DPDK patches and discussions
 help / color / mirror / Atom feed
* [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD
@ 2021-07-16  8:23 Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
                   ` (7 more replies)
  0 siblings, 8 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

This patch set restructures the NFP PMD, aligning it more with the
common layout adopted by most other PMD's. Although the changes look
fairly large, functionally nothing is added or removed from the driver
and the existing code is mostly just reorganized into the familiar
structure seen in other PMD's. Apart form adopting the common PMD layout
this change should also aid in future feature development to the NFP
PMD. The previous layout where most of the logic resided in a single
file (nfp_net.c) would have become tedious to support going forward.

Heinrich Kuhn (7):
  net/nfp: split rxtx headers into separate file
  net/nfp: move rxtx functions to their own file
  net/nfp: move CPP bridge to a separate file
  net/nfp: prototype common functions in header file
  net/nfp: move VF functions into new file
  net/nfp: move PF functions into new file
  net/nfp: batch file rename for consistency

 drivers/net/nfp/meson.build                   |    6 +-
 drivers/net/nfp/nfp_common.c                  | 1322 ++++++
 drivers/net/nfp/nfp_common.h                  |  413 ++
 drivers/net/nfp/nfp_cpp_bridge.c              |  392 ++
 drivers/net/nfp/nfp_cpp_bridge.h              |   36 +
 .../net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h}    |    6 +-
 drivers/net/nfp/nfp_ethdev.c                  | 1099 +++++
 drivers/net/nfp/nfp_ethdev_vf.c               |  504 +++
 .../net/nfp/{nfp_net_logs.h => nfp_logs.h}    |    6 +-
 drivers/net/nfp/nfp_net.c                     | 3921 -----------------
 drivers/net/nfp/nfp_rxtx.c                    | 1002 +++++
 drivers/net/nfp/{nfp_net_pmd.h => nfp_rxtx.h} |  279 +-
 12 files changed, 4815 insertions(+), 4171 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_common.c
 create mode 100644 drivers/net/nfp/nfp_common.h
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.c
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.h
 rename drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h} (99%)
 create mode 100644 drivers/net/nfp/nfp_ethdev.c
 create mode 100644 drivers/net/nfp/nfp_ethdev_vf.c
 rename drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h} (94%)
 delete mode 100644 drivers/net/nfp/nfp_net.c
 create mode 100644 drivers/net/nfp/nfp_rxtx.c
 rename drivers/net/nfp/{nfp_net_pmd.h => nfp_rxtx.h} (54%)

-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH 1/7] net/nfp: split rxtx headers into separate file
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
@ 2021-07-16  8:23 ` Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
                   ` (6 subsequent siblings)
  7 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

This change splits out the rx/tx specific structs and defines from the
main nfp_net_pmd header file and into their own header file.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
---
 drivers/net/nfp/nfp_net.c     |   1 +
 drivers/net/nfp/nfp_net_pmd.h | 248 ------------------------------
 drivers/net/nfp/nfp_rxtx.h    | 276 ++++++++++++++++++++++++++++++++++
 3 files changed, 277 insertions(+), 248 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_rxtx.h

diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index b18edd8c7b..67288abeff 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -38,6 +38,7 @@
 #include "nfpcore/nfp_nsp.h"
 
 #include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
 #include "nfp_net_logs.h"
 #include "nfp_net_ctrl.h"
 
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index 212f9ef162..a3a3ba32d6 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -23,19 +23,6 @@
 /* Forward declaration */
 struct nfp_net_adapter;
 
-/*
- * The maximum number of descriptors is limited by design as
- * DPDK uses uint16_t variables for these values
- */
-#define NFP_NET_MAX_TX_DESC (32 * 1024)
-#define NFP_NET_MIN_TX_DESC 64
-
-#define NFP_NET_MAX_RX_DESC (32 * 1024)
-#define NFP_NET_MIN_RX_DESC 64
-
-/* Descriptor alignment */
-#define NFP_ALIGN_RING_DESC 128
-
 #define NFP_TX_MAX_SEG     UINT8_MAX
 #define NFP_TX_MAX_MTU_SEG 8
 
@@ -150,241 +137,6 @@ static inline void nn_writeq(uint64_t val, volatile void *addr)
 	nn_writel(val, addr);
 }
 
-/* TX descriptor format */
-#define PCIE_DESC_TX_EOP                (1 << 7)
-#define PCIE_DESC_TX_OFFSET_MASK        (0x7f)
-
-/* Flags in the host TX descriptor */
-#define PCIE_DESC_TX_CSUM               (1 << 7)
-#define PCIE_DESC_TX_IP4_CSUM           (1 << 6)
-#define PCIE_DESC_TX_TCP_CSUM           (1 << 5)
-#define PCIE_DESC_TX_UDP_CSUM           (1 << 4)
-#define PCIE_DESC_TX_VLAN               (1 << 3)
-#define PCIE_DESC_TX_LSO                (1 << 2)
-#define PCIE_DESC_TX_ENCAP_NONE         (0)
-#define PCIE_DESC_TX_ENCAP_VXLAN        (1 << 1)
-#define PCIE_DESC_TX_ENCAP_GRE          (1 << 0)
-
-struct nfp_net_tx_desc {
-	union {
-		struct {
-			uint8_t dma_addr_hi; /* High bits of host buf address */
-			__le16 dma_len;     /* Length to DMA for this desc */
-			uint8_t offset_eop; /* Offset in buf where pkt starts +
-					     * highest bit is eop flag.
-					     */
-			__le32 dma_addr_lo; /* Low 32bit of host buf addr */
-
-			__le16 mss;         /* MSS to be used for LSO */
-			uint8_t lso_hdrlen; /* LSO, where the data starts */
-			uint8_t flags;      /* TX Flags, see @PCIE_DESC_TX_* */
-
-			union {
-				struct {
-					/*
-					 * L3 and L4 header offsets required
-					 * for TSOv2
-					 */
-					uint8_t l3_offset;
-					uint8_t l4_offset;
-				};
-				__le16 vlan; /* VLAN tag to add if indicated */
-			};
-			__le16 data_len;    /* Length of frame + meta data */
-		} __rte_packed;
-		__le32 vals[4];
-	};
-};
-
-struct nfp_net_txq {
-	struct nfp_net_hw *hw; /* Backpointer to nfp_net structure */
-
-	/*
-	 * Queue information: @qidx is the queue index from Linux's
-	 * perspective.  @tx_qcidx is the index of the Queue
-	 * Controller Peripheral queue relative to the TX queue BAR.
-	 * @cnt is the size of the queue in number of
-	 * descriptors. @qcp_q is a pointer to the base of the queue
-	 * structure on the NFP
-	 */
-	uint8_t *qcp_q;
-
-	/*
-	 * Read and Write pointers.  @wr_p and @rd_p are host side pointer,
-	 * they are free running and have little relation to the QCP pointers *
-	 * @qcp_rd_p is a local copy queue controller peripheral read pointer
-	 */
-
-	uint32_t wr_p;
-	uint32_t rd_p;
-
-	uint32_t tx_count;
-
-	uint32_t tx_free_thresh;
-
-	/*
-	 * For each descriptor keep a reference to the mbuf and
-	 * DMA address used until completion is signalled.
-	 */
-	struct {
-		struct rte_mbuf *mbuf;
-	} *txbufs;
-
-	/*
-	 * Information about the host side queue location. @txds is
-	 * the virtual address for the queue, @dma is the DMA address
-	 * of the queue and @size is the size in bytes for the queue
-	 * (needed for free)
-	 */
-	struct nfp_net_tx_desc *txds;
-
-	/*
-	 * At this point 48 bytes have been used for all the fields in the
-	 * TX critical path. We have room for 8 bytes and still all placed
-	 * in a cache line. We are not using the threshold values below but
-	 * if we need to, we can add the most used in the remaining bytes.
-	 */
-	uint32_t tx_rs_thresh; /* not used by now. Future? */
-	uint32_t tx_pthresh;   /* not used by now. Future? */
-	uint32_t tx_hthresh;   /* not used by now. Future? */
-	uint32_t tx_wthresh;   /* not used by now. Future? */
-	uint16_t port_id;
-	int qidx;
-	int tx_qcidx;
-	__le64 dma;
-} __rte_aligned(64);
-
-/* RX and freelist descriptor format */
-#define PCIE_DESC_RX_DD                 (1 << 7)
-#define PCIE_DESC_RX_META_LEN_MASK      (0x7f)
-
-/* Flags in the RX descriptor */
-#define PCIE_DESC_RX_RSS                (1 << 15)
-#define PCIE_DESC_RX_I_IP4_CSUM         (1 << 14)
-#define PCIE_DESC_RX_I_IP4_CSUM_OK      (1 << 13)
-#define PCIE_DESC_RX_I_TCP_CSUM         (1 << 12)
-#define PCIE_DESC_RX_I_TCP_CSUM_OK      (1 << 11)
-#define PCIE_DESC_RX_I_UDP_CSUM         (1 << 10)
-#define PCIE_DESC_RX_I_UDP_CSUM_OK      (1 <<  9)
-#define PCIE_DESC_RX_SPARE              (1 <<  8)
-#define PCIE_DESC_RX_EOP                (1 <<  7)
-#define PCIE_DESC_RX_IP4_CSUM           (1 <<  6)
-#define PCIE_DESC_RX_IP4_CSUM_OK        (1 <<  5)
-#define PCIE_DESC_RX_TCP_CSUM           (1 <<  4)
-#define PCIE_DESC_RX_TCP_CSUM_OK        (1 <<  3)
-#define PCIE_DESC_RX_UDP_CSUM           (1 <<  2)
-#define PCIE_DESC_RX_UDP_CSUM_OK        (1 <<  1)
-#define PCIE_DESC_RX_VLAN               (1 <<  0)
-
-#define PCIE_DESC_RX_L4_CSUM_OK         (PCIE_DESC_RX_TCP_CSUM_OK | \
-					 PCIE_DESC_RX_UDP_CSUM_OK)
-struct nfp_net_rx_desc {
-	union {
-		/* Freelist descriptor */
-		struct {
-			uint8_t dma_addr_hi;
-			__le16 spare;
-			uint8_t dd;
-
-			__le32 dma_addr_lo;
-		} __rte_packed fld;
-
-		/* RX descriptor */
-		struct {
-			__le16 data_len;
-			uint8_t reserved;
-			uint8_t meta_len_dd;
-
-			__le16 flags;
-			__le16 vlan;
-		} __rte_packed rxd;
-
-		__le32 vals[2];
-	};
-};
-
-struct nfp_net_rx_buff {
-	struct rte_mbuf *mbuf;
-};
-
-struct nfp_net_rxq {
-	struct nfp_net_hw *hw;	/* Backpointer to nfp_net structure */
-
-	 /*
-	  * @qcp_fl and @qcp_rx are pointers to the base addresses of the
-	  * freelist and RX queue controller peripheral queue structures on the
-	  * NFP
-	  */
-	uint8_t *qcp_fl;
-	uint8_t *qcp_rx;
-
-	/*
-	 * Read and Write pointers.  @wr_p and @rd_p are host side
-	 * pointer, they are free running and have little relation to
-	 * the QCP pointers. @wr_p is where the driver adds new
-	 * freelist descriptors and @rd_p is where the driver start
-	 * reading descriptors for newly arrive packets from.
-	 */
-	uint32_t rd_p;
-
-	/*
-	 * For each buffer placed on the freelist, record the
-	 * associated SKB
-	 */
-	struct nfp_net_rx_buff *rxbufs;
-
-	/*
-	 * Information about the host side queue location.  @rxds is
-	 * the virtual address for the queue
-	 */
-	struct nfp_net_rx_desc *rxds;
-
-	/*
-	 * The mempool is created by the user specifying a mbuf size.
-	 * We save here the reference of the mempool needed in the RX
-	 * path and the mbuf size for checking received packets can be
-	 * safely copied to the mbuf using the NFP_NET_RX_OFFSET
-	 */
-	struct rte_mempool *mem_pool;
-	uint16_t mbuf_size;
-
-	/*
-	 * Next two fields are used for giving more free descriptors
-	 * to the NFP
-	 */
-	uint16_t rx_free_thresh;
-	uint16_t nb_rx_hold;
-
-	 /* the size of the queue in number of descriptors */
-	uint16_t rx_count;
-
-	/*
-	 * Fields above this point fit in a single cache line and are all used
-	 * in the RX critical path. Fields below this point are just used
-	 * during queue configuration or not used at all (yet)
-	 */
-
-	/* referencing dev->data->port_id */
-	uint16_t port_id;
-
-	uint8_t  crc_len; /* Not used by now */
-	uint8_t  drop_en; /* Not used by now */
-
-	/* DMA address of the queue */
-	__le64 dma;
-
-	/*
-	 * Queue information: @qidx is the queue index from Linux's
-	 * perspective.  @fl_qcidx is the index of the Queue
-	 * Controller peripheral queue relative to the RX queue BAR
-	 * used for the freelist and @rx_qcidx is the Queue Controller
-	 * Peripheral index for the RX queue.
-	 */
-	int qidx;
-	int fl_qcidx;
-	int rx_qcidx;
-} __rte_aligned(64);
-
 struct nfp_pf_dev {
 	/* Backpointer to associated pci device */
 	struct rte_pci_device *pci_dev;
diff --git a/drivers/net/nfp/nfp_rxtx.h b/drivers/net/nfp/nfp_rxtx.h
new file mode 100644
index 0000000000..41a3a4b4e7
--- /dev/null
+++ b/drivers/net/nfp/nfp_rxtx.h
@@ -0,0 +1,276 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_rxtx.h
+ *
+ * Netronome NFP Rx/Tx specific header file
+ */
+
+#ifndef _NFP_RXTX_H_
+#define _NFP_RXTX_H_
+
+#include <linux/types.h>
+#include <rte_io.h>
+
+/*
+ * The maximum number of descriptors is limited by design as
+ * DPDK uses uint16_t variables for these values
+ */
+#define NFP_NET_MAX_TX_DESC (32 * 1024)
+#define NFP_NET_MIN_TX_DESC 64
+
+#define NFP_NET_MAX_RX_DESC (32 * 1024)
+#define NFP_NET_MIN_RX_DESC 64
+
+/* Descriptor alignment */
+#define NFP_ALIGN_RING_DESC 128
+
+/* TX descriptor format */
+#define PCIE_DESC_TX_EOP                (1 << 7)
+#define PCIE_DESC_TX_OFFSET_MASK        (0x7f)
+
+/* Flags in the host TX descriptor */
+#define PCIE_DESC_TX_CSUM               (1 << 7)
+#define PCIE_DESC_TX_IP4_CSUM           (1 << 6)
+#define PCIE_DESC_TX_TCP_CSUM           (1 << 5)
+#define PCIE_DESC_TX_UDP_CSUM           (1 << 4)
+#define PCIE_DESC_TX_VLAN               (1 << 3)
+#define PCIE_DESC_TX_LSO                (1 << 2)
+#define PCIE_DESC_TX_ENCAP_NONE         (0)
+#define PCIE_DESC_TX_ENCAP_VXLAN        (1 << 1)
+#define PCIE_DESC_TX_ENCAP_GRE          (1 << 0)
+
+struct nfp_net_tx_desc {
+	union {
+		struct {
+			uint8_t dma_addr_hi; /* High bits of host buf address */
+			__le16 dma_len;     /* Length to DMA for this desc */
+			uint8_t offset_eop; /* Offset in buf where pkt starts +
+					     * highest bit is eop flag.
+					     */
+			__le32 dma_addr_lo; /* Low 32bit of host buf addr */
+
+			__le16 mss;         /* MSS to be used for LSO */
+			uint8_t lso_hdrlen; /* LSO, where the data starts */
+			uint8_t flags;      /* TX Flags, see @PCIE_DESC_TX_* */
+
+			union {
+				struct {
+					/*
+					 * L3 and L4 header offsets required
+					 * for TSOv2
+					 */
+					uint8_t l3_offset;
+					uint8_t l4_offset;
+				};
+				__le16 vlan; /* VLAN tag to add if indicated */
+			};
+			__le16 data_len;    /* Length of frame + meta data */
+		} __rte_packed;
+		__le32 vals[4];
+	};
+};
+
+struct nfp_net_txq {
+	struct nfp_net_hw *hw; /* Backpointer to nfp_net structure */
+
+	/*
+	 * Queue information: @qidx is the queue index from Linux's
+	 * perspective.  @tx_qcidx is the index of the Queue
+	 * Controller Peripheral queue relative to the TX queue BAR.
+	 * @cnt is the size of the queue in number of
+	 * descriptors. @qcp_q is a pointer to the base of the queue
+	 * structure on the NFP
+	 */
+	uint8_t *qcp_q;
+
+	/*
+	 * Read and Write pointers.  @wr_p and @rd_p are host side pointer,
+	 * they are free running and have little relation to the QCP pointers *
+	 * @qcp_rd_p is a local copy queue controller peripheral read pointer
+	 */
+
+	uint32_t wr_p;
+	uint32_t rd_p;
+
+	uint32_t tx_count;
+
+	uint32_t tx_free_thresh;
+
+	/*
+	 * For each descriptor keep a reference to the mbuf and
+	 * DMA address used until completion is signalled.
+	 */
+	struct {
+		struct rte_mbuf *mbuf;
+	} *txbufs;
+
+	/*
+	 * Information about the host side queue location. @txds is
+	 * the virtual address for the queue, @dma is the DMA address
+	 * of the queue and @size is the size in bytes for the queue
+	 * (needed for free)
+	 */
+	struct nfp_net_tx_desc *txds;
+
+	/*
+	 * At this point 48 bytes have been used for all the fields in the
+	 * TX critical path. We have room for 8 bytes and still all placed
+	 * in a cache line. We are not using the threshold values below but
+	 * if we need to, we can add the most used in the remaining bytes.
+	 */
+	uint32_t tx_rs_thresh; /* not used by now. Future? */
+	uint32_t tx_pthresh;   /* not used by now. Future? */
+	uint32_t tx_hthresh;   /* not used by now. Future? */
+	uint32_t tx_wthresh;   /* not used by now. Future? */
+	uint16_t port_id;
+	int qidx;
+	int tx_qcidx;
+	__le64 dma;
+} __rte_aligned(64);
+
+/* RX and freelist descriptor format */
+#define PCIE_DESC_RX_DD                 (1 << 7)
+#define PCIE_DESC_RX_META_LEN_MASK      (0x7f)
+
+/* Flags in the RX descriptor */
+#define PCIE_DESC_RX_RSS                (1 << 15)
+#define PCIE_DESC_RX_I_IP4_CSUM         (1 << 14)
+#define PCIE_DESC_RX_I_IP4_CSUM_OK      (1 << 13)
+#define PCIE_DESC_RX_I_TCP_CSUM         (1 << 12)
+#define PCIE_DESC_RX_I_TCP_CSUM_OK      (1 << 11)
+#define PCIE_DESC_RX_I_UDP_CSUM         (1 << 10)
+#define PCIE_DESC_RX_I_UDP_CSUM_OK      (1 <<  9)
+#define PCIE_DESC_RX_SPARE              (1 <<  8)
+#define PCIE_DESC_RX_EOP                (1 <<  7)
+#define PCIE_DESC_RX_IP4_CSUM           (1 <<  6)
+#define PCIE_DESC_RX_IP4_CSUM_OK        (1 <<  5)
+#define PCIE_DESC_RX_TCP_CSUM           (1 <<  4)
+#define PCIE_DESC_RX_TCP_CSUM_OK        (1 <<  3)
+#define PCIE_DESC_RX_UDP_CSUM           (1 <<  2)
+#define PCIE_DESC_RX_UDP_CSUM_OK        (1 <<  1)
+#define PCIE_DESC_RX_VLAN               (1 <<  0)
+
+#define PCIE_DESC_RX_L4_CSUM_OK         (PCIE_DESC_RX_TCP_CSUM_OK | \
+					 PCIE_DESC_RX_UDP_CSUM_OK)
+
+struct nfp_net_rx_desc {
+	union {
+		/* Freelist descriptor */
+		struct {
+			uint8_t dma_addr_hi;
+			__le16 spare;
+			uint8_t dd;
+
+			__le32 dma_addr_lo;
+		} __rte_packed fld;
+
+		/* RX descriptor */
+		struct {
+			__le16 data_len;
+			uint8_t reserved;
+			uint8_t meta_len_dd;
+
+			__le16 flags;
+			__le16 vlan;
+		} __rte_packed rxd;
+
+		__le32 vals[2];
+	};
+};
+
+struct nfp_net_rx_buff {
+	struct rte_mbuf *mbuf;
+};
+
+struct nfp_net_rxq {
+	struct nfp_net_hw *hw;	/* Backpointer to nfp_net structure */
+
+	 /*
+	  * @qcp_fl and @qcp_rx are pointers to the base addresses of the
+	  * freelist and RX queue controller peripheral queue structures on the
+	  * NFP
+	  */
+	uint8_t *qcp_fl;
+	uint8_t *qcp_rx;
+
+	/*
+	 * Read and Write pointers.  @wr_p and @rd_p are host side
+	 * pointer, they are free running and have little relation to
+	 * the QCP pointers. @wr_p is where the driver adds new
+	 * freelist descriptors and @rd_p is where the driver start
+	 * reading descriptors for newly arrive packets from.
+	 */
+	uint32_t rd_p;
+
+	/*
+	 * For each buffer placed on the freelist, record the
+	 * associated SKB
+	 */
+	struct nfp_net_rx_buff *rxbufs;
+
+	/*
+	 * Information about the host side queue location.  @rxds is
+	 * the virtual address for the queue
+	 */
+	struct nfp_net_rx_desc *rxds;
+
+	/*
+	 * The mempool is created by the user specifying a mbuf size.
+	 * We save here the reference of the mempool needed in the RX
+	 * path and the mbuf size for checking received packets can be
+	 * safely copied to the mbuf using the NFP_NET_RX_OFFSET
+	 */
+	struct rte_mempool *mem_pool;
+	uint16_t mbuf_size;
+
+	/*
+	 * Next two fields are used for giving more free descriptors
+	 * to the NFP
+	 */
+	uint16_t rx_free_thresh;
+	uint16_t nb_rx_hold;
+
+	 /* the size of the queue in number of descriptors */
+	uint16_t rx_count;
+
+	/*
+	 * Fields above this point fit in a single cache line and are all used
+	 * in the RX critical path. Fields below this point are just used
+	 * during queue configuration or not used at all (yet)
+	 */
+
+	/* referencing dev->data->port_id */
+	uint16_t port_id;
+
+	uint8_t  crc_len; /* Not used by now */
+	uint8_t  drop_en; /* Not used by now */
+
+	/* DMA address of the queue */
+	__le64 dma;
+
+	/*
+	 * Queue information: @qidx is the queue index from Linux's
+	 * perspective.  @fl_qcidx is the index of the Queue
+	 * Controller peripheral queue relative to the RX queue BAR
+	 * used for the freelist and @rx_qcidx is the Queue Controller
+	 * Peripheral index for the RX queue.
+	 */
+	int qidx;
+	int fl_qcidx;
+	int rx_qcidx;
+} __rte_aligned(64);
+
+#endif /* _NFP_RXTX_H_ */
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
+
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH 2/7] net/nfp: move rxtx functions to their own file
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
@ 2021-07-16  8:23 ` Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
                   ` (5 subsequent siblings)
  7 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

