net/sfc: implement simple EF10 native Tx datapath

The datapath does not support VLAN insertion, TSO and multi-segment
mbufs.

Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>

diff --git a/doc/guides/nics/sfc_efx.rst b/doc/guides/nics/sfc_efx.rst
index 5c96625..5f825e9 100644 (file)
--- a/doc/guides/nics/sfc_efx.rst
+++ b/doc/guides/nics/sfc_efx.rst
@@ -236,7 +236,7 @@ boolean parameters value.
   more efficient than libefx-based and provides richer packet type
   classification, but lacks Rx scatter support.
 
-- ``tx_datapath`` [auto|efx|ef10] (default **auto**)
+- ``tx_datapath`` [auto|efx|ef10|ef10_simple] (default **auto**)
 
   Choose transmit datapath implementation.
   **auto** allows the driver itself to make a choice based on firmware
@@ -246,6 +246,9 @@ boolean parameters value.
   **ef10** chooses EF10 (SFN7xxx, SFN8xxx) native datapath which is
   more efficient than libefx-based but has no VLAN insertion and TSO
   support yet.
+  **ef10_simple** chooses EF10 (SFN7xxx, SFN8xxx) native datapath which
+  is even more faster then **ef10** but does not support multi-segment
+  mbufs.
 
 - ``perf_profile`` [auto|throughput|low-latency] (default **throughput**)
 

diff --git a/drivers/net/sfc/sfc_dp_tx.h b/drivers/net/sfc/sfc_dp_tx.h
index f99a53d..2bb9a2e 100644 (file)
--- a/drivers/net/sfc/sfc_dp_tx.h
+++ b/drivers/net/sfc/sfc_dp_tx.h
@@ -162,6 +162,7 @@ sfc_dp_find_tx_by_caps(struct sfc_dp_list *head, unsigned int avail_caps)
 
 extern struct sfc_dp_tx sfc_efx_tx;
 extern struct sfc_dp_tx sfc_ef10_tx;
+extern struct sfc_dp_tx sfc_ef10_simple_tx;
 
 #ifdef __cplusplus
 }

diff --git a/drivers/net/sfc/sfc_ef10_tx.c b/drivers/net/sfc/sfc_ef10_tx.c
index 1ef198a..74bd822 100644 (file)
--- a/drivers/net/sfc/sfc_ef10_tx.c
+++ b/drivers/net/sfc/sfc_ef10_tx.c
@@ -313,6 +313,64 @@ sfc_ef10_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
        return pktp - &tx_pkts[0];
 }
 
+static uint16_t
+sfc_ef10_simple_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+                         uint16_t nb_pkts)
+{
+       struct sfc_ef10_txq * const txq = sfc_ef10_txq_by_dp_txq(tx_queue);
+       unsigned int ptr_mask;
+       unsigned int added;
+       unsigned int dma_desc_space;
+       bool reap_done;
+       struct rte_mbuf **pktp;
+       struct rte_mbuf **pktp_end;
+
+       if (unlikely(txq->flags &
+                    (SFC_EF10_TXQ_NOT_RUNNING | SFC_EF10_TXQ_EXCEPTION)))
+               return 0;
+
+       ptr_mask = txq->ptr_mask;
+       added = txq->added;
+       dma_desc_space = SFC_EF10_TXQ_LIMIT(ptr_mask + 1) -
+                        (added - txq->completed);
+
+       reap_done = (dma_desc_space < RTE_MAX(txq->free_thresh, nb_pkts));
+       if (reap_done) {
+               sfc_ef10_tx_reap(txq);
+               dma_desc_space = SFC_EF10_TXQ_LIMIT(ptr_mask + 1) -
+                                (added - txq->completed);
+       }
+
+       pktp_end = &tx_pkts[MIN(nb_pkts, dma_desc_space)];
+       for (pktp = &tx_pkts[0]; pktp != pktp_end; ++pktp) {
+               struct rte_mbuf *pkt = *pktp;
+               unsigned int id = added & ptr_mask;
+
+               SFC_ASSERT(rte_pktmbuf_data_len(pkt) <=
+                          SFC_EF10_TX_DMA_DESC_LEN_MAX);
+
+               sfc_ef10_tx_qdesc_dma_create(rte_mbuf_data_dma_addr(pkt),
+                                            rte_pktmbuf_data_len(pkt),
+                                            true, &txq->txq_hw_ring[id]);
+
+               txq->sw_ring[id].mbuf = pkt;
+
+               ++added;
+       }
+
+       if (likely(added != txq->added)) {
+               sfc_ef10_tx_qpush(txq, added, txq->added);
+               txq->added = added;
+       }
+
+#if SFC_TX_XMIT_PKTS_REAP_AT_LEAST_ONCE
+       if (!reap_done)
+               sfc_ef10_tx_reap(txq);
+#endif
+
+       return pktp - &tx_pkts[0];
+}
+
 
