Re: [PATCH v2] opensm: Multicast root switch calculation

[Sasha Khapyorsky <sashak-smomgflXvOZWk0Htik3J/w@public.gmane.org>]

On 10:39 Thu 28 Jan     , Hal Rosenstock wrote:
> On Wed, Jan 27, 2010 at 5:45 AM, Sasha Khapyorsky <sashak-smomgflXvObQFizaE/u3fw@public.gmane.orgm> wrote:
> > On 13:59 Wed 20 Jan     , Sasha Khapyorsky wrote:
> >> On 13:32 Wed 20 Jan     , Slava Strebkov wrote:
> >> > "average hops" was chosen instead of "max hops" because in root weight
> >> > calculation the number of ports is also important, not only the distance
> >> > (hops).
> >>
> >> But this patch is declared as root switch calculation optimization, not
> >> as algorithm change (actually I even missed this part in V1).
> >
> > I reworked this patch preserving original ("max hops") calculation
> > method. Please look at this.
> >
> > The next step is to evaluate "max hops" -> "average hops" switch and to
> > cleanup OSM_VENDOR_INTF_ANAFA macro.
> >
> > Sasha
> >
> >
> > From: Slava Strebkov <slavas-hKgKHo2Ms0F+cjeuK/JdrQ@public.gmane.org>
> > Date: Thu, 3 Dec 2009 16:11:30 +0200
> > Subject: [PATCH] opensm: Multicast root switch calculation
> >
> > Proposed new algorithm for calculation of root switch for multicast
> > spanning tree. Only edge switches(those connected to hosts)
> 
> What about switches whose peer port is a router ? Shouldn't they be
> included here ?

Yes.

> 
> > and
> > switches - multicast members themselves are involved in root calculation.
> > This gives improvement, especially on large fabrics, since number of
> > switches usually much less then the number of ports, shared same mcast
> > group.
> >
> > Signed-off-by: Slava Strebkov <slavas-smomgflXvOZWk0Htik3J/w@public.gmane.org>
> > Signed-off-by: Sasha Khapyorsky <sashak-smomgflXvOZWk0Htik3J/w@public.gmane.org>
> > ---
> >  opensm/include/opensm/osm_switch.h |   12 +++
> >  opensm/opensm/osm_mcast_mgr.c      |  149 ++++++++++++++++++++++++++---------
> >  2 files changed, 122 insertions(+), 39 deletions(-)
> >
> > diff --git a/opensm/include/opensm/osm_switch.h b/opensm/include/opensm/osm_switch.h
> > index 205896d..cb6e5ac 100644
> > --- a/opensm/include/opensm/osm_switch.h
> > +++ b/opensm/include/opensm/osm_switch.h
> > @@ -109,6 +109,9 @@ typedef struct osm_switch {
> >        unsigned endport_links;
> >        unsigned need_update;
> >        void *priv;
> > +       cl_map_item_t mgrp_item;
> > +       uint32_t num_of_mcm;
> > +       uint8_t is_mc_member;
> >  } osm_switch_t;
> >  /*
> >  * FIELDS
> > @@ -151,6 +154,15 @@ typedef struct osm_switch {
> >  *              When set indicates that switch was probably reset, so
> >  *              fwd tables and rest cached data should be flushed
> >  *
> > +*      mgrp_item
> > +*              map item for switch in building mcast tree
> > +*
> > +*      num_of_mcm
> > +*              number of mcast members(ports) connected to switch
> > +*
> > +*      is_mc_member
> > +*              whether switch is a mcast member itself
> > +*
> >  * SEE ALSO
> >  *      Switch object
> >  *********/
> > diff --git a/opensm/opensm/osm_mcast_mgr.c b/opensm/opensm/osm_mcast_mgr.c
> > index dce9f2b..5c9d0bc 100644
> > --- a/opensm/opensm/osm_mcast_mgr.c
> > +++ b/opensm/opensm/osm_mcast_mgr.c
> > @@ -157,50 +157,119 @@ static void mcast_mgr_purge_tree(osm_sm_t * sm, IN osm_mgrp_box_t * mbox)
> >        OSM_LOG_EXIT(sm->p_log);
> >  }
> >
> > -static float osm_mcast_mgr_compute_avg_hops(osm_sm_t * sm, cl_qlist_t * l,
> > -                                           const osm_switch_t * p_sw)
> > +static void mcast_mgr_build_switch_map(osm_sm_t * sm,
> > +                                      const cl_qlist_t * port_list,
> > +                                      cl_qmap_t * p_mcast_member_sw_tbl)
> >  {
> > -       float avg_hops = 0;
> > -       uint32_t hops = 0;
> > -       uint32_t num_ports = 0;
> > -       cl_list_item_t *i;
> > +       osm_switch_t *remote_sw;
> > +       cl_list_item_t *list_item;
> > +       osm_port_t *p_port;
> > +       ib_net64_t port_guid;
> > +       osm_physp_t *p_physp_remote;
> > +       osm_node_t *remote_node;
> >        osm_mcast_work_obj_t *wobj;
> >
> >        OSM_LOG_ENTER(sm->p_log);
> >
> > -       /*
> > -          For each member of the multicast group, compute the
> > -          number of hops to its base LID.
> > -        */
> > -       for (i = cl_qlist_head(l); i != cl_qlist_end(l); i = cl_qlist_next(i)) {
> > -               wobj = cl_item_obj(i, wobj, list_item);
> > -               hops += osm_switch_get_port_least_hops(p_sw, wobj->p_port);
> > -               num_ports++;
> > +       cl_qmap_init(p_mcast_member_sw_tbl);
> > +       for (list_item = cl_qlist_head(port_list);
> > +            list_item != cl_qlist_end(port_list);
> > +            list_item = cl_qlist_next(list_item)) {
> > +               wobj = cl_item_obj(list_item, wobj, list_item);
> > +               p_port = wobj->p_port;
> > +               if (!p_port)
> > +                       continue;
> > +               if (p_port->p_node->sw) {
> > +                       /* for switches - remote switch would be the switch itself */
> > +                       remote_node = osm_physp_get_node_ptr(p_port->p_physp);
> > +               } else {
> > +                       p_physp_remote = osm_physp_get_remote(p_port->p_physp);
> > +                       remote_node = osm_physp_get_node_ptr(p_physp_remote);
> > +               }
> > +               /* get the remote switch of the mcmember */
> > +               remote_sw = remote_node->sw;
> > +               port_guid = osm_node_get_node_guid(remote_node);
> > +               if (cl_qmap_get(p_mcast_member_sw_tbl, port_guid) ==
> > +                       cl_qmap_end(p_mcast_member_sw_tbl)) {
> > +                               /* insert switch to table */
> > +                               cl_qmap_insert(p_mcast_member_sw_tbl, port_guid, &remote_sw->mgrp_item);
> > +                               /* New element in the table */
> > +                               if (osm_node_get_type(p_port->p_node) == IB_NODE_TYPE_CA)
> > +                                       /* for HCA update the MC count on the remote switch */
> 
> Should this be != IB_NODE_TYPE_SWITCH so that both CAs and routers are
> included here ?

Yes, obviously.

Sasha
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html