opensm_t * const p_osm, IN const ib_net64_t guid)
>  		goto Exit;
>  #endif				/* ENABLE_OSM_PERF_MGR */
>  
> +	/* setting IS_SM in capability mask */
> +	OSM_LOG(&p_osm->log, OSM_LOG_INFO, "Setting IS_SM on port 0x%016" PRIx64 "\n",
> +			cl_ntoh64(guid));
> +	osm_vendor_set_sm(p_osm->sm.mad_ctrl.h_bind, TRUE);

This looks correct fix for me. Let's wait day or two for potential
another comments.

> @@ -139,16 +115,7 @@ __osm_pi_rcv_process_endport(IN osm_sm_t * sm,
>  		}
>  	}
>  
> -	if (port_guid == sm->p_subn->sm_port_guid) {
> -		/*
> -		   We received the PortInfo for our own port.
> -		 */
> -		if (!(p_pi->capability_mask & IB_PORT_CAP_IS_SM))
> -			/*
> -			   Set the IS_SM bit to indicate our port hosts an SM.
> -			 */
> -			__osm_pi_rcv_set_sm(sm, p_physp);
> -	} else {
> +	if (port_guid != sm->p_subn->sm_port_guid) {
>  		p_sm_tbl = &sm->p_subn->sm_guid_tbl;
>  		if (p_pi->capability_mask & IB_PORT_CAP_IS_SM) {

Just thought unrelated to this patch. In the past I started to think
about redoing sm_guid_tbl so that it will keep our SM as well, for first
look it should simplify some flows and SA SMinfoRecord processing. Was
interrupted by something else then...

Sasha