[ofa-general] two questions about RDMA_CM_EVENT_TIMEWAIT_EXIT and the TimeWait state
Isaac Huang
He.Huang at Sun.COM
Wed May 27 21:33:19 PDT 2009
On Tue, May 26, 2009 at 11:22:24PM +0300, Or Gerlitz wrote:
> On Tue, May 26, 2009 at 11:03 PM, Isaac Huang <[1]He.Huang at sun.com>
> wrote:
>
> If rdma_destroy_qp is called on a QP before it exits the TimeWait
> state (i.e. after RDMA_CM_EVENT_DISCONNECTED but before
> RDMA_CM_EVENT_TIMEWAIT_EXIT), is it possible that a subsequent
> rdma_create_qp would reuse the same QP while it's still in TimeWait?
>
> YES - as rdma_destroy/create_qp are basically wrappers to
> ib_destroy/create_qp and the latter two are not aware by any means to
> the QP state from the CM point of view.
Thanks, they should probably be called CM ID states instead if QP
states.
Isaac
More information about the general
mailing list