[openib-general] [PATCH] osm: reviewing osmtest - osmt_multicast.c

[Yevgeny Kliteynik]

Hal Rosenstock wrote:
> On Wed, 2006-10-11 at 06:23, Yevgeny Kliteynik wrote:
>> Hi Hal
>>
>> Fixing a few problems in the multicast test flow,
>> plus some cosmetics.
>>
>> Yevgeny
>>
>> Signed-off-by:  Yevgeny Kliteynik <kliteyn at dev.mellanox.co.il>
> 
> Thanks. Applied. See question below...
> 
>> Index: osmt_multicast.c
>> ===================================================================
>> --- osmt_multicast.c	(revision 9776)
>> +++ osmt_multicast.c	(working copy)
> 
> [snip...]
> 
>> @@ -1808,14 +1865,54 @@ osmt_run_mcast_flow( IN osmtest_t * cons
>>  
>>    /*  o15.0.1.6: */
>>    /*  - Create a new MCG with valid requested MGID. */
>> +  osmt_init_mc_query_rec(p_osmt, &mc_req_rec);
>> +  mc_req_rec.mgid = good_mgid;
>>  
>>    osm_log( &p_osmt->log, OSM_LOG_INFO,
>>             "osmt_run_mcast_flow: "
>> -           "Checking Create given MGID=0x%016" PRIx64 " : "
>> +           "Checking Create given valid MGID=0x%016" PRIx64 " : "
>> +           "0x%016" PRIx64 " (o15.0.1.6)...\n",
>> +           cl_ntoh64(mc_req_rec.mgid.unicast.prefix),
>> +           cl_ntoh64(mc_req_rec.mgid.unicast.interface_id));
>> +
>> +  /* Before creation, need to check that this group doesn't exist */
>> +  osm_log( &p_osmt->log, OSM_LOG_INFO,
>> +           "osmt_run_mcast_flow: "
>> +           "Verifying that MCGroup with this MGID doesn't exist by trying to Join it (o15.0.1.13)...\n");
>> +
>> +  ib_member_set_join_state(&mc_req_rec, IB_MC_REC_STATE_NON_MEMBER);
>> +
>> +  osm_log( &p_osmt->log, OSM_LOG_ERROR,
>> +           "osmt_run_mcast_flow: " EXPECTING_ERRORS_START "\n" );
>> +  status = osmt_send_mcast_request( p_osmt, 1, /*  join */
>> +                                    &mc_req_rec,
>> +                                    comp_mask,
>> +                                    &res_sa_mad );
>> +  osm_log( &p_osmt->log, OSM_LOG_ERROR,
>> +           "osmt_run_mcast_flow: " EXPECTING_ERRORS_END "\n" );
>> +
>> +  if ((status != IB_REMOTE_ERROR) ||
>> +      (res_sa_mad.status != IB_SA_MAD_STATUS_REQ_INVALID))
>> +  {
>> +    osm_log( &p_osmt->log, OSM_LOG_ERROR,
>> +             "osmt_run_mcast_flow: ERR 0301: "
>> +             "Tried joining group that shouldn't have existed - got %s/%s\n",
>> +             ib_get_err_str( status ),
>> +             ib_get_mad_status_str( (ib_mad_t*)(&res_sa_mad) )
>> +             );
>> +    status = IB_ERROR;
>> +    goto Exit;
>> +  }
> 
> In the event this works, the SA is potentially left with some bad state
> because of this. Should the join be removed for this case ?

When osmtest completes w/o error, at the end of the test flow
SM is returned to its initial state.
When osmtest runs into some error, it exits, leaving SM as is.
I think it's better to leave SM state as is after discovering 
a failure for further examination.
But even if we wanted to 'clean up', it is not always possible.
For instance, what happens when multicast group was created by 
osmtest, and the group removal fails?

--
Yevgeny
> 
> -- Hal
>