[openib-general] Re: IPoIB Completion Handling

[Sean Hefty]

Hal Rosenstock wrote:

> On Tue, 2004-11-09 at 10:37, Roland Dreier wrote:
> 
>>By the way, reposting the receives is not the right thing to do on
>>error -- the QP will be in the error state, so any new work requests
>>will just complete with a flush status.  We need to reset the QP and
>>start over to recover from errors.
> 
> 
> Is the same thing true for QP0/1 ? If so, this needs to be done there as
> well. (There used to be a port restart there but this was excised).

Btw, I have plans to get to this shortly.  I have the send queuing code 
complete (need to re-merge after the patches this morning), but I 
haven't been able to debug the code yet.  I'm running into some issues 
configuring a point-to-point "fabric", with opensm running on the 
sourceforge stack on the other node.  I have some changes to handle send 
queuing that are needed when recovering from QP errors as well.

- Sean