On Mon, Sep 03, 2018 at 06:19:38PM +0800, Yanjun Zhu wrote:
>
>
> On 2018/6/14 9:47, Vijay Immanuel wrote:
> >Error retries can occur due to timeouts, NAKs or receiving
> >packets beyond the current read request. Avoid back-to-back
> >retries due to packet processing, by only retrying the initial
> >attempt immediately. Subsequent retries must be due to timeouts.
> >
> >Continue to process completion packets after scheduling a retry.
> >
> >Signed-off-by: Vijay Immanuel <vijayi@xxxxxxxxxxxxxxxxx>
> >---
> > drivers/infiniband/sw/rxe/rxe_comp.c | 19 ++++++++++++++++++-
> > drivers/infiniband/sw/rxe/rxe_verbs.h | 1 +
> > 2 files changed, 19 insertions(+), 1 deletion(-)
> >
> >diff --git a/drivers/infiniband/sw/rxe/rxe_comp.c b/drivers/infiniband/sw/rxe/rxe_comp.c
> >index 6cdc40e..b1d5624 100644
> >--- a/drivers/infiniband/sw/rxe/rxe_comp.c
> >+++ b/drivers/infiniband/sw/rxe/rxe_comp.c
> >@@ -191,6 +191,7 @@ static inline void reset_retry_counters(struct rxe_qp *qp)
> > {
> > qp->comp.retry_cnt = qp->attr.retry_cnt;
> > qp->comp.rnr_retry = qp->attr.rnr_retry;
> >+ qp->comp.started_retry = 0;
> > }
> > static inline enum comp_state check_psn(struct rxe_qp *qp,
> >@@ -679,6 +680,21 @@ int rxe_completer(void *arg)
> > goto exit;
> > }
> >+ /* if we've started a retry, don't start another
> >+ * retry sequence, unless this is a timeout.
> >+ */
>
> Sorry. Is it possible that several SKBs require retry at almost the same
> time?
>
Yes, they will all get retried. We only need to set the qp->req.need_retry
once. rxe_requester() will retry the send queue, when it sees that flag.
> >+ if (qp->comp.started_retry &&
> >+ !qp->comp.timeout_retry) {
> >+ if (pkt) {
> >+ rxe_drop_ref(pkt->qp);
> >+ kfree_skb(skb);
> >+ skb = NULL;
> >+ }
> >+
> >+ WARN_ON_ONCE(skb);
> >+ goto done;
> >+ }
> >+
> > if (qp->comp.retry_cnt > 0) {
> > if (qp->comp.retry_cnt != 7)
> > qp->comp.retry_cnt--;
> >@@ -695,6 +711,7 @@ int rxe_completer(void *arg)
> > rxe_counter_inc(rxe,
> > RXE_CNT_COMP_RETRY);
> > qp->req.need_retry = 1;
> >+ qp->comp.started_retry = 1;
> > rxe_run_task(&qp->req.task, 1);
> > }
> >@@ -705,7 +722,7 @@ int rxe_completer(void *arg)
> > }
> > WARN_ON_ONCE(skb);
> >- goto exit;
> >+ goto done;
> > } else {
> > rxe_counter_inc(rxe, RXE_CNT_RETRY_EXCEEDED);
> >diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
> >index af1470d..51a515b 100644
> >--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
> >+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
> >@@ -158,6 +158,7 @@ struct rxe_comp_info {
> > int opcode;
> > int timeout;
> > int timeout_retry;
> >+ int started_retry;
> > u32 retry_cnt;
> > u32 rnr_retry;
> > struct rxe_task task;
>