 static sfc_dp_tx_qcreate_t sfc_ef10_tx_qcreate;
 static int
@@ -449,3 +507,18 @@ struct sfc_dp_tx sfc_ef10_tx = {
        .qreap                  = sfc_ef10_tx_qreap,
        .pkt_burst              = sfc_ef10_xmit_pkts,
 };
+
+struct sfc_dp_tx sfc_ef10_simple_tx = {
+       .dp = {
+               .name           = SFC_KVARG_DATAPATH_EF10_SIMPLE,
+               .type           = SFC_DP_TX,
+       },
+       .features               = 0,
+       .qcreate                = sfc_ef10_tx_qcreate,
+       .qdestroy               = sfc_ef10_tx_qdestroy,
+       .qstart                 = sfc_ef10_tx_qstart,
+       .qtx_ev                 = sfc_ef10_tx_qtx_ev,
+       .qstop                  = sfc_ef10_tx_qstop,
+       .qreap                  = sfc_ef10_tx_qreap,
+       .pkt_burst              = sfc_ef10_simple_xmit_pkts,
+};

diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c
index 8f36450..bba8b01 100644 (file)
--- a/drivers/net/sfc/sfc_ethdev.c
+++ b/drivers/net/sfc/sfc_ethdev.c
@@ -1470,6 +1470,7 @@ sfc_register_dp(void)
 
                sfc_dp_register(&sfc_dp_head, &sfc_ef10_tx.dp);
                sfc_dp_register(&sfc_dp_head, &sfc_efx_tx.dp);
+               sfc_dp_register(&sfc_dp_head, &sfc_ef10_simple_tx.dp);
        }
 }
 

diff --git a/drivers/net/sfc/sfc_kvargs.h b/drivers/net/sfc/sfc_kvargs.h
index e38ec17..d9c3b1d 100644 (file)
--- a/drivers/net/sfc/sfc_kvargs.h
+++ b/drivers/net/sfc/sfc_kvargs.h
@@ -58,6 +58,7 @@ extern "C" {
 
 #define SFC_KVARG_DATAPATH_EFX         "efx"
 #define SFC_KVARG_DATAPATH_EF10                "ef10"
+#define SFC_KVARG_DATAPATH_EF10_SIMPLE "ef10_simple"
 
 #define SFC_KVARG_RX_DATAPATH          "rx_datapath"
 #define SFC_KVARG_VALUES_RX_DATAPATH \
@@ -67,7 +68,8 @@ extern "C" {
 #define SFC_KVARG_TX_DATAPATH          "tx_datapath"
 #define SFC_KVARG_VALUES_TX_DATAPATH \
        "[" SFC_KVARG_DATAPATH_EFX "|" \
-           SFC_KVARG_DATAPATH_EF10 "]"
+           SFC_KVARG_DATAPATH_EF10 "|" \
+           SFC_KVARG_DATAPATH_EF10_SIMPLE "]"
 
 struct sfc_adapter;