Create a new rxtx file and move the Rx/Tx functions to this file. This
commit will also move the needed shared functions to the nfp_net_pmd.h
file as needed.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
---
 drivers/net/nfp/meson.build   |    1 +
 drivers/net/nfp/nfp_net.c     | 1090 ---------------------------------
 drivers/net/nfp/nfp_net_pmd.h |  184 ++++--
 drivers/net/nfp/nfp_rxtx.c    | 1002 ++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_rxtx.h    |   27 +
 5 files changed, 1173 insertions(+), 1131 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_rxtx.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index b51e2e5f20..1b289e2354 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -19,4 +19,5 @@ sources = files(
         'nfpcore/nfp_nsp_eth.c',
         'nfpcore/nfp_hwinfo.c',
         'nfp_net.c',
+        'nfp_rxtx.c',
 )
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 67288abeff..5bfc23ba04 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -66,29 +66,11 @@ static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
 static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
 static int nfp_net_promisc_enable(struct rte_eth_dev *dev);
 static int nfp_net_promisc_disable(struct rte_eth_dev *dev);
-static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
-static uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
-				       uint16_t queue_idx);
-static uint16_t nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
-				  uint16_t nb_pkts);
-static void nfp_net_rx_queue_release(void *rxq);
-static int nfp_net_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-				  uint16_t nb_desc, unsigned int socket_id,
-				  const struct rte_eth_rxconf *rx_conf,
-				  struct rte_mempool *mp);
-static int nfp_net_tx_free_bufs(struct nfp_net_txq *txq);
-static void nfp_net_tx_queue_release(void *txq);
-static int nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-				  uint16_t nb_desc, unsigned int socket_id,
-				  const struct rte_eth_txconf *tx_conf);
 static int nfp_net_start(struct rte_eth_dev *dev);
 static int nfp_net_stats_get(struct rte_eth_dev *dev,
 			      struct rte_eth_stats *stats);
 static int nfp_net_stats_reset(struct rte_eth_dev *dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
-static uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
-				  uint16_t nb_pkts);
-
 static int nfp_net_rss_config_default(struct rte_eth_dev *dev);
 static int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 				   struct rte_eth_rss_conf *rss_conf);
@@ -106,184 +88,6 @@ static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_eth_table *nfp_eth_table,
 			struct nfp_hwinfo *hwinfo);
 
-
-/* The offset of the queue controller queues in the PCIe Target */
-#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
-
-/* Maximum value which can be added to a queue with one transaction */
-#define NFP_QCP_MAX_ADD	0x7f
-
-#define RTE_MBUF_DMA_ADDR_DEFAULT(mb) \
-	(uint64_t)((mb)->buf_iova + RTE_PKTMBUF_HEADROOM)
-
-/* nfp_qcp_ptr - Read or Write Pointer of a queue */
-enum nfp_qcp_ptr {
-	NFP_QCP_READ_PTR = 0,
-	NFP_QCP_WRITE_PTR
-};
-
-/*
- * nfp_qcp_ptr_add - Add the value to the selected pointer of a queue
- * @q: Base address for queue structure
- * @ptr: Add to the Read or Write pointer
- * @val: Value to add to the queue pointer
- *
- * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed.
- */
-static inline void
-nfp_qcp_ptr_add(uint8_t *q, enum nfp_qcp_ptr ptr, uint32_t val)
-{
-	uint32_t off;
-
-	if (ptr == NFP_QCP_READ_PTR)
-		off = NFP_QCP_QUEUE_ADD_RPTR;
-	else
-		off = NFP_QCP_QUEUE_ADD_WPTR;
-
-	while (val > NFP_QCP_MAX_ADD) {
-		nn_writel(rte_cpu_to_le_32(NFP_QCP_MAX_ADD), q + off);
-		val -= NFP_QCP_MAX_ADD;
-	}
-
-	nn_writel(rte_cpu_to_le_32(val), q + off);
-}
-
-/*
- * nfp_qcp_read - Read the current Read/Write pointer value for a queue
- * @q:  Base address for queue structure
- * @ptr: Read or Write pointer
- */
-static inline uint32_t
-nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
-{
-	uint32_t off;
-	uint32_t val;
-
-	if (ptr == NFP_QCP_READ_PTR)
-		off = NFP_QCP_QUEUE_STS_LO;
-	else
-		off = NFP_QCP_QUEUE_STS_HI;
-
-	val = rte_cpu_to_le_32(nn_readl(q + off));
-
-	if (ptr == NFP_QCP_READ_PTR)
-		return val & NFP_QCP_QUEUE_STS_LO_READPTR_mask;
-	else
-		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
-}
-
-/*
- * Functions to read/write from/to Config BAR
- * Performs any endian conversion necessary.
- */
-static inline uint8_t
-nn_cfg_readb(struct nfp_net_hw *hw, int off)
-{
-	return nn_readb(hw->ctrl_bar + off);
-}
-
-static inline void
-nn_cfg_writeb(struct nfp_net_hw *hw, int off, uint8_t val)
-{
-	nn_writeb(val, hw->ctrl_bar + off);
-}
-
-static inline uint32_t
-nn_cfg_readl(struct nfp_net_hw *hw, int off)
-{
-	return rte_le_to_cpu_32(nn_readl(hw->ctrl_bar + off));
-}
-
-static inline void
-nn_cfg_writel(struct nfp_net_hw *hw, int off, uint32_t val)
-{
-	nn_writel(rte_cpu_to_le_32(val), hw->ctrl_bar + off);
-}
-
-static inline uint64_t
-nn_cfg_readq(struct nfp_net_hw *hw, int off)
-{
-	return rte_le_to_cpu_64(nn_readq(hw->ctrl_bar + off));
-}
-
-static inline void
-nn_cfg_writeq(struct nfp_net_hw *hw, int off, uint64_t val)
-{
-	nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
-}
-
-static void
-nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
-{
-	unsigned i;
-
-	if (rxq->rxbufs == NULL)
-		return;
-
-	for (i = 0; i < rxq->rx_count; i++) {
-		if (rxq->rxbufs[i].mbuf) {
-			rte_pktmbuf_free_seg(rxq->rxbufs[i].mbuf);
-			rxq->rxbufs[i].mbuf = NULL;
-		}
-	}
-}
-
-static void
-nfp_net_rx_queue_release(void *rx_queue)
-{
-	struct nfp_net_rxq *rxq = rx_queue;
-
-	if (rxq) {
-		nfp_net_rx_queue_release_mbufs(rxq);
-		rte_free(rxq->rxbufs);
-		rte_free(rxq);
-	}
-}
-
-static void
-nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
-{
-	nfp_net_rx_queue_release_mbufs(rxq);
-	rxq->rd_p = 0;
-	rxq->nb_rx_hold = 0;
-}
-
-static void
-nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq)
-{
-	unsigned i;
-
-	if (txq->txbufs == NULL)
-		return;
-
-	for (i = 0; i < txq->tx_count; i++) {
-		if (txq->txbufs[i].mbuf) {
-			rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
-			txq->txbufs[i].mbuf = NULL;
-		}
-	}
-}
-
-static void
-nfp_net_tx_queue_release(void *tx_queue)
-{
-	struct nfp_net_txq *txq = tx_queue;
-
-	if (txq) {
-		nfp_net_tx_queue_release_mbufs(txq);
-		rte_free(txq->txbufs);
-		rte_free(txq);
-	}
-}
-
-static void
-nfp_net_reset_tx_queue(struct nfp_net_txq *txq)
-{
-	nfp_net_tx_queue_release_mbufs(txq);
-	txq->wr_p = 0;
-	txq->rd_p = 0;
-}
-
 static int
 __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
 {
@@ -461,18 +265,6 @@ nfp_net_disable_queues(struct rte_eth_dev *dev)
 	hw->ctrl = new_ctrl;
 }
 
-static int
-nfp_net_rx_freelist_setup(struct rte_eth_dev *dev)
-{
-	int i;
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		if (nfp_net_rx_fill_freelist(dev->data->rx_queues[i]) < 0)
-			return -1;
-	}
-	return 0;
-}
-
 static void
 nfp_net_params_setup(struct nfp_net_hw *hw)
 {
@@ -1349,44 +1141,6 @@ nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 	return NULL;
 }
 
-static uint32_t
-nfp_net_rx_queue_count(struct rte_eth_dev *dev, uint16_t queue_idx)
-{
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_rx_desc *rxds;
-	uint32_t idx;
-	uint32_t count;
-
-	rxq = (struct nfp_net_rxq *)dev->data->rx_queues[queue_idx];
-
-	idx = rxq->rd_p;
-
-	count = 0;
-
-	/*
-	 * Other PMDs are just checking the DD bit in intervals of 4
-	 * descriptors and counting all four if the first has the DD
-	 * bit on. Of course, this is not accurate but can be good for
-	 * performance. But ideally that should be done in descriptors
-	 * chunks belonging to the same cache line
-	 */
-
-	while (count < rxq->rx_count) {
-		rxds = &rxq->rxds[idx];
-		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
-			break;
-
-		count++;
-		idx++;
-
-		/* Wrapping? */
-		if ((idx) == rxq->rx_count)
-			idx = 0;
-	}
-
-	return count;
-}
-
 static int
 nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
@@ -1568,850 +1322,6 @@ nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	return 0;
 }
 
