[openib-general] [PATCH 0/4] Dispatch communication relatedevents to the IB CM

[Michael S. Tsirkin]

Quoting r. Or Gerlitz <ogerlitz at voltaire.com>:
> Subject: Re: [PATCH 0/4] Dispatch communication relatedevents to the IB CM
> 
> Sean Hefty wrote:
> >> - Do _all_ IB devices (not just mthca) queue WR posted on the SQ when
> >> the QP is in RTR or can they generate an error if I don't wait until the
> >> QP --> RTS to reply to a request.
> > 
> > All IB devices should allow posting of WR to the SQ while in the RTR state.
> 
> Sean,
> 
> Is this statement based on the IB spec, the mthca driver or both?
> 
>  From the mthca code it seems there is a place holder for checking and 
> doing some sort of special handing for this case, but this is not 
> implemented yet.
> 
> mthca/mthca_qp.c:1507:  /* XXX check that state is OK to post send */
> mthca/mthca_qp.c:1732:  /* XXX check that state is OK to post receive */
> mthca/mthca_qp.c:1848:  /* XXX check that state is OK to post send */
> mthca/mthca_qp.c:2101:  /* XXX check that state is OK to post receive */
> 
> I think special handling (eg queuing of TX WRs if the QP is still in 
> RTR) is some sort of over doing, but lets first see if indeed its not 
> implemented and what others think.
> 
> Or.

IB spec 1.2 vol 1 page 456

C10-29: If a Work Request is submitted to a Send Queue while its corresponding
QP is in the RTR State, an immediate error shall be returned.

So, while you technically *can* post on a QP in RTR, you won't, and are not
supposed to, get anything useful out of this.

Sean, could you please restate why is this actually useful?

-- 
MST