]> err.no Git - linux-2.6/commit
RDMA/cxgb3: QP flush fixes
authorSteve Wise <swise@opengridcomputing.com>
Fri, 2 May 2008 16:17:41 +0000 (11:17 -0500)
committerRoland Dreier <rolandd@cisco.com>
Fri, 2 May 2008 17:56:57 +0000 (10:56 -0700)
commitc8286944b802c5ce4063ec3c334b38c6757a9434
treef1cbf7c3baa83cc2e4e92ad43cf5083cd956ad32
parent57ce41d1d18279cc90223f3deadca70c7de1cfca
RDMA/cxgb3: QP flush fixes

- Flush the QP only after the HW disables the connection.  Currently
  we flush the QP when transitioning to CLOSING.  This exposes a race
  condition where the HW can complete a RECV WR, for instance, -and-
  the SW can flush that same WR.

- Only call CQ event handlers on flush IFF we actually flushed something.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
drivers/infiniband/hw/cxgb3/cxio_hal.c
drivers/infiniband/hw/cxgb3/cxio_hal.h
drivers/infiniband/hw/cxgb3/iwch_qp.c