3.16.55-rc1 review patch. If anyone has any objections, please let me know.
------------------
From: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
commit f55688c45442bc863f40ad678c638785b26cdce6 upstream.
If the RECV CQE is in error, ignore the MSN check. This was causing
recvs that were flushed into the sw cq to be completed with the wrong
status (BAD_MSN instead of FLUSHED).
Signed-off-by: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Jason Gunthorpe <jgg@xxxxxxxxxxxx>
[bwh: Backported to 3.16: adjust context]
Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx>
---
drivers/infiniband/hw/cxgb4/cq.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
--- a/drivers/infiniband/hw/cxgb4/cq.c
+++ b/drivers/infiniband/hw/cxgb4/cq.c
@@ -574,10 +574,10 @@ static int poll_cq(struct t4_wq *wq, str
ret = -EAGAIN;
goto skip_cqe;
}
- if (unlikely((CQE_WRID_MSN(hw_cqe) != (wq->rq.msn)))) {
+ if (unlikely(!CQE_STATUS(hw_cqe) &&
+ CQE_WRID_MSN(hw_cqe) != wq->rq.msn)) {
t4_set_wq_in_error(wq);
- hw_cqe->header |= htonl(V_CQE_STATUS(T4_ERR_MSN));
- goto proc_cqe;
+ hw_cqe->header |= cpu_to_be32(V_CQE_STATUS(T4_ERR_MSN));
}
goto proc_cqe;
}