[openib-general] osm: osmtest new flow of informinfo fails

[Hal Rosenstock]

Hi Eitan,

On Thu, 2006-12-07 at 05:28, Eitan Zahavi wrote:
> Hi Hal,
> 
> All osmtest flows fail for me with the following error:

By all flows, you mean osmtest -a (the all flows test).

> I start the log from the first inform info related message to give you 
> the context.
> 
> Dec 07 11:48:17 656752 [B7FD48E0] -> osmtest_get_node_rec_by_lid: 
> Getting node record for LID 0xFFFF
> Dec 07 11:48:17 663592 [B6BD1BB0] -> __osmv_sa_mad_rcv_cb: ERR 0501: 
> Remote error:0x0C00 .
> Dec 07 11:48:17 663642 [B6BD1BB0] -> osmtest_query_res_cb: ERR 0003: 
> Error on query (IB_REMOTE_ERROR)
> Dec 07 11:48:17 663694 [B7FD48E0] -> osmtest_informinfo_request: ERR 
> 008F: ib_query failed (IB_REMOTE_ERROR)
> Dec 07 11:48:17 663729 [B7FD48E0] -> osmtest_informinfo_request: Remote 
> error = IB_MAD_STATUS_UNSUP_METHOD_ATTR
> Dec 07 11:48:17 663759 [B7FD48E0] -> osmtest_informinfo_request: 
> InformInfoRecord IS EXPECTED ERROR ^^^^
> Dec 07 11:48:17 667671 [B6BD1BB0] -> __osmv_sa_mad_rcv_cb: ERR 0501: 
> Remote error:0x0C00 .
> Dec 07 11:48:17 667705 [B6BD1BB0] -> osmtest_query_res_cb: ERR 0003: 
> Error on query (IB_REMOTE_ERROR)
> Dec 07 11:48:17 667756 [B7FD48E0] -> osmtest_informinfo_request: ERR 
> 008F: ib_query failed (IB_REMOTE_ERROR)
> Dec 07 11:48:17 667789 [B7FD48E0] -> osmtest_informinfo_request: Remote 
> error = IB_MAD_STATUS_UNSUP_METHOD_ATTR
> Dec 07 11:48:17 667820 [B7FD48E0] -> osmtest_informinfo_request: 
> InformInfo IS EXPECTED ERROR ^^^^
> Dec 07 11:48:17 669403 [B6BD1BB0] -> __osmv_sa_mad_rcv_cb: ERR 0501: 
> Remote error:0x0002 .
> Dec 07 11:48:17 669436 [B6BD1BB0] -> osmtest_query_res_cb: ERR 0003: 
> Error on query (IB_REMOTE_ERROR)
> Dec 07 11:48:17 669489 [B7FD48E0] -> osmtest_informinfo_request: ERR 
> 008F: ib_query failed (IB_REMOTE_ERROR)
> Dec 07 11:48:17 669561 [B7FD48E0] -> osmtest_informinfo_request: Remote 
> error = IB_SA_MAD_STATUS_REQ_INVALID
> Dec 07 11:48:17 669590 [B7FD48E0] -> osmtest_informinfo_request: 
> InformInfo UnSubscribe IS EXPECTED ERROR ^^^^
> Dec 07 11:48:17 672731 [B6BD1BB0] -> __osmv_sa_mad_rcv_cb: ERR 0501: 
> Remote error:0x0002 .
> Dec 07 11:48:17 672772 [B6BD1BB0] -> osmtest_query_res_cb: ERR 0003: 
> Error on query (IB_REMOTE_ERROR)
> Dec 07 11:48:17 672826 [B7FD48E0] -> osmtest_informinfo_request: ERR 
> 008F: ib_query failed (IB_REMOTE_ERROR)
> Dec 07 11:48:17 672859 [B7FD48E0] -> osmtest_informinfo_request: Remote 
> error = IB_SA_MAD_STATUS_REQ_INVALID
> Dec 07 11:48:17 672894 [B7FD48E0] -> osmtest_run: ERR 0146: SA 
> validation database failure (IB_INSUFFICIENT_MEMORY)

This is a failure of the first subscribe.

> OpenSM log says:
> Dec 07 11:48:17 668513 [B57DABB0] -> osm_infr_rcv_process_set_method: 
> ERR 4307: Failed to UnSubscribe to non existin
> g inform object
> Dec 07 11:48:17 671896 [B75DDBB0] -> osm_infr_rcv_process_set_method: 
> ERR 4307: Failed to UnSubscribe to non existin
> g inform object

The first one is correct. The second one is due to bad treatment on the
valid subscribe. Evidently, it is now somehow being treated as an
unsubscribe rather than a subscribe. Can you run opensm with -V to see
all the log messages which will give a better indication of what path it
is taking in osm_infr_rcv_process_set_method. Thanks.

> Please let me know if you want me to debug it.

This works for me. Not sure what is different.

-- Hal

> Eitan