-static int
-nfp_net_rx_queue_setup(struct rte_eth_dev *dev,
-		       uint16_t queue_idx, uint16_t nb_desc,
-		       unsigned int socket_id,
-		       const struct rte_eth_rxconf *rx_conf,
-		       struct rte_mempool *mp)
-{
-	const struct rte_memzone *tz;
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_hw *hw;
-	uint32_t rx_desc_sz;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	PMD_INIT_FUNC_TRACE();
-
-	/* Validating number of descriptors */
-	rx_desc_sz = nb_desc * sizeof(struct nfp_net_rx_desc);
-	if (rx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
-	    nb_desc > NFP_NET_MAX_RX_DESC ||
-	    nb_desc < NFP_NET_MIN_RX_DESC) {
-		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
-		return -EINVAL;
-	}
-
-	/*
-	 * Free memory prior to re-allocation if needed. This is the case after
-	 * calling nfp_net_stop
-	 */
-	if (dev->data->rx_queues[queue_idx]) {
-		nfp_net_rx_queue_release(dev->data->rx_queues[queue_idx]);
-		dev->data->rx_queues[queue_idx] = NULL;
-	}
-
-	/* Allocating rx queue data structure */
-	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct nfp_net_rxq),
-				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (rxq == NULL)
-		return -ENOMEM;
-
-	/* Hw queues mapping based on firmware configuration */
-	rxq->qidx = queue_idx;
-	rxq->fl_qcidx = queue_idx * hw->stride_rx;
-	rxq->rx_qcidx = rxq->fl_qcidx + (hw->stride_rx - 1);
-	rxq->qcp_fl = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->fl_qcidx);
-	rxq->qcp_rx = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->rx_qcidx);
-
-	/*
-	 * Tracking mbuf size for detecting a potential mbuf overflow due to
-	 * RX offset
-	 */
-	rxq->mem_pool = mp;
-	rxq->mbuf_size = rxq->mem_pool->elt_size;
-	rxq->mbuf_size -= (sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM);
-	hw->flbufsz = rxq->mbuf_size;
-
-	rxq->rx_count = nb_desc;
-	rxq->port_id = dev->data->port_id;
-	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
-	rxq->drop_en = rx_conf->rx_drop_en;
-
-	/*
-	 * Allocate RX ring hardware descriptors. A memzone large enough to
-	 * handle the maximum ring size is allocated in order to allow for
-	 * resizing in later calls to the queue setup function.
-	 */
-	tz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
-				   sizeof(struct nfp_net_rx_desc) *
-				   NFP_NET_MAX_RX_DESC, NFP_MEMZONE_ALIGN,
-				   socket_id);
-
-	if (tz == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating rx dma");
-		nfp_net_rx_queue_release(rxq);
-		return -ENOMEM;
-	}
-
-	/* Saving physical and virtual addresses for the RX ring */
-	rxq->dma = (uint64_t)tz->iova;
-	rxq->rxds = (struct nfp_net_rx_desc *)tz->addr;
-
-	/* mbuf pointers array for referencing mbufs linked to RX descriptors */
-	rxq->rxbufs = rte_zmalloc_socket("rxq->rxbufs",
-					 sizeof(*rxq->rxbufs) * nb_desc,
-					 RTE_CACHE_LINE_SIZE, socket_id);
-	if (rxq->rxbufs == NULL) {
-		nfp_net_rx_queue_release(rxq);
-		return -ENOMEM;
-	}
-
-	PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
-		   rxq->rxbufs, rxq->rxds, (unsigned long int)rxq->dma);
-
-	nfp_net_reset_rx_queue(rxq);
-
-	dev->data->rx_queues[queue_idx] = rxq;
-	rxq->hw = hw;
-
-	/*
-	 * Telling the HW about the physical address of the RX ring and number
-	 * of descriptors in log2 format
-	 */
-	nn_cfg_writeq(hw, NFP_NET_CFG_RXR_ADDR(queue_idx), rxq->dma);
-	nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), rte_log2_u32(nb_desc));
-
-	return 0;
-}
-
-static int
-nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq)
-{
-	struct nfp_net_rx_buff *rxe = rxq->rxbufs;
-	uint64_t dma_addr;
-	unsigned i;
-
-	PMD_RX_LOG(DEBUG, "nfp_net_rx_fill_freelist for %u descriptors",
-		   rxq->rx_count);
-
-	for (i = 0; i < rxq->rx_count; i++) {
-		struct nfp_net_rx_desc *rxd;
-		struct rte_mbuf *mbuf = rte_pktmbuf_alloc(rxq->mem_pool);
-
-		if (mbuf == NULL) {
-			PMD_DRV_LOG(ERR, "RX mbuf alloc failed queue_id=%u",
-				(unsigned)rxq->qidx);
-			return -ENOMEM;
-		}
-
-		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(mbuf));
-
-		rxd = &rxq->rxds[i];
-		rxd->fld.dd = 0;
-		rxd->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
-		rxd->fld.dma_addr_lo = dma_addr & 0xffffffff;
-		rxe[i].mbuf = mbuf;
-		PMD_RX_LOG(DEBUG, "[%d]: %" PRIx64, i, dma_addr);
-	}
-
-	/* Make sure all writes are flushed before telling the hardware */
-	rte_wmb();
-
-	/* Not advertising the whole ring as the firmware gets confused if so */
-	PMD_RX_LOG(DEBUG, "Increment FL write pointer in %u",
-		   rxq->rx_count - 1);
-
-	nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, rxq->rx_count - 1);
-
-	return 0;
-}
-
-static int
-nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-		       uint16_t nb_desc, unsigned int socket_id,
-		       const struct rte_eth_txconf *tx_conf)
-{
-	const struct rte_memzone *tz;
-	struct nfp_net_txq *txq;
-	uint16_t tx_free_thresh;
-	struct nfp_net_hw *hw;
-	uint32_t tx_desc_sz;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	PMD_INIT_FUNC_TRACE();
-
-	/* Validating number of descriptors */
-	tx_desc_sz = nb_desc * sizeof(struct nfp_net_tx_desc);
-	if (tx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
-	    nb_desc > NFP_NET_MAX_TX_DESC ||
-	    nb_desc < NFP_NET_MIN_TX_DESC) {
-		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
-		return -EINVAL;
-	}
-
-	tx_free_thresh = (uint16_t)((tx_conf->tx_free_thresh) ?
-				    tx_conf->tx_free_thresh :
-				    DEFAULT_TX_FREE_THRESH);
-
-	if (tx_free_thresh > (nb_desc)) {
-		PMD_DRV_LOG(ERR,
-			"tx_free_thresh must be less than the number of TX "
-			"descriptors. (tx_free_thresh=%u port=%d "
-			"queue=%d)", (unsigned int)tx_free_thresh,
-			dev->data->port_id, (int)queue_idx);
-		return -(EINVAL);
-	}
-
-	/*
-	 * Free memory prior to re-allocation if needed. This is the case after
-	 * calling nfp_net_stop
-	 */
-	if (dev->data->tx_queues[queue_idx]) {
-		PMD_TX_LOG(DEBUG, "Freeing memory prior to re-allocation %d",
-			   queue_idx);
-		nfp_net_tx_queue_release(dev->data->tx_queues[queue_idx]);
-		dev->data->tx_queues[queue_idx] = NULL;
-	}
-
-	/* Allocating tx queue data structure */
-	txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct nfp_net_txq),
-				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating tx dma");
-		return -ENOMEM;
-	}
-
-	/*
-	 * Allocate TX ring hardware descriptors. A memzone large enough to
-	 * handle the maximum ring size is allocated in order to allow for
-	 * resizing in later calls to the queue setup function.
-	 */
-	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
-				   sizeof(struct nfp_net_tx_desc) *
-				   NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN,
-				   socket_id);
-	if (tz == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating tx dma");
-		nfp_net_tx_queue_release(txq);
-		return -ENOMEM;
-	}
-
-	txq->tx_count = nb_desc;
-	txq->tx_free_thresh = tx_free_thresh;
-	txq->tx_pthresh = tx_conf->tx_thresh.pthresh;
-	txq->tx_hthresh = tx_conf->tx_thresh.hthresh;
-	txq->tx_wthresh = tx_conf->tx_thresh.wthresh;
-
-	/* queue mapping based on firmware configuration */
-	txq->qidx = queue_idx;
-	txq->tx_qcidx = queue_idx * hw->stride_tx;
-	txq->qcp_q = hw->tx_bar + NFP_QCP_QUEUE_OFF(txq->tx_qcidx);
-
-	txq->port_id = dev->data->port_id;
-
-	/* Saving physical and virtual addresses for the TX ring */
-	txq->dma = (uint64_t)tz->iova;
-	txq->txds = (struct nfp_net_tx_desc *)tz->addr;
-
-	/* mbuf pointers array for referencing mbufs linked to TX descriptors */
-	txq->txbufs = rte_zmalloc_socket("txq->txbufs",
-					 sizeof(*txq->txbufs) * nb_desc,
-					 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq->txbufs == NULL) {
-		nfp_net_tx_queue_release(txq);
-		return -ENOMEM;
-	}
-	PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
-		   txq->txbufs, txq->txds, (unsigned long int)txq->dma);
-
-	nfp_net_reset_tx_queue(txq);
-
-	dev->data->tx_queues[queue_idx] = txq;
-	txq->hw = hw;
-
-	/*
-	 * Telling the HW about the physical address of the TX ring and number
-	 * of descriptors in log2 format
-	 */
-	nn_cfg_writeq(hw, NFP_NET_CFG_TXR_ADDR(queue_idx), txq->dma);
-	nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), rte_log2_u32(nb_desc));
-
-	return 0;
-}
-
-/* nfp_net_tx_tso - Set TX descriptor for TSO */
-static inline void
-nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
-	       struct rte_mbuf *mb)
-{
-	uint64_t ol_flags;
-	struct nfp_net_hw *hw = txq->hw;
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY))
-		goto clean_txd;
-
-	ol_flags = mb->ol_flags;
-
-	if (!(ol_flags & PKT_TX_TCP_SEG))
-		goto clean_txd;
-
-	txd->l3_offset = mb->l2_len;
-	txd->l4_offset = mb->l2_len + mb->l3_len;
-	txd->lso_hdrlen = mb->l2_len + mb->l3_len + mb->l4_len;
-	txd->mss = rte_cpu_to_le_16(mb->tso_segsz);
-	txd->flags = PCIE_DESC_TX_LSO;
-	return;
-
-clean_txd:
-	txd->flags = 0;
-	txd->l3_offset = 0;
-	txd->l4_offset = 0;
-	txd->lso_hdrlen = 0;
-	txd->mss = 0;
-}
-
-/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
-static inline void
-nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
-		 struct rte_mbuf *mb)
-{
-	uint64_t ol_flags;
-	struct nfp_net_hw *hw = txq->hw;
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_TXCSUM))
-		return;
-
-	ol_flags = mb->ol_flags;
-
-	/* IPv6 does not need checksum */
-	if (ol_flags & PKT_TX_IP_CKSUM)
-		txd->flags |= PCIE_DESC_TX_IP4_CSUM;
-
-	switch (ol_flags & PKT_TX_L4_MASK) {
-	case PKT_TX_UDP_CKSUM:
-		txd->flags |= PCIE_DESC_TX_UDP_CSUM;
-		break;
-	case PKT_TX_TCP_CKSUM:
-		txd->flags |= PCIE_DESC_TX_TCP_CSUM;
-		break;
-	}
-
-	if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
-		txd->flags |= PCIE_DESC_TX_CSUM;
-}
-
-/* nfp_net_rx_cksum - set mbuf checksum flags based on RX descriptor flags */
-static inline void
-nfp_net_rx_cksum(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
-		 struct rte_mbuf *mb)
-{
-	struct nfp_net_hw *hw = rxq->hw;
-
-	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RXCSUM))
-		return;
-
-	/* If IPv4 and IP checksum error, fail */
-	if (unlikely((rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM) &&
-	    !(rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM_OK)))
-		mb->ol_flags |= PKT_RX_IP_CKSUM_BAD;
-	else
-		mb->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
-
-	/* If neither UDP nor TCP return */
-	if (!(rxd->rxd.flags & PCIE_DESC_RX_TCP_CSUM) &&
-	    !(rxd->rxd.flags & PCIE_DESC_RX_UDP_CSUM))
-		return;
-
-	if (likely(rxd->rxd.flags & PCIE_DESC_RX_L4_CSUM_OK))
-		mb->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
-	else
-		mb->ol_flags |= PKT_RX_L4_CKSUM_BAD;
-}
-
-#define NFP_HASH_OFFSET      ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 4)
-#define NFP_HASH_TYPE_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 8)
-
-#define NFP_DESC_META_LEN(d) (d->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
-
-/*
- * nfp_net_set_hash - Set mbuf hash data
- *
- * The RSS hash and hash-type are pre-pended to the packet data.
- * Extract and decode it and set the mbuf fields.
- */
-static inline void
-nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
-		 struct rte_mbuf *mbuf)
-{
-	struct nfp_net_hw *hw = rxq->hw;
-	uint8_t *meta_offset;
-	uint32_t meta_info;
-	uint32_t hash = 0;
-	uint32_t hash_type = 0;
-
-	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
-		return;
-
-	/* this is true for new firmwares */
-	if (likely(((hw->cap & NFP_NET_CFG_CTRL_RSS2) ||
-	    (NFD_CFG_MAJOR_VERSION_of(hw->ver) == 4)) &&
-	     NFP_DESC_META_LEN(rxd))) {
-		/*
-		 * new metadata api:
-		 * <----  32 bit  ----->
-		 * m    field type word
-		 * e     data field #2
-		 * t     data field #1
-		 * a     data field #0
-		 * ====================
-		 *    packet data
-		 *
-		 * Field type word contains up to 8 4bit field types
-		 * A 4bit field type refers to a data field word
-		 * A data field word can have several 4bit field types
-		 */
-		meta_offset = rte_pktmbuf_mtod(mbuf, uint8_t *);
-		meta_offset -= NFP_DESC_META_LEN(rxd);
-		meta_info = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
-		meta_offset += 4;
-		/* NFP PMD just supports metadata for hashing */
-		switch (meta_info & NFP_NET_META_FIELD_MASK) {
-		case NFP_NET_META_HASH:
-			/* next field type is about the hash type */
-			meta_info >>= NFP_NET_META_FIELD_SIZE;
-			/* hash value is in the data field */
-			hash = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
-			hash_type = meta_info & NFP_NET_META_FIELD_MASK;
-			break;
-		default:
-			/* Unsupported metadata can be a performance issue */
-			return;
-		}
-	} else {
-		if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
-			return;
-
-		hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
-		hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
-	}
-
-	mbuf->hash.rss = hash;
-	mbuf->ol_flags |= PKT_RX_RSS_HASH;
-
-	switch (hash_type) {
-	case NFP_NET_RSS_IPV4:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV4;
-		break;
-	case NFP_NET_RSS_IPV6:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6;
-		break;
-	case NFP_NET_RSS_IPV6_EX:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV4_TCP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV6_TCP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV4_UDP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV6_UDP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	default:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L4_MASK;
-	}
-}
-
-static inline void
-nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq)
-{
-	rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
-}
-
-#define NFP_DESC_META_LEN(d) (d->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
-
-/*
- * RX path design:
- *
- * There are some decisions to take:
- * 1) How to check DD RX descriptors bit
- * 2) How and when to allocate new mbufs
- *
- * Current implementation checks just one single DD bit each loop. As each
- * descriptor is 8 bytes, it is likely a good idea to check descriptors in
- * a single cache line instead. Tests with this change have not shown any
- * performance improvement but it requires further investigation. For example,
- * depending on which descriptor is next, the number of descriptors could be
- * less than 8 for just checking those in the same cache line. This implies
- * extra work which could be counterproductive by itself. Indeed, last firmware
- * changes are just doing this: writing several descriptors with the DD bit
- * for saving PCIe bandwidth and DMA operations from the NFP.
- *
- * Mbuf allocation is done when a new packet is received. Then the descriptor
- * is automatically linked with the new mbuf and the old one is given to the
- * user. The main drawback with this design is mbuf allocation is heavier than
- * using bulk allocations allowed by DPDK with rte_mempool_get_bulk. From the
- * cache point of view it does not seem allocating the mbuf early on as we are
- * doing now have any benefit at all. Again, tests with this change have not
- * shown any improvement. Also, rte_mempool_get_bulk returns all or nothing
- * so looking at the implications of this type of allocation should be studied
- * deeply
- */
-
-static uint16_t
-nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
-{
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_rx_desc *rxds;
-	struct nfp_net_rx_buff *rxb;
-	struct nfp_net_hw *hw;
-	struct rte_mbuf *mb;
-	struct rte_mbuf *new_mb;
-	uint16_t nb_hold;
-	uint64_t dma_addr;
-	int avail;
-
-	rxq = rx_queue;
-	if (unlikely(rxq == NULL)) {
-		/*
-		 * DPDK just checks the queue is lower than max queues
-		 * enabled. But the queue needs to be configured
-		 */
-		RTE_LOG_DP(ERR, PMD, "RX Bad queue\n");
-		return -EINVAL;
-	}
-
-	hw = rxq->hw;
-	avail = 0;
-	nb_hold = 0;
-
-	while (avail < nb_pkts) {
-		rxb = &rxq->rxbufs[rxq->rd_p];
-		if (unlikely(rxb == NULL)) {
-			RTE_LOG_DP(ERR, PMD, "rxb does not exist!\n");
-			break;
-		}
-
-		rxds = &rxq->rxds[rxq->rd_p];
-		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
-			break;
-
-		/*
-		 * Memory barrier to ensure that we won't do other
-		 * reads before the DD bit.
-		 */
-		rte_rmb();
-
-		/*
-		 * We got a packet. Let's alloc a new mbuf for refilling the
-		 * free descriptor ring as soon as possible
-		 */
-		new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
-		if (unlikely(new_mb == NULL)) {
-			RTE_LOG_DP(DEBUG, PMD,
-			"RX mbuf alloc failed port_id=%u queue_id=%u\n",
-				rxq->port_id, (unsigned int)rxq->qidx);
-			nfp_net_mbuf_alloc_failed(rxq);
-			break;
-		}
-
-		nb_hold++;
-
-		/*
-		 * Grab the mbuf and refill the descriptor with the
-		 * previously allocated mbuf
-		 */
-		mb = rxb->mbuf;
-		rxb->mbuf = new_mb;
-
-		PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u",
-			   rxds->rxd.data_len, rxq->mbuf_size);
-
-		/* Size of this segment */
-		mb->data_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
-		/* Size of the whole packet. We just support 1 segment */
-		mb->pkt_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
-
-		if (unlikely((mb->data_len + hw->rx_offset) >
-			     rxq->mbuf_size)) {
-			/*
-			 * This should not happen and the user has the
-			 * responsibility of avoiding it. But we have
-			 * to give some info about the error
-			 */
-			RTE_LOG_DP(ERR, PMD,
-				"mbuf overflow likely due to the RX offset.\n"
-				"\t\tYour mbuf size should have extra space for"
-				" RX offset=%u bytes.\n"
-				"\t\tCurrently you just have %u bytes available"
-				" but the received packet is %u bytes long",
-				hw->rx_offset,
-				rxq->mbuf_size - hw->rx_offset,
-				mb->data_len);
-			return -EINVAL;
-		}
-
-		/* Filling the received mbuf with packet info */
-		if (hw->rx_offset)
-			mb->data_off = RTE_PKTMBUF_HEADROOM + hw->rx_offset;
-		else
-			mb->data_off = RTE_PKTMBUF_HEADROOM +
-				       NFP_DESC_META_LEN(rxds);
-
-		/* No scatter mode supported */
-		mb->nb_segs = 1;
-		mb->next = NULL;
-
-		mb->port = rxq->port_id;
-
-		/* Checking the RSS flag */
-		nfp_net_set_hash(rxq, rxds, mb);
-
-		/* Checking the checksum flag */
-		nfp_net_rx_cksum(rxq, rxds, mb);
-
-		if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
-		    (hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
-			mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
-			mb->ol_flags |= PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
-		}
-
-		/* Adding the mbuf to the mbuf array passed by the app */
-		rx_pkts[avail++] = mb;
-
-		/* Now resetting and updating the descriptor */
-		rxds->vals[0] = 0;
-		rxds->vals[1] = 0;
-		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(new_mb));
-		rxds->fld.dd = 0;
-		rxds->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
-		rxds->fld.dma_addr_lo = dma_addr & 0xffffffff;
-
-		rxq->rd_p++;
-		if (unlikely(rxq->rd_p == rxq->rx_count)) /* wrapping?*/
-			rxq->rd_p = 0;
-	}
-
-	if (nb_hold == 0)
-		return nb_hold;
-
-	PMD_RX_LOG(DEBUG, "RX  port_id=%u queue_id=%u, %d packets received",
-		   rxq->port_id, (unsigned int)rxq->qidx, nb_hold);
-
-	nb_hold += rxq->nb_rx_hold;
-
-	/*
-	 * FL descriptors needs to be written before incrementing the
-	 * FL queue WR pointer
-	 */
-	rte_wmb();
-	if (nb_hold > rxq->rx_free_thresh) {
-		PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u",
-			   rxq->port_id, (unsigned int)rxq->qidx,
-			   (unsigned)nb_hold, (unsigned)avail);
-		nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, nb_hold);
-		nb_hold = 0;
-	}
-	rxq->nb_rx_hold = nb_hold;
-
-	return avail;
-}
-
-/*
- * nfp_net_tx_free_bufs - Check for descriptors with a complete
- * status
- * @txq: TX queue to work with
- * Returns number of descriptors freed
- */
-int
-nfp_net_tx_free_bufs(struct nfp_net_txq *txq)
-{
-	uint32_t qcp_rd_p;
-	int todo;
-
-	PMD_TX_LOG(DEBUG, "queue %u. Check for descriptor with a complete"
-		   " status", txq->qidx);
-
-	/* Work out how many packets have been sent */
-	qcp_rd_p = nfp_qcp_read(txq->qcp_q, NFP_QCP_READ_PTR);
-
-	if (qcp_rd_p == txq->rd_p) {
-		PMD_TX_LOG(DEBUG, "queue %u: It seems harrier is not sending "
-			   "packets (%u, %u)", txq->qidx,
-			   qcp_rd_p, txq->rd_p);
-		return 0;
-	}
-
-	if (qcp_rd_p > txq->rd_p)
-		todo = qcp_rd_p - txq->rd_p;
-	else
-		todo = qcp_rd_p + txq->tx_count - txq->rd_p;
-
-	PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->rd_p: %u, qcp->rd_p: %u",
-		   qcp_rd_p, txq->rd_p, txq->rd_p);
-
-	if (todo == 0)
-		return todo;
-
-	txq->rd_p += todo;
-	if (unlikely(txq->rd_p >= txq->tx_count))
-		txq->rd_p -= txq->tx_count;
-
-	return todo;
-}
-
-/* Leaving always free descriptors for avoiding wrapping confusion */
-static inline
-uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq)
-{
-	if (txq->wr_p >= txq->rd_p)
-		return txq->tx_count - (txq->wr_p - txq->rd_p) - 8;
-	else
-		return txq->rd_p - txq->wr_p - 8;
-}
-
-/*
- * nfp_net_txq_full - Check if the TX queue free descriptors
- * is below tx_free_threshold
- *
- * @txq: TX queue to check
- *
- * This function uses the host copy* of read/write pointers
- */
-static inline
-uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
-{
-	return (nfp_free_tx_desc(txq) < txq->tx_free_thresh);
-}
-
-static uint16_t
-nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
-{
-	struct nfp_net_txq *txq;
-	struct nfp_net_hw *hw;
-	struct nfp_net_tx_desc *txds, txd;
-	struct rte_mbuf *pkt;
-	uint64_t dma_addr;
-	int pkt_size, dma_size;
-	uint16_t free_descs, issued_descs;
-	struct rte_mbuf **lmbuf;
-	int i;
-
-	txq = tx_queue;
-	hw = txq->hw;
-	txds = &txq->txds[txq->wr_p];
-
-	PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets",
-		   txq->qidx, txq->wr_p, nb_pkts);
-
-	if ((nfp_free_tx_desc(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
-		nfp_net_tx_free_bufs(txq);
-
-	free_descs = (uint16_t)nfp_free_tx_desc(txq);
-	if (unlikely(free_descs == 0))
-		return 0;
-
-	pkt = *tx_pkts;
-
-	i = 0;
-	issued_descs = 0;
-	PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets",
-		   txq->qidx, nb_pkts);
-	/* Sending packets */
-	while ((i < nb_pkts) && free_descs) {
-		/* Grabbing the mbuf linked to the current descriptor */
-		lmbuf = &txq->txbufs[txq->wr_p].mbuf;
-		/* Warming the cache for releasing the mbuf later on */
-		RTE_MBUF_PREFETCH_TO_FREE(*lmbuf);
-
-		pkt = *(tx_pkts + i);
-
-		if (unlikely((pkt->nb_segs > 1) &&
-			     !(hw->cap & NFP_NET_CFG_CTRL_GATHER))) {
-			PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set");
-			rte_panic("Multisegment packet unsupported\n");
-		}
-
-		/* Checking if we have enough descriptors */
-		if (unlikely(pkt->nb_segs > free_descs))
-			goto xmit_end;
-
-		/*
-		 * Checksum and VLAN flags just in the first descriptor for a
-		 * multisegment packet, but TSO info needs to be in all of them.
-		 */
-		txd.data_len = pkt->pkt_len;
-		nfp_net_tx_tso(txq, &txd, pkt);
-		nfp_net_tx_cksum(txq, &txd, pkt);
-
-		if ((pkt->ol_flags & PKT_TX_VLAN_PKT) &&
-		    (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)) {
-			txd.flags |= PCIE_DESC_TX_VLAN;
-			txd.vlan = pkt->vlan_tci;
-		}
-
-		/*
-		 * mbuf data_len is the data in one segment and pkt_len data
-		 * in the whole packet. When the packet is just one segment,
-		 * then data_len = pkt_len
-		 */
-		pkt_size = pkt->pkt_len;
-
-		while (pkt) {
-			/* Copying TSO, VLAN and cksum info */
-			*txds = txd;
-
-			/* Releasing mbuf used by this descriptor previously*/
-			if (*lmbuf)
-				rte_pktmbuf_free_seg(*lmbuf);
-
-			/*
-			 * Linking mbuf with descriptor for being released
-			 * next time descriptor is used
-			 */
-			*lmbuf = pkt;
-
-			dma_size = pkt->data_len;
-			dma_addr = rte_mbuf_data_iova(pkt);
-			PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
-				   "%" PRIx64 "", dma_addr);
-
-			/* Filling descriptors fields */
-			txds->dma_len = dma_size;
-			txds->data_len = txd.data_len;
-			txds->dma_addr_hi = (dma_addr >> 32) & 0xff;
-			txds->dma_addr_lo = (dma_addr & 0xffffffff);
-			ASSERT(free_descs > 0);
-			free_descs--;
-
-			txq->wr_p++;
-			if (unlikely(txq->wr_p == txq->tx_count)) /* wrapping?*/
-				txq->wr_p = 0;
-
-			pkt_size -= dma_size;
-
-			/*
-			 * Making the EOP, packets with just one segment
-			 * the priority
-			 */
-			if (likely(!pkt_size))
-				txds->offset_eop = PCIE_DESC_TX_EOP;
-			else
-				txds->offset_eop = 0;
-
-			pkt = pkt->next;
-			/* Referencing next free TX descriptor */
-			txds = &txq->txds[txq->wr_p];
-			lmbuf = &txq->txbufs[txq->wr_p].mbuf;
-			issued_descs++;
-		}
-		i++;
-	}
-
-xmit_end:
-	/* Increment write pointers. Force memory write before we let HW know */
-	rte_wmb();
-	nfp_qcp_ptr_add(txq->qcp_q, NFP_QCP_WRITE_PTR, issued_descs);
-
-	return i;
-}
-
 static int
 nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 {
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index a3a3ba32d6..9265496bf0 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -41,6 +41,12 @@ struct nfp_net_adapter;
 #define NFP_QCP_QUEUE_STS_HI                    0x000c
 #define NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask    (0x3ffff)
 
