[openib-general] Re: [PATCH] mthca: Don't call CQ completion handler if it doesn't exist

Roland Dreier roland at topspin.com
Mon Apr 11 12:07:35 PDT 2005


    Roland> Why do we want to add this test?  This is adding a
    Roland> conditional branch in what I think is a fast path, and I
    Roland> would consider it a bug in the consumer if it creates a CQ
    Roland> with an invalid completion handler and then requests a
    Roland> completion event for that CQ.  Am I missing something?

    Hal> Then shouldn't this be indicated as an error at create_cq time ?

How can it be?  We don't know if the consumer is going to call
ib_req_notify_cq() or not.

 - R.




More information about the general mailing list