RDMA/cxgb3: Don't complete flushed send work requests twice

When the SQ is flushed, mark the flushed entries as not signaled so
the poll logic doesn't re-insert the CQ entry thinking its an out of
order completion.

The bug can cause the NFS/RDMA server to crash due to processing the
same completed work request twice.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/cxgb3/cxio_hal.c b/drivers/infiniband/hw/cxgb3/cxio_hal.c
index 8d71086f5a1c2b8c0e55bb6168ee0cdb474de520..62f9cf2f94ec647756dcd5f2ae06c9f01a9eafac 100644 (file)
--- a/drivers/infiniband/hw/cxgb3/cxio_hal.c
+++ b/drivers/infiniband/hw/cxgb3/cxio_hal.c
@@ -410,6 +410,7 @@ int cxio_flush_sq(struct t3_wq *wq, struct t3_cq *cq, int count)
        ptr = wq->sq_rptr + count;
        sqp = wq->sq + Q_PTR2IDX(ptr, wq->sq_size_log2);
        while (ptr != wq->sq_wptr) {
+               sqp->signaled = 0;
                insert_sq_cqe(wq, cq, sqp);
                ptr++;
                sqp = wq->sq + Q_PTR2IDX(ptr, wq->sq_size_log2);