+/* The offset of the queue controller queues in the PCIe Target */
+#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
+
+/* Maximum value which can be added to a queue with one transaction */
+#define NFP_QCP_MAX_ADD	0x7f
+
 /* Interrupt definitions */
 #define NFP_NET_IRQ_LSC_IDX             0
 
@@ -95,47 +101,11 @@ struct nfp_net_adapter;
 #include <linux/types.h>
 #include <rte_io.h>
 
-static inline uint8_t nn_readb(volatile const void *addr)
-{
-	return rte_read8(addr);
-}
-
-static inline void nn_writeb(uint8_t val, volatile void *addr)
-{
-	rte_write8(val, addr);
-}
-
-static inline uint32_t nn_readl(volatile const void *addr)
-{
-	return rte_read32(addr);
-}
-
-static inline void nn_writel(uint32_t val, volatile void *addr)
-{
-	rte_write32(val, addr);
-}
-
-static inline void nn_writew(uint16_t val, volatile void *addr)
-{
-	rte_write16(val, addr);
-}
-
-static inline uint64_t nn_readq(volatile void *addr)
-{
-	const volatile uint32_t *p = addr;
-	uint32_t low, high;
-
-	high = nn_readl((volatile const void *)(p + 1));
-	low = nn_readl((volatile const void *)p);
-
-	return low + ((uint64_t)high << 32);
-}
-
-static inline void nn_writeq(uint64_t val, volatile void *addr)
-{
-	nn_writel(val >> 32, (volatile char *)addr + 4);
-	nn_writel(val, addr);
-}
+/* nfp_qcp_ptr - Read or Write Pointer of a queue */
+enum nfp_qcp_ptr {
+	NFP_QCP_READ_PTR = 0,
+	NFP_QCP_WRITE_PTR
+};
 
 struct nfp_pf_dev {
 	/* Backpointer to associated pci device */
@@ -247,6 +217,138 @@ struct nfp_net_adapter {
 	struct nfp_net_hw hw;
 };
 
+static inline uint8_t nn_readb(volatile const void *addr)
+{
+	return rte_read8(addr);
+}
+
+static inline void nn_writeb(uint8_t val, volatile void *addr)
+{
+	rte_write8(val, addr);
+}
+
+static inline uint32_t nn_readl(volatile const void *addr)
+{
+	return rte_read32(addr);
+}
+
+static inline void nn_writel(uint32_t val, volatile void *addr)
+{
+	rte_write32(val, addr);
+}
+
+static inline void nn_writew(uint16_t val, volatile void *addr)
+{
+	rte_write16(val, addr);
+}
+
+static inline uint64_t nn_readq(volatile void *addr)
+{
+	const volatile uint32_t *p = addr;
+	uint32_t low, high;
+
+	high = nn_readl((volatile const void *)(p + 1));
+	low = nn_readl((volatile const void *)p);
+
+	return low + ((uint64_t)high << 32);
+}
+
+static inline void nn_writeq(uint64_t val, volatile void *addr)
+{
+	nn_writel(val >> 32, (volatile char *)addr + 4);
+	nn_writel(val, addr);
+}
+
+/*
+ * Functions to read/write from/to Config BAR
+ * Performs any endian conversion necessary.
+ */
+static inline uint8_t
+nn_cfg_readb(struct nfp_net_hw *hw, int off)
+{
+	return nn_readb(hw->ctrl_bar + off);
+}
+
+static inline void
+nn_cfg_writeb(struct nfp_net_hw *hw, int off, uint8_t val)
+{
+	nn_writeb(val, hw->ctrl_bar + off);
+}
+
+static inline uint32_t
+nn_cfg_readl(struct nfp_net_hw *hw, int off)
+{
+	return rte_le_to_cpu_32(nn_readl(hw->ctrl_bar + off));
+}
+
+static inline void
+nn_cfg_writel(struct nfp_net_hw *hw, int off, uint32_t val)
+{
+	nn_writel(rte_cpu_to_le_32(val), hw->ctrl_bar + off);
+}
+
+static inline uint64_t
+nn_cfg_readq(struct nfp_net_hw *hw, int off)
+{
+	return rte_le_to_cpu_64(nn_readq(hw->ctrl_bar + off));
+}
+
+static inline void
+nn_cfg_writeq(struct nfp_net_hw *hw, int off, uint64_t val)
+{
+	nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
+}
+
+/*
+ * nfp_qcp_ptr_add - Add the value to the selected pointer of a queue
+ * @q: Base address for queue structure
+ * @ptr: Add to the Read or Write pointer
+ * @val: Value to add to the queue pointer
+ *
+ * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed.
+ */
+static inline void
+nfp_qcp_ptr_add(uint8_t *q, enum nfp_qcp_ptr ptr, uint32_t val)
+{
+	uint32_t off;
+
+	if (ptr == NFP_QCP_READ_PTR)
+		off = NFP_QCP_QUEUE_ADD_RPTR;
+	else
+		off = NFP_QCP_QUEUE_ADD_WPTR;
+
+	while (val > NFP_QCP_MAX_ADD) {
+		nn_writel(rte_cpu_to_le_32(NFP_QCP_MAX_ADD), q + off);
+		val -= NFP_QCP_MAX_ADD;
+}
+
+nn_writel(rte_cpu_to_le_32(val), q + off);
+}
+
+/*
+ * nfp_qcp_read - Read the current Read/Write pointer value for a queue
+ * @q:  Base address for queue structure
+ * @ptr: Read or Write pointer
+ */
+static inline uint32_t
+nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
+{
+	uint32_t off;
+	uint32_t val;
+
+	if (ptr == NFP_QCP_READ_PTR)
+		off = NFP_QCP_QUEUE_STS_LO;
+	else
+		off = NFP_QCP_QUEUE_STS_HI;
+
+	val = rte_cpu_to_le_32(nn_readl(q + off));
+
+	if (ptr == NFP_QCP_READ_PTR)
+		return val & NFP_QCP_QUEUE_STS_LO_READPTR_mask;
+	else
+		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
+}
+
 #define NFP_NET_DEV_PRIVATE_TO_HW(adapter)\
 	(&((struct nfp_net_adapter *)adapter)->hw)
 
