[ofa-general] rdma_resolve_route() returning -EINVAL

[Talpey, Thomas]

At 05:08 AM 10/6/2008, Or Gerlitz wrote:
>Roland Dreier wrote:
>> The issue is more of spec compliance than a likely real-life 
>scenario... and as for why no one else is worrying about it, I think 
>it's because the only other user of rdma_connect() in the tree is 
>iSER, and I guess no one worried too much there.  SRP uses the IB CM 
>directly, and waits for timewait exit before calling a connection closed.
>Roland,
>
>I guess there's some tradeoff here between the time connection recovery 
>would take when the ULPs does wait for the timewait event vs the risk of 
>getting into the target considering the REQ as stale and rejecting. Does 
>SRP just wait for the event, or the new connection is established in 
>parallel, which also means it would use a different QP number always.

That's what the NFS/RDMA client does - I always create a new cm_id and qp.
So the TIMEWAIT upcall isn't very interesting, unless I change that. My
problem was that I started getting the upcall, when I didn't before.

Tom.