[openib-general] 3513 DAPL is Broken

[James Lentini]

Is there a way to reproduce this with dapltest? 

Our 6 canonical regression tests (see 
test/dapltest/scripts/regress.sh), don't encounter this problem on 
revision 3521.

Are you sure the application is hanging in DAPL? Can you enable DAPL 
debugging and send the output (see doc/dapl_environ.txt and 
doc/dat_environ.txt)?

Thanks,
james

On Wed, 21 Sep 2005, Woodruff, Robert J wrote:

> Seems to hang around the time of the modify QP.
> 
> ibv_rc_pingpong seems to work OK and also your 
> DAPL-socket CM version that you gave me yesterday seems 
> to work, but the DAPL I pulled from SVN that uses the IB AT/CM
> has the following problem. 
> 
> I am starting to think that pushing out your socket CM
> version until things stabilize with the IBAT/IBCM version
> might be worth considering, so that people that
> want to use DAPL now have something that is reliable.
> 
> woody
> 
> Here is the dapl trace when running Intel MPI on top of uDAPL 3513, 
> 
> dapl_ia_query (0x522860, (nil), 0x0, (nil), 0x3ffffff, 0x7fbfffe510)
> dapl_ia_query () returns 0x0
> dapl_evd_create () returns 0x0
>  setup_listener(ia_ptr 0x522860 SID 3545 sp 0x5238e0 conn 0x5239a0 id
> 5389248)
>  setup_listener(conn=0x5239a0 cm_id=5389248)
> dapl_ep_create (0x522860, 0x5235a0, 0x523620, 0x523620, 0x523780,
> 0x7fbfffecb0, 0x5201e8)
>  query_hca: MAX msg 2147483648 dto 65535 iov 59 rdma i4,o4
>  qp_alloc: ia_ptr 0x522860 ep_ptr 0x526740 ep_ctx_ptr 0x526740
>  qp_alloc: qpn 0xc0409 sq 1000,9 rq 1000,1
>  modify_qp: qp 0x523c50, state 1 qp_num 0xc0409
> dapl_ep_create (0x522860, 0x5235a0, 0x523620, 0x523620, 0x523780,
> 0x7fbfffecb0, 0x5203b8)
>  query_hca: MAX msg 2147483648 dto 65535 iov 59 rdma i4,o4
>  qp_alloc: ia_ptr 0x522860 ep_ptr 0x526a20 ep_ctx_ptr 0x526a20
>  qp_alloc: qpn 0xc040a sq 1000,9 rq 1000,1
>  modify_qp: qp 0x526d00, state 1 qp_num 0xc040a
> dapl_ep_create (0x522860, 0x5235a0, 0x523620, 0x523620, 0x523780,
> 0x7fbfffecb0, 0x520758)
>  query_hca: MAX msg 2147483648 dto 65535 iov 59 rdma i4,o4
>  qp_alloc: ia_ptr 0x522860 ep_ptr 0x593470 ep_ctx_ptr 0x593470
>  qp_alloc: qpn 0xc040b sq 1000,9 rq 1000,1
>  modify_qp: qp 0x526e40, state 1 qp_num 0xc040b