diff --git a/drivers/net/nfp/nfp_rxtx.c b/drivers/net/nfp/nfp_rxtx.c
new file mode 100644
index 0000000000..9ee9e5c9a3
--- /dev/null
+++ b/drivers/net/nfp/nfp_rxtx.c
@@ -0,0 +1,1002 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_rxtx.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: Rx/Tx functions
+ */
+
+#include <ethdev_driver.h>
+#include <ethdev_pci.h>
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+
+/* Prototypes */
+static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
+static inline void nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq);
+static inline void nfp_net_set_hash(struct nfp_net_rxq *rxq,
+				    struct nfp_net_rx_desc *rxd,
+				    struct rte_mbuf *mbuf);
+static inline void nfp_net_rx_cksum(struct nfp_net_rxq *rxq,
+				    struct nfp_net_rx_desc *rxd,
+				    struct rte_mbuf *mb);
+static void nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq);
+static int nfp_net_tx_free_bufs(struct nfp_net_txq *txq);
+static void nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq);
+static inline uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq);
+static inline uint32_t nfp_net_txq_full(struct nfp_net_txq *txq);
+static inline void nfp_net_tx_tso(struct nfp_net_txq *txq,
+				  struct nfp_net_tx_desc *txd,
+				  struct rte_mbuf *mb);
+static inline void nfp_net_tx_cksum(struct nfp_net_txq *txq,
+				    struct nfp_net_tx_desc *txd,
+				    struct rte_mbuf *mb);
+
+static int
+nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq)
+{
+	struct nfp_net_rx_buff *rxe = rxq->rxbufs;
+	uint64_t dma_addr;
+	unsigned int i;
+
+	PMD_RX_LOG(DEBUG, "Fill Rx Freelist for %u descriptors",
+		   rxq->rx_count);
+
+	for (i = 0; i < rxq->rx_count; i++) {
+		struct nfp_net_rx_desc *rxd;
+		struct rte_mbuf *mbuf = rte_pktmbuf_alloc(rxq->mem_pool);
+
+		if (mbuf == NULL) {
+			PMD_DRV_LOG(ERR, "RX mbuf alloc failed queue_id=%u",
+				(unsigned int)rxq->qidx);
+			return -ENOMEM;
+		}
+
+		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(mbuf));
+
+		rxd = &rxq->rxds[i];
+		rxd->fld.dd = 0;
+		rxd->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
+		rxd->fld.dma_addr_lo = dma_addr & 0xffffffff;
+		rxe[i].mbuf = mbuf;
+		PMD_RX_LOG(DEBUG, "[%d]: %" PRIx64, i, dma_addr);
+	}
+
+	/* Make sure all writes are flushed before telling the hardware */
+	rte_wmb();
+
+	/* Not advertising the whole ring as the firmware gets confused if so */
+	PMD_RX_LOG(DEBUG, "Increment FL write pointer in %u",
+		   rxq->rx_count - 1);
+
+	nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, rxq->rx_count - 1);
+
+	return 0;
+}
+
+int
+nfp_net_rx_freelist_setup(struct rte_eth_dev *dev)
+{
+	int i;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		if (nfp_net_rx_fill_freelist(dev->data->rx_queues[i]) < 0)
+			return -1;
+	}
+	return 0;
+}
+
+uint32_t
+nfp_net_rx_queue_count(struct rte_eth_dev *dev, uint16_t queue_idx)
+{
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_rx_desc *rxds;
+	uint32_t idx;
+	uint32_t count;
+
+	rxq = (struct nfp_net_rxq *)dev->data->rx_queues[queue_idx];
+
+	idx = rxq->rd_p;
+
+	count = 0;
+
+	/*
+	 * Other PMDs are just checking the DD bit in intervals of 4
+	 * descriptors and counting all four if the first has the DD
+	 * bit on. Of course, this is not accurate but can be good for
+	 * performance. But ideally that should be done in descriptors
+	 * chunks belonging to the same cache line
+	 */
+
+	while (count < rxq->rx_count) {
+		rxds = &rxq->rxds[idx];
+		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
+			break;
+
+		count++;
+		idx++;
+
+		/* Wrapping? */
+		if ((idx) == rxq->rx_count)
+			idx = 0;
+	}
+
+	return count;
+}
+
+static inline void
+nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq)
+{
+	rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
+}
+
+/*
+ * nfp_net_set_hash - Set mbuf hash data
+ *
+ * The RSS hash and hash-type are pre-pended to the packet data.
+ * Extract and decode it and set the mbuf fields.
+ */
+static inline void
+nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
+		 struct rte_mbuf *mbuf)
+{
+	struct nfp_net_hw *hw = rxq->hw;
+	uint8_t *meta_offset;
+	uint32_t meta_info;
+	uint32_t hash = 0;
+	uint32_t hash_type = 0;
+
+	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
+		return;
+
+	/* this is true for new firmwares */
+	if (likely(((hw->cap & NFP_NET_CFG_CTRL_RSS2) ||
+	    (NFD_CFG_MAJOR_VERSION_of(hw->ver) == 4)) &&
+	     NFP_DESC_META_LEN(rxd))) {
+		/*
+		 * new metadata api:
+		 * <----  32 bit  ----->
+		 * m    field type word
+		 * e     data field #2
+		 * t     data field #1
+		 * a     data field #0
+		 * ====================
+		 *    packet data
+		 *
+		 * Field type word contains up to 8 4bit field types
+		 * A 4bit field type refers to a data field word
+		 * A data field word can have several 4bit field types
+		 */
+		meta_offset = rte_pktmbuf_mtod(mbuf, uint8_t *);
+		meta_offset -= NFP_DESC_META_LEN(rxd);
+		meta_info = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+		meta_offset += 4;
+		/* NFP PMD just supports metadata for hashing */
+		switch (meta_info & NFP_NET_META_FIELD_MASK) {
+		case NFP_NET_META_HASH:
+			/* next field type is about the hash type */
+			meta_info >>= NFP_NET_META_FIELD_SIZE;
+			/* hash value is in the data field */
+			hash = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+			hash_type = meta_info & NFP_NET_META_FIELD_MASK;
+			break;
+		default:
+			/* Unsupported metadata can be a performance issue */
+			return;
+		}
+	} else {
+		if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
+			return;
+
+		hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
+		hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
+	}
+
+	mbuf->hash.rss = hash;
+	mbuf->ol_flags |= PKT_RX_RSS_HASH;
+
+	switch (hash_type) {
+	case NFP_NET_RSS_IPV4:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV4;
+		break;
+	case NFP_NET_RSS_IPV6:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6;
+		break;
+	case NFP_NET_RSS_IPV6_EX:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV4_TCP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV6_TCP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV4_UDP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV6_UDP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	default:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L4_MASK;
+	}
+}
+
+/* nfp_net_rx_cksum - set mbuf checksum flags based on RX descriptor flags */
+static inline void
+nfp_net_rx_cksum(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
+		 struct rte_mbuf *mb)
+{
+	struct nfp_net_hw *hw = rxq->hw;
+
+	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RXCSUM))
+		return;
+
+	/* If IPv4 and IP checksum error, fail */
+	if (unlikely((rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM) &&
+	    !(rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM_OK)))
+		mb->ol_flags |= PKT_RX_IP_CKSUM_BAD;
+	else
+		mb->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+
+	/* If neither UDP nor TCP return */
+	if (!(rxd->rxd.flags & PCIE_DESC_RX_TCP_CSUM) &&
+	    !(rxd->rxd.flags & PCIE_DESC_RX_UDP_CSUM))
+		return;
+
+	if (likely(rxd->rxd.flags & PCIE_DESC_RX_L4_CSUM_OK))
+		mb->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
+	else
+		mb->ol_flags |= PKT_RX_L4_CKSUM_BAD;
+}
+
+/*
+ * RX path design:
+ *
+ * There are some decisions to take:
+ * 1) How to check DD RX descriptors bit
+ * 2) How and when to allocate new mbufs
+ *
+ * Current implementation checks just one single DD bit each loop. As each
+ * descriptor is 8 bytes, it is likely a good idea to check descriptors in
+ * a single cache line instead. Tests with this change have not shown any
+ * performance improvement but it requires further investigation. For example,
+ * depending on which descriptor is next, the number of descriptors could be
+ * less than 8 for just checking those in the same cache line. This implies
+ * extra work which could be counterproductive by itself. Indeed, last firmware
+ * changes are just doing this: writing several descriptors with the DD bit
+ * for saving PCIe bandwidth and DMA operations from the NFP.
+ *
+ * Mbuf allocation is done when a new packet is received. Then the descriptor
+ * is automatically linked with the new mbuf and the old one is given to the
+ * user. The main drawback with this design is mbuf allocation is heavier than
+ * using bulk allocations allowed by DPDK with rte_mempool_get_bulk. From the
+ * cache point of view it does not seem allocating the mbuf early on as we are
+ * doing now have any benefit at all. Again, tests with this change have not
+ * shown any improvement. Also, rte_mempool_get_bulk returns all or nothing
+ * so looking at the implications of this type of allocation should be studied
+ * deeply
+ */
+
+uint16_t
+nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_rx_desc *rxds;
+	struct nfp_net_rx_buff *rxb;
+	struct nfp_net_hw *hw;
+	struct rte_mbuf *mb;
+	struct rte_mbuf *new_mb;
+	uint16_t nb_hold;
+	uint64_t dma_addr;
+	int avail;
+
+	rxq = rx_queue;
+	if (unlikely(rxq == NULL)) {
+		/*
+		 * DPDK just checks the queue is lower than max queues
+		 * enabled. But the queue needs to be configured
+		 */
+		RTE_LOG_DP(ERR, PMD, "RX Bad queue\n");
+		return -EINVAL;
+	}
+
+	hw = rxq->hw;
+	avail = 0;
+	nb_hold = 0;
+
+	while (avail < nb_pkts) {
+		rxb = &rxq->rxbufs[rxq->rd_p];
+		if (unlikely(rxb == NULL)) {
+			RTE_LOG_DP(ERR, PMD, "rxb does not exist!\n");
+			break;
+		}
+
+		rxds = &rxq->rxds[rxq->rd_p];
+		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
+			break;
+
+		/*
+		 * Memory barrier to ensure that we won't do other
+		 * reads before the DD bit.
+		 */
+		rte_rmb();
+
+		/*
+		 * We got a packet. Let's alloc a new mbuf for refilling the
+		 * free descriptor ring as soon as possible
+		 */
+		new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
+		if (unlikely(new_mb == NULL)) {
+			RTE_LOG_DP(DEBUG, PMD,
+			"RX mbuf alloc failed port_id=%u queue_id=%u\n",
+				rxq->port_id, (unsigned int)rxq->qidx);
+			nfp_net_mbuf_alloc_failed(rxq);
+			break;
+		}
+
+		nb_hold++;
+
+		/*
+		 * Grab the mbuf and refill the descriptor with the
+		 * previously allocated mbuf
+		 */
+		mb = rxb->mbuf;
+		rxb->mbuf = new_mb;
+
+		PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u",
+			   rxds->rxd.data_len, rxq->mbuf_size);
+
+		/* Size of this segment */
+		mb->data_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
+		/* Size of the whole packet. We just support 1 segment */
+		mb->pkt_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
+
+		if (unlikely((mb->data_len + hw->rx_offset) >
+			     rxq->mbuf_size)) {
+			/*
+			 * This should not happen and the user has the
+			 * responsibility of avoiding it. But we have
+			 * to give some info about the error
+			 */
+			RTE_LOG_DP(ERR, PMD,
+				"mbuf overflow likely due to the RX offset.\n"
+				"\t\tYour mbuf size should have extra space for"
+				" RX offset=%u bytes.\n"
+				"\t\tCurrently you just have %u bytes available"
+				" but the received packet is %u bytes long",
+				hw->rx_offset,
+				rxq->mbuf_size - hw->rx_offset,
+				mb->data_len);
+			return -EINVAL;
+		}
+
+		/* Filling the received mbuf with packet info */
+		if (hw->rx_offset)
+			mb->data_off = RTE_PKTMBUF_HEADROOM + hw->rx_offset;
+		else
+			mb->data_off = RTE_PKTMBUF_HEADROOM +
+				       NFP_DESC_META_LEN(rxds);
+
+		/* No scatter mode supported */
+		mb->nb_segs = 1;
+		mb->next = NULL;
+
+		mb->port = rxq->port_id;
+
+		/* Checking the RSS flag */
+		nfp_net_set_hash(rxq, rxds, mb);
+
+		/* Checking the checksum flag */
+		nfp_net_rx_cksum(rxq, rxds, mb);
+
+		if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
+		    (hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
+			mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
+			mb->ol_flags |= PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
+		}
+
+		/* Adding the mbuf to the mbuf array passed by the app */
+		rx_pkts[avail++] = mb;
+
+		/* Now resetting and updating the descriptor */
+		rxds->vals[0] = 0;
+		rxds->vals[1] = 0;
+		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(new_mb));
+		rxds->fld.dd = 0;
+		rxds->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
+		rxds->fld.dma_addr_lo = dma_addr & 0xffffffff;
+
+		rxq->rd_p++;
+		if (unlikely(rxq->rd_p == rxq->rx_count)) /* wrapping?*/
+			rxq->rd_p = 0;
+	}
+
+	if (nb_hold == 0)
+		return nb_hold;
+
+	PMD_RX_LOG(DEBUG, "RX  port_id=%u queue_id=%u, %d packets received",
+		   rxq->port_id, (unsigned int)rxq->qidx, nb_hold);
+
+	nb_hold += rxq->nb_rx_hold;
+
+	/*
+	 * FL descriptors needs to be written before incrementing the
+	 * FL queue WR pointer
+	 */
+	rte_wmb();
+	if (nb_hold > rxq->rx_free_thresh) {
+		PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u",
+			   rxq->port_id, (unsigned int)rxq->qidx,
+			   (unsigned int)nb_hold, (unsigned int)avail);
+		nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, nb_hold);
+		nb_hold = 0;
+	}
+	rxq->nb_rx_hold = nb_hold;
+
+	return avail;
+}
+
+static void
+nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
+{
+	unsigned int i;
+
+	if (rxq->rxbufs == NULL)
+		return;
+
+	for (i = 0; i < rxq->rx_count; i++) {
+		if (rxq->rxbufs[i].mbuf) {
+			rte_pktmbuf_free_seg(rxq->rxbufs[i].mbuf);
+			rxq->rxbufs[i].mbuf = NULL;
+		}
+	}
+}
+
+void
+nfp_net_rx_queue_release(void *rx_queue)
+{
+	struct nfp_net_rxq *rxq = rx_queue;
+
+	if (rxq) {
+		nfp_net_rx_queue_release_mbufs(rxq);
+		rte_free(rxq->rxbufs);
+		rte_free(rxq);
+	}
+}
+
+void
+nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
+{
+	nfp_net_rx_queue_release_mbufs(rxq);
+	rxq->rd_p = 0;
+	rxq->nb_rx_hold = 0;
+}
+
+int
+nfp_net_rx_queue_setup(struct rte_eth_dev *dev,
+		       uint16_t queue_idx, uint16_t nb_desc,
+		       unsigned int socket_id,
+		       const struct rte_eth_rxconf *rx_conf,
+		       struct rte_mempool *mp)
+{
+	const struct rte_memzone *tz;
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_hw *hw;
