Re: [dpdk-dev] [PATCH 1/9] net/sfc: receive prepared packets even in Rx exception case

Re: [dpdk-dev] [PATCH 1/9] net/sfc: receive prepared packets even in Rx exception case

[Ivan Malov]

On 10/3/2018 12:03 PM, Andrew Rybchenko wrote:
> Make sure that number of prepared packets, completed and added
> Rx ring pointers are reset to zeros on queue purge at stop.
> 
> Fixes: 638bddc99faa ("net/sfc: implement EF10 native Rx datapath")
> Cc: stable at dpdk.org
> 
> Signed-off-by: Andrew Rybchenko <arybchenko at solarflare.com>

Reviewed-by: Ivan Malov <Ivan.Malov@oktetlabs.ru>

[dpdk-dev] [PATCH 0/9] net/sfc: support more features in EF10 Rx

[Andrew Rybchenko]

Andrew Rybchenko (8):
  net/sfc: receive prepared packets even in Rx exception case
  net/sfc: use mbuf raw free instead of mempool put directly
  net/sfc: check mbufs allocated using mempool API for Rx
  net/sfc: avoid dummy writes to Rx queue state structure
  net/sfc: decrease number of variables maintained on EF10 Rx
  net/sfc: avoid usage of prepared packets number in EF10 Rx
  net/sfc: rename variable to prepare for scatter support
  net/sfc: support Rx scatter in EF10 Rx datapath

Igor Romanov (1):
  net/sfc: support Rx descriptor status on EF10 datapath

 doc/guides/nics/sfc_efx.rst            |   2 +-
 doc/guides/rel_notes/release_18_11.rst |   7 +
 drivers/net/sfc/sfc_ef10_essb_rx.c     |  16 +-
 drivers/net/sfc/sfc_ef10_rx.c          | 227 +++++++++++++++++--------
 drivers/net/sfc/sfc_rx.c               |   9 +-
 5 files changed, 184 insertions(+), 77 deletions(-)

-- 
2.17.1

[dpdk-dev] [PATCH 1/9] net/sfc: receive prepared packets even in Rx exception case

[Andrew Rybchenko]

Make sure that number of prepared packets, completed and added
Rx ring pointers are reset to zeros on queue purge at stop.

Fixes: 638bddc99faa ("net/sfc: implement EF10 native Rx datapath")
Cc: stable@dpdk.org

Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
---
 drivers/net/sfc/sfc_ef10_rx.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/net/sfc/sfc_ef10_rx.c b/drivers/net/sfc/sfc_ef10_rx.c
index 6a5052b93..5c0152d01 100644
--- a/drivers/net/sfc/sfc_ef10_rx.c
+++ b/drivers/net/sfc/sfc_ef10_rx.c
@@ -359,11 +359,11 @@ sfc_ef10_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 	uint16_t n_rx_pkts;
 	efx_qword_t rx_ev;
 
+	n_rx_pkts = sfc_ef10_rx_prepared(rxq, rx_pkts, nb_pkts);
+
 	if (unlikely(rxq->flags &
 		     (SFC_EF10_RXQ_NOT_RUNNING | SFC_EF10_RXQ_EXCEPTION)))
-		return 0;
-
-	n_rx_pkts = sfc_ef10_rx_prepared(rxq, rx_pkts, nb_pkts);
+		goto done;
 
 	evq_old_read_ptr = rxq->evq_read_ptr;
 	while (n_rx_pkts != nb_pkts && sfc_ef10_rx_get_event(rxq, &rx_ev)) {
@@ -383,6 +383,7 @@ sfc_ef10_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
 	/* It is not a problem if we refill in the case of exception */
 	sfc_ef10_rx_qrefill(rxq);
 
+done:
 	return n_rx_pkts;
 }
 
@@ -594,8 +595,9 @@ sfc_ef10_rx_qstart(struct sfc_dp_rxq *dp_rxq, unsigned int evq_read_ptr)
 {
 	struct sfc_ef10_rxq *rxq = sfc_ef10_rxq_by_dp_rxq(dp_rxq);
 
-	rxq->prepared = 0;
-	rxq->completed = rxq->added = 0;
+	SFC_ASSERT(rxq->prepared == 0);
+	SFC_ASSERT(rxq->completed == 0);
+	SFC_ASSERT(rxq->added == 0);
 
 	sfc_ef10_rx_qrefill(rxq);
 
@@ -642,12 +644,16 @@ sfc_ef10_rx_qpurge(struct sfc_dp_rxq *dp_rxq)
 	unsigned int i;
 	struct sfc_ef10_rx_sw_desc *rxd;
 
+	rxq->prepared = 0;
+
 	for (i = rxq->completed; i != rxq->added; ++i) {
 		rxd = &rxq->sw_ring[i & rxq->ptr_mask];
 		rte_mempool_put(rxq->refill_mb_pool, rxd->mbuf);
 		rxd->mbuf = NULL;
 	}
 
+	rxq->completed = rxq->added = 0;
+
 	rxq->flags &= ~SFC_EF10_RXQ_STARTED;
 }
 
-- 
2.17.1