[openib-general] Re: race in mthca_cq.c?

[Roland Dreier]

 > The following seems to work. How does it look?

I don't think it's needed, and anyway I don't see how it fixes
things.  The problem only happens when the new CQ or QP has the same
number as an old CQ/QP, so the test of cq->cqn == cqn might still pass
even if the cq has changed (there's no guarantee the upper bits won't
repeat -- or someone could be using 24 bits for index)

 - R.