+	uint32_t rx_desc_sz;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* Validating number of descriptors */
+	rx_desc_sz = nb_desc * sizeof(struct nfp_net_rx_desc);
+	if (rx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
+	    nb_desc > NFP_NET_MAX_RX_DESC ||
+	    nb_desc < NFP_NET_MIN_RX_DESC) {
+		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
+		return -EINVAL;
+	}
+
+	/*
+	 * Free memory prior to re-allocation if needed. This is the case after
+	 * calling nfp_net_stop
+	 */
+	if (dev->data->rx_queues[queue_idx]) {
+		nfp_net_rx_queue_release(dev->data->rx_queues[queue_idx]);
+		dev->data->rx_queues[queue_idx] = NULL;
+	}
+
+	/* Allocating rx queue data structure */
+	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct nfp_net_rxq),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxq == NULL)
+		return -ENOMEM;
+
+	/* Hw queues mapping based on firmware configuration */
+	rxq->qidx = queue_idx;
+	rxq->fl_qcidx = queue_idx * hw->stride_rx;
+	rxq->rx_qcidx = rxq->fl_qcidx + (hw->stride_rx - 1);
+	rxq->qcp_fl = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->fl_qcidx);
+	rxq->qcp_rx = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->rx_qcidx);
+
+	/*
+	 * Tracking mbuf size for detecting a potential mbuf overflow due to
+	 * RX offset
+	 */
+	rxq->mem_pool = mp;
+	rxq->mbuf_size = rxq->mem_pool->elt_size;
+	rxq->mbuf_size -= (sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM);
+	hw->flbufsz = rxq->mbuf_size;
+
+	rxq->rx_count = nb_desc;
+	rxq->port_id = dev->data->port_id;
+	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
+	rxq->drop_en = rx_conf->rx_drop_en;
+
+	/*
+	 * Allocate RX ring hardware descriptors. A memzone large enough to
+	 * handle the maximum ring size is allocated in order to allow for
+	 * resizing in later calls to the queue setup function.
+	 */
+	tz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
+				   sizeof(struct nfp_net_rx_desc) *
+				   NFP_NET_MAX_RX_DESC, NFP_MEMZONE_ALIGN,
+				   socket_id);
+
+	if (tz == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating rx dma");
+		nfp_net_rx_queue_release(rxq);
+		return -ENOMEM;
+	}
+
+	/* Saving physical and virtual addresses for the RX ring */
+	rxq->dma = (uint64_t)tz->iova;
+	rxq->rxds = (struct nfp_net_rx_desc *)tz->addr;
+
+	/* mbuf pointers array for referencing mbufs linked to RX descriptors */
+	rxq->rxbufs = rte_zmalloc_socket("rxq->rxbufs",
+					 sizeof(*rxq->rxbufs) * nb_desc,
+					 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxq->rxbufs == NULL) {
+		nfp_net_rx_queue_release(rxq);
+		return -ENOMEM;
+	}
+
+	PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
+		   rxq->rxbufs, rxq->rxds, (unsigned long)rxq->dma);
+
+	nfp_net_reset_rx_queue(rxq);
+
+	dev->data->rx_queues[queue_idx] = rxq;
+	rxq->hw = hw;
+
+	/*
+	 * Telling the HW about the physical address of the RX ring and number
+	 * of descriptors in log2 format
+	 */
+	nn_cfg_writeq(hw, NFP_NET_CFG_RXR_ADDR(queue_idx), rxq->dma);
+	nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), rte_log2_u32(nb_desc));
+
+	return 0;
+}
+
+/*
+ * nfp_net_tx_free_bufs - Check for descriptors with a complete
+ * status
+ * @txq: TX queue to work with
+ * Returns number of descriptors freed
+ */
+static int
+nfp_net_tx_free_bufs(struct nfp_net_txq *txq)
+{
+	uint32_t qcp_rd_p;
+	int todo;
+
+	PMD_TX_LOG(DEBUG, "queue %u. Check for descriptor with a complete"
+		   " status", txq->qidx);
+
+	/* Work out how many packets have been sent */
+	qcp_rd_p = nfp_qcp_read(txq->qcp_q, NFP_QCP_READ_PTR);
+
+	if (qcp_rd_p == txq->rd_p) {
+		PMD_TX_LOG(DEBUG, "queue %u: It seems harrier is not sending "
+			   "packets (%u, %u)", txq->qidx,
+			   qcp_rd_p, txq->rd_p);
+		return 0;
+	}
+
+	if (qcp_rd_p > txq->rd_p)
+		todo = qcp_rd_p - txq->rd_p;
+	else
+		todo = qcp_rd_p + txq->tx_count - txq->rd_p;
+
+	PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->rd_p: %u, qcp->rd_p: %u",
+		   qcp_rd_p, txq->rd_p, txq->rd_p);
+
+	if (todo == 0)
+		return todo;
+
+	txq->rd_p += todo;
+	if (unlikely(txq->rd_p >= txq->tx_count))
+		txq->rd_p -= txq->tx_count;
+
+	return todo;
+}
+
+static void
+nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq)
+{
+	unsigned int i;
+
+	if (txq->txbufs == NULL)
+		return;
+
+	for (i = 0; i < txq->tx_count; i++) {
+		if (txq->txbufs[i].mbuf) {
+			rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
+			txq->txbufs[i].mbuf = NULL;
+		}
+	}
+}
+
+void
+nfp_net_tx_queue_release(void *tx_queue)
+{
+	struct nfp_net_txq *txq = tx_queue;
+
+	if (txq) {
+		nfp_net_tx_queue_release_mbufs(txq);
+		rte_free(txq->txbufs);
+		rte_free(txq);
+	}
+}
+
+void
+nfp_net_reset_tx_queue(struct nfp_net_txq *txq)
+{
+	nfp_net_tx_queue_release_mbufs(txq);
+	txq->wr_p = 0;
+	txq->rd_p = 0;
+}
+
+int
+nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+		       uint16_t nb_desc, unsigned int socket_id,
+		       const struct rte_eth_txconf *tx_conf)
+{
+	const struct rte_memzone *tz;
+	struct nfp_net_txq *txq;
+	uint16_t tx_free_thresh;
+	struct nfp_net_hw *hw;
+	uint32_t tx_desc_sz;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* Validating number of descriptors */
+	tx_desc_sz = nb_desc * sizeof(struct nfp_net_tx_desc);
+	if (tx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
+	    nb_desc > NFP_NET_MAX_TX_DESC ||
+	    nb_desc < NFP_NET_MIN_TX_DESC) {
+		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
+		return -EINVAL;
+	}
+
+	tx_free_thresh = (uint16_t)((tx_conf->tx_free_thresh) ?
+				    tx_conf->tx_free_thresh :
+				    DEFAULT_TX_FREE_THRESH);
+
+	if (tx_free_thresh > (nb_desc)) {
+		PMD_DRV_LOG(ERR,
+			"tx_free_thresh must be less than the number of TX "
+			"descriptors. (tx_free_thresh=%u port=%d "
+			"queue=%d)", (unsigned int)tx_free_thresh,
+			dev->data->port_id, (int)queue_idx);
+		return -(EINVAL);
+	}
+
+	/*
+	 * Free memory prior to re-allocation if needed. This is the case after
+	 * calling nfp_net_stop
+	 */
+	if (dev->data->tx_queues[queue_idx]) {
+		PMD_TX_LOG(DEBUG, "Freeing memory prior to re-allocation %d",
+			   queue_idx);
+		nfp_net_tx_queue_release(dev->data->tx_queues[queue_idx]);
+		dev->data->tx_queues[queue_idx] = NULL;
+	}
+
+	/* Allocating tx queue data structure */
+	txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct nfp_net_txq),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating tx dma");
+		return -ENOMEM;
+	}
+
+	/*
+	 * Allocate TX ring hardware descriptors. A memzone large enough to
+	 * handle the maximum ring size is allocated in order to allow for
+	 * resizing in later calls to the queue setup function.
+	 */
+	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
+				   sizeof(struct nfp_net_tx_desc) *
+				   NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN,
+				   socket_id);
+	if (tz == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating tx dma");
+		nfp_net_tx_queue_release(txq);
+		return -ENOMEM;
+	}
+
+	txq->tx_count = nb_desc;
+	txq->tx_free_thresh = tx_free_thresh;
+	txq->tx_pthresh = tx_conf->tx_thresh.pthresh;
+	txq->tx_hthresh = tx_conf->tx_thresh.hthresh;
+	txq->tx_wthresh = tx_conf->tx_thresh.wthresh;
+
+	/* queue mapping based on firmware configuration */
+	txq->qidx = queue_idx;
+	txq->tx_qcidx = queue_idx * hw->stride_tx;
+	txq->qcp_q = hw->tx_bar + NFP_QCP_QUEUE_OFF(txq->tx_qcidx);
+
+	txq->port_id = dev->data->port_id;
+
+	/* Saving physical and virtual addresses for the TX ring */
+	txq->dma = (uint64_t)tz->iova;
+	txq->txds = (struct nfp_net_tx_desc *)tz->addr;
+
+	/* mbuf pointers array for referencing mbufs linked to TX descriptors */
+	txq->txbufs = rte_zmalloc_socket("txq->txbufs",
+					 sizeof(*txq->txbufs) * nb_desc,
+					 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq->txbufs == NULL) {
+		nfp_net_tx_queue_release(txq);
+		return -ENOMEM;
+	}
+	PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
+		   txq->txbufs, txq->txds, (unsigned long)txq->dma);
+
+	nfp_net_reset_tx_queue(txq);
+
+	dev->data->tx_queues[queue_idx] = txq;
+	txq->hw = hw;
+
+	/*
+	 * Telling the HW about the physical address of the TX ring and number
+	 * of descriptors in log2 format
+	 */
+	nn_cfg_writeq(hw, NFP_NET_CFG_TXR_ADDR(queue_idx), txq->dma);
+	nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), rte_log2_u32(nb_desc));
+
+	return 0;
+}
+
+/* Leaving always free descriptors for avoiding wrapping confusion */
+static inline
+uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq)
+{
+	if (txq->wr_p >= txq->rd_p)
+		return txq->tx_count - (txq->wr_p - txq->rd_p) - 8;
+	else
+		return txq->rd_p - txq->wr_p - 8;
+}
+
+/*
+ * nfp_net_txq_full - Check if the TX queue free descriptors
+ * is below tx_free_threshold
+ *
+ * @txq: TX queue to check
+ *
+ * This function uses the host copy* of read/write pointers
+ */
+static inline
+uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
+{
+	return (nfp_free_tx_desc(txq) < txq->tx_free_thresh);
+}
+
+/* nfp_net_tx_tso - Set TX descriptor for TSO */
+static inline void
+nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+	       struct rte_mbuf *mb)
+{
+	uint64_t ol_flags;
+	struct nfp_net_hw *hw = txq->hw;
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY))
+		goto clean_txd;
+
+	ol_flags = mb->ol_flags;
+
+	if (!(ol_flags & PKT_TX_TCP_SEG))
+		goto clean_txd;
+
+	txd->l3_offset = mb->l2_len;
+	txd->l4_offset = mb->l2_len + mb->l3_len;
+	txd->lso_hdrlen = mb->l2_len + mb->l3_len + mb->l4_len;
+	txd->mss = rte_cpu_to_le_16(mb->tso_segsz);
+	txd->flags = PCIE_DESC_TX_LSO;
+	return;
+
+clean_txd:
+	txd->flags = 0;
+	txd->l3_offset = 0;
+	txd->l4_offset = 0;
+	txd->lso_hdrlen = 0;
+	txd->mss = 0;
+}
+
+/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
+static inline void
+nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+		 struct rte_mbuf *mb)
+{
+	uint64_t ol_flags;
+	struct nfp_net_hw *hw = txq->hw;
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_TXCSUM))
+		return;
+
+	ol_flags = mb->ol_flags;
+
+	/* IPv6 does not need checksum */
+	if (ol_flags & PKT_TX_IP_CKSUM)
+		txd->flags |= PCIE_DESC_TX_IP4_CSUM;
+
+	switch (ol_flags & PKT_TX_L4_MASK) {
+	case PKT_TX_UDP_CKSUM:
+		txd->flags |= PCIE_DESC_TX_UDP_CSUM;
+		break;
+	case PKT_TX_TCP_CKSUM:
+		txd->flags |= PCIE_DESC_TX_TCP_CSUM;
+		break;
+	}
+
+	if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
+		txd->flags |= PCIE_DESC_TX_CSUM;
+}
+
+uint16_t
+nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+	struct nfp_net_txq *txq;
+	struct nfp_net_hw *hw;
+	struct nfp_net_tx_desc *txds, txd;
+	struct rte_mbuf *pkt;
+	uint64_t dma_addr;
+	int pkt_size, dma_size;
+	uint16_t free_descs, issued_descs;
+	struct rte_mbuf **lmbuf;
+	int i;
+
+	txq = tx_queue;
+	hw = txq->hw;
+	txds = &txq->txds[txq->wr_p];
+
+	PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets",
+		   txq->qidx, txq->wr_p, nb_pkts);
+
+	if ((nfp_free_tx_desc(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
+		nfp_net_tx_free_bufs(txq);
+
+	free_descs = (uint16_t)nfp_free_tx_desc(txq);
+	if (unlikely(free_descs == 0))
+		return 0;
+
+	pkt = *tx_pkts;
+
+	i = 0;
+	issued_descs = 0;
+	PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets",
+		   txq->qidx, nb_pkts);
+	/* Sending packets */
+	while ((i < nb_pkts) && free_descs) {
+		/* Grabbing the mbuf linked to the current descriptor */
+		lmbuf = &txq->txbufs[txq->wr_p].mbuf;
+		/* Warming the cache for releasing the mbuf later on */
+		RTE_MBUF_PREFETCH_TO_FREE(*lmbuf);
+
+		pkt = *(tx_pkts + i);
+
+		if (unlikely(pkt->nb_segs > 1 &&
+			     !(hw->cap & NFP_NET_CFG_CTRL_GATHER))) {
+			PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set");
+			rte_panic("Multisegment packet unsupported\n");
+		}
+
+		/* Checking if we have enough descriptors */
+		if (unlikely(pkt->nb_segs > free_descs))
+			goto xmit_end;
+
+		/*
+		 * Checksum and VLAN flags just in the first descriptor for a
+		 * multisegment packet, but TSO info needs to be in all of them.
+		 */
+		txd.data_len = pkt->pkt_len;
+		nfp_net_tx_tso(txq, &txd, pkt);
+		nfp_net_tx_cksum(txq, &txd, pkt);
+
+		if ((pkt->ol_flags & PKT_TX_VLAN_PKT) &&
+		    (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)) {
+			txd.flags |= PCIE_DESC_TX_VLAN;
+			txd.vlan = pkt->vlan_tci;
+		}
+
+		/*
+		 * mbuf data_len is the data in one segment and pkt_len data
+		 * in the whole packet. When the packet is just one segment,
+		 * then data_len = pkt_len
+		 */
+		pkt_size = pkt->pkt_len;
+
+		while (pkt) {
+			/* Copying TSO, VLAN and cksum info */
+			*txds = txd;
+
+			/* Releasing mbuf used by this descriptor previously*/
+			if (*lmbuf)
+				rte_pktmbuf_free_seg(*lmbuf);
+
+			/*
+			 * Linking mbuf with descriptor for being released
+			 * next time descriptor is used
+			 */
+			*lmbuf = pkt;
+
+			dma_size = pkt->data_len;
+			dma_addr = rte_mbuf_data_iova(pkt);
+			PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
+				   "%" PRIx64 "", dma_addr);
+
+			/* Filling descriptors fields */
+			txds->dma_len = dma_size;
+			txds->data_len = txd.data_len;
+			txds->dma_addr_hi = (dma_addr >> 32) & 0xff;
+			txds->dma_addr_lo = (dma_addr & 0xffffffff);
+			ASSERT(free_descs > 0);
+			free_descs--;
+
+			txq->wr_p++;
+			if (unlikely(txq->wr_p == txq->tx_count)) /* wrapping?*/
+				txq->wr_p = 0;
+
+			pkt_size -= dma_size;
+
+			/*
+			 * Making the EOP, packets with just one segment
+			 * the priority
+			 */
+			if (likely(!pkt_size))
+				txds->offset_eop = PCIE_DESC_TX_EOP;
+			else
+				txds->offset_eop = 0;
+
+			pkt = pkt->next;
+			/* Referencing next free TX descriptor */
+			txds = &txq->txds[txq->wr_p];
+			lmbuf = &txq->txbufs[txq->wr_p].mbuf;
+			issued_descs++;
+		}
+		i++;
+	}
+
+xmit_end:
+	/* Increment write pointers. Force memory write before we let HW know */
+	rte_wmb();
+	nfp_qcp_ptr_add(txq->qcp_q, NFP_QCP_WRITE_PTR, issued_descs);
+
+	return i;
+}
diff --git a/drivers/net/nfp/nfp_rxtx.h b/drivers/net/nfp/nfp_rxtx.h
index 41a3a4b4e7..d2d0f3f175 100644
--- a/drivers/net/nfp/nfp_rxtx.h
+++ b/drivers/net/nfp/nfp_rxtx.h
@@ -17,6 +17,14 @@
 #include <linux/types.h>
 #include <rte_io.h>
 
