iw_cxgb4: do not send RX_DATA_ACK CPLs after close/abort

[ Upstream commit 3bcf96e0183f5c863657cb6ae9adad307a0f6071 ]

Function rx_data(), which handles ingress CPL_RX_DATA messages, was
always sending an RX_DATA_ACK with the goal of updating the credits.
However, if the RDMA connection is moved out of FPDU mode abruptly,
then it is possible for iw_cxgb4 to process queued RX_DATA CPLs after HW
has aborted the connection.  These CPLs should not trigger RX_DATA_ACKS.
If they do, HW can see a READ after DELETE of the DB_LE hash entry for
the tid and post a LE_DB HashTblMemCrcError.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/infiniband/hw/cxgb4/cm.c b/drivers/infiniband/hw/cxgb4/cm.c
index f1510cc..9398143 100644 (file)
--- a/drivers/infiniband/hw/cxgb4/cm.c
+++ b/drivers/infiniband/hw/cxgb4/cm.c
@@ -1804,20 +1804,21 @@ static int rx_data(struct c4iw_dev *dev, struct sk_buff *skb)
        skb_trim(skb, dlen);
        mutex_lock(&ep->com.mutex);
 
-       /* update RX credits */
-       update_rx_credits(ep, dlen);
-
        switch (ep->com.state) {
        case MPA_REQ_SENT:
+               update_rx_credits(ep, dlen);
                ep->rcv_seq += dlen;
                disconnect = process_mpa_reply(ep, skb);
                break;
        case MPA_REQ_WAIT:
+               update_rx_credits(ep, dlen);
                ep->rcv_seq += dlen;
                disconnect = process_mpa_request(ep, skb);
                break;
        case FPDU_MODE: {
                struct c4iw_qp_attributes attrs;
+
+               update_rx_credits(ep, dlen);
                BUG_ON(!ep->com.qp);
                if (status)
                        pr_err("%s Unexpected streaming data." \