+#define NFP_DESC_META_LEN(d) ((d)->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
+
+#define NFP_HASH_OFFSET      ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 4)
+#define NFP_HASH_TYPE_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 8)
+
+#define RTE_MBUF_DMA_ADDR_DEFAULT(mb) \
+	((uint64_t)((mb)->buf_iova + RTE_PKTMBUF_HEADROOM))
+
 /*
  * The maximum number of descriptors is limited by design as
  * DPDK uses uint16_t variables for these values
@@ -266,6 +274,25 @@ struct nfp_net_rxq {
 	int rx_qcidx;
 } __rte_aligned(64);
 
+int nfp_net_rx_freelist_setup(struct rte_eth_dev *dev);
+uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
+				       uint16_t queue_idx);
+uint16_t nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts);
+void nfp_net_rx_queue_release(void *rxq);
+void nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq);
+int nfp_net_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+				  uint16_t nb_desc, unsigned int socket_id,
+				  const struct rte_eth_rxconf *rx_conf,
+				  struct rte_mempool *mp);
+void nfp_net_tx_queue_release(void *txq);
+void nfp_net_reset_tx_queue(struct nfp_net_txq *txq);
+int nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+				  uint16_t nb_desc, unsigned int socket_id,
+				  const struct rte_eth_txconf *tx_conf);
+uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+				  uint16_t nb_pkts);
+
 #endif /* _NFP_RXTX_H_ */
 /*
  * Local variables:
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH 3/7] net/nfp: move CPP bridge to a separate file
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
@ 2021-07-16  8:23 ` Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
                   ` (4 subsequent siblings)
  7 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

This commit moves the CPP bridge logic to a separate file. A new
corresponding header file is also created.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
---
 drivers/net/nfp/meson.build      |   1 +
 drivers/net/nfp/nfp_cpp_bridge.c | 392 +++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_cpp_bridge.h |  36 +++
 drivers/net/nfp/nfp_net.c        | 367 +----------------------------
 4 files changed, 430 insertions(+), 366 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.c
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.h

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 1b289e2354..b46ac2d40f 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -20,4 +20,5 @@ sources = files(
         'nfpcore/nfp_hwinfo.c',
         'nfp_net.c',
         'nfp_rxtx.c',
+        'nfp_cpp_bridge.c',
 )
diff --git a/drivers/net/nfp/nfp_cpp_bridge.c b/drivers/net/nfp/nfp_cpp_bridge.c
new file mode 100644
index 0000000000..d916793338
--- /dev/null
+++ b/drivers/net/nfp/nfp_cpp_bridge.c
@@ -0,0 +1,392 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_cpp_bridge.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: CPP Bridge
+ */
+
+#include <rte_service_component.h>
+
+#include "nfpcore/nfp_cpp.h"
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_nsp.h"
+
+#include "nfp_net_logs.h"
+#include "nfp_cpp_bridge.h"
+
+#include <sys/ioctl.h>
+
+/* Prototypes */
+static int nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp);
+static int nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp);
+static int nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp);
+
+void nfp_register_cpp_service(struct nfp_cpp *cpp)
+{
+	uint32_t *cpp_service_id = NULL;
+	struct rte_service_spec service;
+
+	memset(&service, 0, sizeof(struct rte_service_spec));
+	snprintf(service.name, sizeof(service.name), "nfp_cpp_service");
+	service.callback = nfp_cpp_bridge_service_func;
+	service.callback_userdata = (void *)cpp;
+
+	if (rte_service_component_register(&service,
+					   cpp_service_id))
+		RTE_LOG(WARNING, PMD, "NFP CPP bridge service register() failed");
+	else
+		RTE_LOG(DEBUG, PMD, "NFP CPP bridge service registered");
+}
+
+/*
+ * Serving a write request to NFP from host programs. The request
+ * sends the write size and the CPP target. The bridge makes use
+ * of CPP interface handler configured by the PMD setup.
+ */
+static int
+nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp)
+{
+	struct nfp_cpp_area *area;
+	off_t offset, nfp_offset;
+	uint32_t cpp_id, pos, len;
+	uint32_t tmpbuf[16];
+	size_t count, curlen, totlen = 0;
+	int err = 0;
+
+	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
+		sizeof(off_t), sizeof(size_t));
+
+	/* Reading the count param */
+	err = recv(sockfd, &count, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	curlen = count;
+
+	/* Reading the offset param */
+	err = recv(sockfd, &offset, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	/* Obtain target's CPP ID and offset in target */
+	cpp_id = (offset >> 40) << 8;
+	nfp_offset = offset & ((1ull << 40) - 1);
+
+	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
+		offset);
+	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
+		cpp_id, nfp_offset);
+
+	/* Adjust length if not aligned */
+	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
+	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
+		curlen = NFP_CPP_MEMIO_BOUNDARY -
+			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
+	}
+
+	while (count > 0) {
+		/* configure a CPP PCIe2CPP BAR for mapping the CPP target */
+		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
+						    nfp_offset, curlen);
+		if (!area) {
+			RTE_LOG(ERR, PMD, "%s: area alloc fail\n", __func__);
+			return -EIO;
+		}
+
+		/* mapping the target */
+		err = nfp_cpp_area_acquire(area);
+		if (err < 0) {
+			RTE_LOG(ERR, PMD, "area acquire failed\n");
+			nfp_cpp_area_free(area);
+			return -EIO;
+		}
+
+		for (pos = 0; pos < curlen; pos += len) {
+			len = curlen - pos;
+			if (len > sizeof(tmpbuf))
+				len = sizeof(tmpbuf);
+
+			PMD_CPP_LOG(DEBUG, "%s: Receive %u of %zu\n", __func__,
+					   len, count);
+			err = recv(sockfd, tmpbuf, len, MSG_WAITALL);
+			if (err != (int)len) {
+				RTE_LOG(ERR, PMD,
+					"%s: error when receiving, %d of %zu\n",
+					__func__, err, count);
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+			err = nfp_cpp_area_write(area, pos, tmpbuf, len);
+			if (err < 0) {
+				RTE_LOG(ERR, PMD, "nfp_cpp_area_write error\n");
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+		}
+
+		nfp_offset += pos;
+		totlen += pos;
+		nfp_cpp_area_release(area);
+		nfp_cpp_area_free(area);
+
+		count -= pos;
+		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
+			 NFP_CPP_MEMIO_BOUNDARY : count;
+	}
+
+	return 0;
+}
+
+/*
+ * Serving a read request to NFP from host programs. The request
+ * sends the read size and the CPP target. The bridge makes use
+ * of CPP interface handler configured by the PMD setup. The read
+ * data is sent to the requester using the same socket.
+ */
+static int
+nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp)
+{
+	struct nfp_cpp_area *area;
+	off_t offset, nfp_offset;
+	uint32_t cpp_id, pos, len;
+	uint32_t tmpbuf[16];
+	size_t count, curlen, totlen = 0;
+	int err = 0;
+
+	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
+		sizeof(off_t), sizeof(size_t));
+
+	/* Reading the count param */
+	err = recv(sockfd, &count, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	curlen = count;
+
+	/* Reading the offset param */
+	err = recv(sockfd, &offset, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	/* Obtain target's CPP ID and offset in target */
+	cpp_id = (offset >> 40) << 8;
+	nfp_offset = offset & ((1ull << 40) - 1);
+
+	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
+			   offset);
+	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
+			   cpp_id, nfp_offset);
+
+	/* Adjust length if not aligned */
+	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
+	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
+		curlen = NFP_CPP_MEMIO_BOUNDARY -
+			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
+	}
+
+	while (count > 0) {
+		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
+						    nfp_offset, curlen);
+		if (!area) {
+			RTE_LOG(ERR, PMD, "%s: area alloc failed\n", __func__);
+			return -EIO;
+		}
+
+		err = nfp_cpp_area_acquire(area);
+		if (err < 0) {
+			RTE_LOG(ERR, PMD, "area acquire failed\n");
+			nfp_cpp_area_free(area);
+			return -EIO;
+		}
+
+		for (pos = 0; pos < curlen; pos += len) {
+			len = curlen - pos;
+			if (len > sizeof(tmpbuf))
+				len = sizeof(tmpbuf);
+
+			err = nfp_cpp_area_read(area, pos, tmpbuf, len);
+			if (err < 0) {
+				RTE_LOG(ERR, PMD, "nfp_cpp_area_read error\n");
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+			PMD_CPP_LOG(DEBUG, "%s: sending %u of %zu\n", __func__,
+					   len, count);
+
+			err = send(sockfd, tmpbuf, len, 0);
+			if (err != (int)len) {
+				RTE_LOG(ERR, PMD,
+					"%s: error when sending: %d of %zu\n",
+					__func__, err, count);
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+		}
+
+		nfp_offset += pos;
+		totlen += pos;
+		nfp_cpp_area_release(area);
+		nfp_cpp_area_free(area);
+
+		count -= pos;
+		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
+			NFP_CPP_MEMIO_BOUNDARY : count;
+	}
+	return 0;
+}
+
+/*
+ * Serving a ioctl command from host NFP tools. This usually goes to
+ * a kernel driver char driver but it is not available when the PF is
+ * bound to the PMD. Currently just one ioctl command is served and it
+ * does not require any CPP access at all.
+ */
+static int
+nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp)
+{
+	uint32_t cmd, ident_size, tmp;
+	int err;
+
+	/* Reading now the IOCTL command */
+	err = recv(sockfd, &cmd, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
+		return -EIO;
+	}
+
+	/* Only supporting NFP_IOCTL_CPP_IDENTIFICATION */
+	if (cmd != NFP_IOCTL_CPP_IDENTIFICATION) {
+		RTE_LOG(ERR, PMD, "%s: unknown cmd %d\n", __func__, cmd);
+		return -EINVAL;
+	}
+
+	err = recv(sockfd, &ident_size, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
+		return -EIO;
+	}
+
+	tmp = nfp_cpp_model(cpp);
+
+	PMD_CPP_LOG(DEBUG, "%s: sending NFP model %08x\n", __func__, tmp);
+
+	err = send(sockfd, &tmp, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
+		return -EIO;
+	}
+
+	tmp = cpp->interface;
+
+	PMD_CPP_LOG(DEBUG, "%s: sending NFP interface %08x\n", __func__, tmp);
+
+	err = send(sockfd, &tmp, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
+		return -EIO;
+	}
+
+	return 0;
+}
+
+/*
+ * This is the code to be executed by a service core. The CPP bridge interface
+ * is based on a unix socket and requests usually received by a kernel char
+ * driver, read, write and ioctl, are handled by the CPP bridge. NFP host tools
+ * can be executed with a wrapper library and LD_LIBRARY being completely
+ * unaware of the CPP bridge performing the NFP kernel char driver for CPP
+ * accesses.
+ */
+int32_t
+nfp_cpp_bridge_service_func(void *args)
+{
+	struct sockaddr address;
+	struct nfp_cpp *cpp = args;
+	int sockfd, datafd, op, ret;
+
+	unlink("/tmp/nfp_cpp");
+	sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
+	if (sockfd < 0) {
+		RTE_LOG(ERR, PMD, "%s: socket creation error. Service failed\n",
+			__func__);
+		return -EIO;
+	}
+
+	memset(&address, 0, sizeof(struct sockaddr));
+
+	address.sa_family = AF_UNIX;
+	strcpy(address.sa_data, "/tmp/nfp_cpp");
+
+	ret = bind(sockfd, (const struct sockaddr *)&address,
+		   sizeof(struct sockaddr));
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "%s: bind error (%d). Service failed\n",
+				  __func__, errno);
+		close(sockfd);
+		return ret;
+	}
+
+	ret = listen(sockfd, 20);
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "%s: listen error(%d). Service failed\n",
+				  __func__, errno);
+		close(sockfd);
+		return ret;
+	}
+
+	for (;;) {
+		datafd = accept(sockfd, NULL, NULL);
+		if (datafd < 0) {
+			RTE_LOG(ERR, PMD, "%s: accept call error (%d)\n",
+					  __func__, errno);
+			RTE_LOG(ERR, PMD, "%s: service failed\n", __func__);
+			close(sockfd);
+			return -EIO;
+		}
+
+		while (1) {
+			ret = recv(datafd, &op, 4, 0);
+			if (ret <= 0) {
+				PMD_CPP_LOG(DEBUG, "%s: socket close\n",
+						   __func__);
+				break;
+			}
+
+			PMD_CPP_LOG(DEBUG, "%s: getting op %u\n", __func__, op);
+
+			if (op == NFP_BRIDGE_OP_READ)
+				nfp_cpp_bridge_serve_read(datafd, cpp);
+
+			if (op == NFP_BRIDGE_OP_WRITE)
+				nfp_cpp_bridge_serve_write(datafd, cpp);
+
+			if (op == NFP_BRIDGE_OP_IOCTL)
+				nfp_cpp_bridge_serve_ioctl(datafd, cpp);
+
+			if (op == 0)
+				break;
+		}
+		close(datafd);
+	}
+	close(sockfd);
+
+	return 0;
+}
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_cpp_bridge.h b/drivers/net/nfp/nfp_cpp_bridge.h
new file mode 100644
index 0000000000..aea5fdc784
--- /dev/null
+++ b/drivers/net/nfp/nfp_cpp_bridge.h
@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_cpp_bridge.h
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: CPP Bridge header file
+ */
+
+#ifndef _NFP_CPP_BRIDGE_H_
+#define _NFP_CPP_BRIDGE_H_
+
+#define NFP_CPP_MEMIO_BOUNDARY	(1 << 20)
+#define NFP_BRIDGE_OP_READ	20
+#define NFP_BRIDGE_OP_WRITE	30
+#define NFP_BRIDGE_OP_IOCTL	40
+
+#define NFP_IOCTL 'n'
+#define NFP_IOCTL_CPP_IDENTIFICATION _IOW(NFP_IOCTL, 0x8f, uint32_t)
+
+void nfp_register_cpp_service(struct nfp_cpp *cpp);
+int32_t nfp_cpp_bridge_service_func(void *args);
+
+#endif /* _NFP_CPP_BRIDGE_H_ */
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 5bfc23ba04..d79c70c5b7 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -41,6 +41,7 @@
 #include "nfp_rxtx.h"
 #include "nfp_net_logs.h"
 #include "nfp_net_ctrl.h"
+#include "nfp_cpp_bridge.h"
 
 #include <sys/types.h>
 #include <sys/socket.h>
@@ -81,8 +82,6 @@ static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf);
 static int nfp_set_mac_addr(struct rte_eth_dev *dev,
 			     struct rte_ether_addr *mac_addr);
-static int32_t nfp_cpp_bridge_service_func(void *args);
-static void nfp_register_cpp_service(struct nfp_cpp *cpp);
 static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_cpp *cpp,
 			struct nfp_eth_table *nfp_eth_table,
@@ -1926,353 +1925,6 @@ nfp_net_init(struct rte_eth_dev *eth_dev)
 	return err;
 }
 
-#define NFP_CPP_MEMIO_BOUNDARY		(1 << 20)
-
-/*
- * Serving a write request to NFP from host programs. The request
- * sends the write size and the CPP target. The bridge makes use
- * of CPP interface handler configured by the PMD setup.
- */
-static int
-nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp)
-{
-	struct nfp_cpp_area *area;
-	off_t offset, nfp_offset;
-	uint32_t cpp_id, pos, len;
-	uint32_t tmpbuf[16];
-	size_t count, curlen, totlen = 0;
-	int err = 0;
-
-	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
-		sizeof(off_t), sizeof(size_t));
-
-	/* Reading the count param */
-	err = recv(sockfd, &count, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	curlen = count;
-
-	/* Reading the offset param */
-	err = recv(sockfd, &offset, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	/* Obtain target's CPP ID and offset in target */
-	cpp_id = (offset >> 40) << 8;
-	nfp_offset = offset & ((1ull << 40) - 1);
-
-	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
-		offset);
-	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
-		cpp_id, nfp_offset);
-
-	/* Adjust length if not aligned */
-	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
-	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
-		curlen = NFP_CPP_MEMIO_BOUNDARY -
-			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
-	}
-
-	while (count > 0) {
-		/* configure a CPP PCIe2CPP BAR for mapping the CPP target */
-		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
-						    nfp_offset, curlen);
-		if (!area) {
-			RTE_LOG(ERR, PMD, "%s: area alloc fail\n", __func__);
-			return -EIO;
-		}
-
-		/* mapping the target */
-		err = nfp_cpp_area_acquire(area);
-		if (err < 0) {
-			RTE_LOG(ERR, PMD, "area acquire failed\n");
-			nfp_cpp_area_free(area);
-			return -EIO;
-		}
-
-		for (pos = 0; pos < curlen; pos += len) {
-			len = curlen - pos;
-			if (len > sizeof(tmpbuf))
-				len = sizeof(tmpbuf);
-
-			PMD_CPP_LOG(DEBUG, "%s: Receive %u of %zu\n", __func__,
-					   len, count);
-			err = recv(sockfd, tmpbuf, len, MSG_WAITALL);
-			if (err != (int)len) {
-				RTE_LOG(ERR, PMD,
-					"%s: error when receiving, %d of %zu\n",
-					__func__, err, count);
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-			err = nfp_cpp_area_write(area, pos, tmpbuf, len);
-			if (err < 0) {
-				RTE_LOG(ERR, PMD, "nfp_cpp_area_write error\n");
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-		}
-
-		nfp_offset += pos;
-		totlen += pos;
-		nfp_cpp_area_release(area);
-		nfp_cpp_area_free(area);
-
-		count -= pos;
-		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
-			 NFP_CPP_MEMIO_BOUNDARY : count;
-	}
-
-	return 0;
-}
-
-/*
- * Serving a read request to NFP from host programs. The request
- * sends the read size and the CPP target. The bridge makes use
- * of CPP interface handler configured by the PMD setup. The read
- * data is sent to the requester using the same socket.
- */
-static int
-nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp)
-{
-	struct nfp_cpp_area *area;
-	off_t offset, nfp_offset;
-	uint32_t cpp_id, pos, len;
-	uint32_t tmpbuf[16];
-	size_t count, curlen, totlen = 0;
-	int err = 0;
-
-	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
-		sizeof(off_t), sizeof(size_t));
-
-	/* Reading the count param */
-	err = recv(sockfd, &count, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	curlen = count;
-
-	/* Reading the offset param */
-	err = recv(sockfd, &offset, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	/* Obtain target's CPP ID and offset in target */
-	cpp_id = (offset >> 40) << 8;
-	nfp_offset = offset & ((1ull << 40) - 1);
-
-	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
-			   offset);
-	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
-			   cpp_id, nfp_offset);
-
-	/* Adjust length if not aligned */
-	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
-	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
-		curlen = NFP_CPP_MEMIO_BOUNDARY -
-			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
-	}
-
-	while (count > 0) {
-		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
-						    nfp_offset, curlen);
-		if (!area) {
-			RTE_LOG(ERR, PMD, "%s: area alloc failed\n", __func__);
-			return -EIO;
-		}
-
-		err = nfp_cpp_area_acquire(area);
-		if (err < 0) {
-			RTE_LOG(ERR, PMD, "area acquire failed\n");
-			nfp_cpp_area_free(area);
-			return -EIO;
-		}
-
-		for (pos = 0; pos < curlen; pos += len) {
-			len = curlen - pos;
-			if (len > sizeof(tmpbuf))
-				len = sizeof(tmpbuf);
-
-			err = nfp_cpp_area_read(area, pos, tmpbuf, len);
-			if (err < 0) {
-				RTE_LOG(ERR, PMD, "nfp_cpp_area_read error\n");
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-			PMD_CPP_LOG(DEBUG, "%s: sending %u of %zu\n", __func__,
-					   len, count);
-
-			err = send(sockfd, tmpbuf, len, 0);
-			if (err != (int)len) {
-				RTE_LOG(ERR, PMD,
-					"%s: error when sending: %d of %zu\n",
-					__func__, err, count);
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-		}
-
-		nfp_offset += pos;
-		totlen += pos;
-		nfp_cpp_area_release(area);
-		nfp_cpp_area_free(area);
-
-		count -= pos;
-		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
-			NFP_CPP_MEMIO_BOUNDARY : count;
-	}
-	return 0;
-}
-
-#define NFP_IOCTL 'n'
-#define NFP_IOCTL_CPP_IDENTIFICATION _IOW(NFP_IOCTL, 0x8f, uint32_t)
-/*
- * Serving a ioctl command from host NFP tools. This usually goes to
- * a kernel driver char driver but it is not available when the PF is
- * bound to the PMD. Currently just one ioctl command is served and it
- * does not require any CPP access at all.
- */
-static int
-nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp)
-{
-	uint32_t cmd, ident_size, tmp;
-	int err;
-
-	/* Reading now the IOCTL command */
-	err = recv(sockfd, &cmd, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
-		return -EIO;
-	}
-
-	/* Only supporting NFP_IOCTL_CPP_IDENTIFICATION */
-	if (cmd != NFP_IOCTL_CPP_IDENTIFICATION) {
-		RTE_LOG(ERR, PMD, "%s: unknown cmd %d\n", __func__, cmd);
-		return -EINVAL;
-	}
-
-	err = recv(sockfd, &ident_size, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
-		return -EIO;
-	}
-
-	tmp = nfp_cpp_model(cpp);
-
-	PMD_CPP_LOG(DEBUG, "%s: sending NFP model %08x\n", __func__, tmp);
-
-	err = send(sockfd, &tmp, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
-		return -EIO;
-	}
-
-	tmp = cpp->interface;
-
-	PMD_CPP_LOG(DEBUG, "%s: sending NFP interface %08x\n", __func__, tmp);
-
-	err = send(sockfd, &tmp, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
-		return -EIO;
-	}
-
-	return 0;
-}
-
-#define NFP_BRIDGE_OP_READ	20
-#define NFP_BRIDGE_OP_WRITE	30
-#define NFP_BRIDGE_OP_IOCTL	40
-
-/*
- * This is the code to be executed by a service core. The CPP bridge interface
- * is based on a unix socket and requests usually received by a kernel char
- * driver, read, write and ioctl, are handled by the CPP bridge. NFP host tools
- * can be executed with a wrapper library and LD_LIBRARY being completely
- * unaware of the CPP bridge performing the NFP kernel char driver for CPP
- * accesses.
- */
-static int32_t
-nfp_cpp_bridge_service_func(void *args)
-{
-	struct sockaddr address;
-	struct nfp_cpp *cpp = args;
-	int sockfd, datafd, op, ret;
-
-	unlink("/tmp/nfp_cpp");
-	sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
-	if (sockfd < 0) {
-		RTE_LOG(ERR, PMD, "%s: socket creation error. Service failed\n",
-			__func__);
-		return -EIO;
-	}
-
-	memset(&address, 0, sizeof(struct sockaddr));
-
-	address.sa_family = AF_UNIX;
-	strcpy(address.sa_data, "/tmp/nfp_cpp");
-
-	ret = bind(sockfd, (const struct sockaddr *)&address,
-		   sizeof(struct sockaddr));
-	if (ret < 0) {
-		RTE_LOG(ERR, PMD, "%s: bind error (%d). Service failed\n",
-				  __func__, errno);
-		close(sockfd);
-		return ret;
-	}
-
-	ret = listen(sockfd, 20);
-	if (ret < 0) {
-		RTE_LOG(ERR, PMD, "%s: listen error(%d). Service failed\n",
-				  __func__, errno);
-		close(sockfd);
-		return ret;
-	}
-
-	for (;;) {
-		datafd = accept(sockfd, NULL, NULL);
-		if (datafd < 0) {
-			RTE_LOG(ERR, PMD, "%s: accept call error (%d)\n",
-					  __func__, errno);
-			RTE_LOG(ERR, PMD, "%s: service failed\n", __func__);
-			close(sockfd);
-			return -EIO;
-		}
-
-		while (1) {
-			ret = recv(datafd, &op, 4, 0);
-			if (ret <= 0) {
-				PMD_CPP_LOG(DEBUG, "%s: socket close\n",
-						   __func__);
-				break;
-			}
-
-			PMD_CPP_LOG(DEBUG, "%s: getting op %u\n", __func__, op);
-
-			if (op == NFP_BRIDGE_OP_READ)
-				nfp_cpp_bridge_serve_read(datafd, cpp);
-
-			if (op == NFP_BRIDGE_OP_WRITE)
-				nfp_cpp_bridge_serve_write(datafd, cpp);
-
-			if (op == NFP_BRIDGE_OP_IOCTL)
-				nfp_cpp_bridge_serve_ioctl(datafd, cpp);
-
-			if (op == 0)
-				break;
-		}
-		close(datafd);
-	}
-	close(sockfd);
-
-	return 0;
-}
-
 #define DEFAULT_FW_PATH       "/lib/firmware/netronome"
 
 static int
@@ -2491,23 +2143,6 @@ static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
 	return ret;
 }
 
-static void nfp_register_cpp_service(struct nfp_cpp *cpp)
-{
-	uint32_t *cpp_service_id = NULL;
-	struct rte_service_spec service;
-
-	memset(&service, 0, sizeof(struct rte_service_spec));
-	snprintf(service.name, sizeof(service.name), "nfp_cpp_service");
-	service.callback = nfp_cpp_bridge_service_func;
-	service.callback_userdata = (void *)cpp;
-
-	if (rte_service_component_register(&service,
-					   cpp_service_id))
-		RTE_LOG(WARNING, PMD, "NFP CPP bridge service register() failed");
-	else
-		RTE_LOG(DEBUG, PMD, "NFP CPP bridge service registered");
-}
-
 static int nfp_pf_init(struct rte_pci_device *pci_dev)
 {
 	struct nfp_pf_dev *pf_dev = NULL;
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH 4/7] net/nfp: prototype common functions in header file
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
                   ` (2 preceding siblings ...)
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
@ 2021-07-16  8:23 ` Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
                   ` (3 subsequent siblings)
  7 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

The majority of "ethdev" type functions are used for both PF devices and
VF devices. Prototype these functions in the nfp_net_pmd header file in
preparation of splitting PF and VF specific functions.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
---
 drivers/net/nfp/nfp_net.c     | 87 +++++++++++++----------------------
 drivers/net/nfp/nfp_net_pmd.h | 49 ++++++++++++++++++++
 2 files changed, 81 insertions(+), 55 deletions(-)

diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index d79c70c5b7..da35bba4ef 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -53,35 +53,12 @@
 
 /* Prototypes */
 static int nfp_net_close(struct rte_eth_dev *dev);
-static int nfp_net_configure(struct rte_eth_dev *dev);
-static void nfp_net_dev_interrupt_handler(void *param);
-static void nfp_net_dev_interrupt_delayed_handler(void *param);
-static int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
-static int nfp_net_infos_get(struct rte_eth_dev *dev,
-			     struct rte_eth_dev_info *dev_info);
 static int nfp_net_init(struct rte_eth_dev *eth_dev);
 static int nfp_pf_init(struct rte_pci_device *pci_dev);
 static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
 static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
-static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
-static int nfp_net_promisc_enable(struct rte_eth_dev *dev);
-static int nfp_net_promisc_disable(struct rte_eth_dev *dev);
-static int nfp_net_start(struct rte_eth_dev *dev);
-static int nfp_net_stats_get(struct rte_eth_dev *dev,
-			      struct rte_eth_stats *stats);
-static int nfp_net_stats_reset(struct rte_eth_dev *dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
-static int nfp_net_rss_config_default(struct rte_eth_dev *dev);
-static int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
-				   struct rte_eth_rss_conf *rss_conf);
-static int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
-		    struct rte_eth_rss_reta_entry64 *reta_conf,
-		    uint16_t reta_size);
-static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
-			struct rte_eth_rss_conf *rss_conf);
-static int nfp_set_mac_addr(struct rte_eth_dev *dev,
-			     struct rte_ether_addr *mac_addr);
 static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_cpp *cpp,
 			struct nfp_eth_table *nfp_eth_table,
@@ -136,7 +113,7 @@ __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
  * Write the update word to the BAR and ping the reconfig queue. Then poll
  * until the firmware has acknowledged the update by zeroing the update word.
  */
-static int
+int
 nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update)
 {
 	uint32_t err;
@@ -172,7 +149,7 @@ nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update)
  * before any other function in the Ethernet API. This function can
  * also be re-invoked when a device is in the stopped state.
  */
-static int
+int
 nfp_net_configure(struct rte_eth_dev *dev)
 {
 	struct rte_eth_conf *dev_conf;
@@ -215,7 +192,7 @@ nfp_net_configure(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static void
+void
 nfp_net_enable_queues(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -239,7 +216,7 @@ nfp_net_enable_queues(struct rte_eth_dev *dev)
 	nn_cfg_writeq(hw, NFP_NET_CFG_RXRS_ENABLE, enabled_queues);
 }
 
-static void
+void
 nfp_net_disable_queues(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -264,14 +241,14 @@ nfp_net_disable_queues(struct rte_eth_dev *dev)
 	hw->ctrl = new_ctrl;
 }
 
-static void
+void
 nfp_net_params_setup(struct nfp_net_hw *hw)
 {
 	nn_cfg_writel(hw, NFP_NET_CFG_MTU, hw->mtu);
 	nn_cfg_writel(hw, NFP_NET_CFG_FLBUFSZ, hw->flbufsz);
 }
 
-static void
+void
 nfp_net_cfg_queue_setup(struct nfp_net_hw *hw)
 {
 	hw->qcp_cfg = hw->tx_bar + NFP_QCP_QUEUE_ADDR_SZ;
@@ -279,7 +256,7 @@ nfp_net_cfg_queue_setup(struct nfp_net_hw *hw)
 
 #define ETH_ADDR_LEN	6
 
-static void
+void
 nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
 {
 	int i;
@@ -318,7 +295,7 @@ nfp_net_vf_read_mac(struct nfp_net_hw *hw)
 	memcpy(&hw->mac_addr[4], &tmp, 2);
 }
 
-static void
+void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
 	uint32_t mac0 = *(uint32_t *)mac;
@@ -366,7 +343,7 @@ nfp_set_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr)
 	return 0;
 }
 
-static int
+int
 nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
 			   struct rte_intr_handle *intr_handle)
 {
@@ -410,7 +387,7 @@ nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static uint32_t
+uint32_t
 nfp_check_offloads(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -746,7 +723,7 @@ nfp_net_close(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_promisc_enable(struct rte_eth_dev *dev)
 {
 	uint32_t new_ctrl, update = 0;
@@ -783,7 +760,7 @@ nfp_net_promisc_enable(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_promisc_disable(struct rte_eth_dev *dev)
 {
 	uint32_t new_ctrl, update = 0;
@@ -819,7 +796,7 @@ nfp_net_promisc_disable(struct rte_eth_dev *dev)
  * Wait to complete is needed as it can take up to 9 seconds to get the Link
  * status.
  */
-static int
+int
 nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
 {
 	struct nfp_net_hw *hw;
@@ -869,7 +846,7 @@ nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
 	return ret;
 }
 
-static int
+int
 nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 {
 	int i;
@@ -964,7 +941,7 @@ nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 	return -EINVAL;
 }
 
-static int
+int
 nfp_net_stats_reset(struct rte_eth_dev *dev)
 {
 	int i;
@@ -1029,7 +1006,7 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 {
 	struct nfp_net_hw *hw;
@@ -1123,7 +1100,7 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 	return 0;
 }
 
-static const uint32_t *
+const uint32_t *
 nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 {
 	static const uint32_t ptypes[] = {
@@ -1140,7 +1117,7 @@ nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 	return NULL;
 }
 
-static int
+int
 nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
 	struct rte_pci_device *pci_dev;
@@ -1160,7 +1137,7 @@ nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 	return 0;
 }
 
-static int
+int
 nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
 	struct rte_pci_device *pci_dev;
@@ -1179,7 +1156,7 @@ nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
 	return 0;
 }
 
-static void
+void
 nfp_net_dev_link_status_print(struct rte_eth_dev *dev)
 {
 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
@@ -1208,7 +1185,7 @@ nfp_net_dev_link_status_print(struct rte_eth_dev *dev)
  * If MSI-X auto-masking is enabled clear the mask bit, otherwise
  * clear the ICR for the entry.
  */
-static void
+void
 nfp_net_irq_unmask(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -1229,7 +1206,7 @@ nfp_net_irq_unmask(struct rte_eth_dev *dev)
 	}
 }
 
-static void
+void
 nfp_net_dev_interrupt_handler(void *param)
 {
 	int64_t timeout;
@@ -1272,7 +1249,7 @@ nfp_net_dev_interrupt_handler(void *param)
  *
  * @return  void
  */
-static void
+void
 nfp_net_dev_interrupt_delayed_handler(void *param)
 {
 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
@@ -1286,7 +1263,7 @@ nfp_net_dev_interrupt_delayed_handler(void *param)
 	nfp_net_irq_unmask(dev);
 }
 
-static int
+int
 nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 {
 	struct nfp_net_hw *hw;
@@ -1321,7 +1298,7 @@ nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	return 0;
 }
 
-static int
+int
 nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 {
 	uint32_t new_ctrl, update;
@@ -1353,7 +1330,7 @@ nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 	return ret;
 }
 
-static int
+int
 nfp_net_rss_reta_write(struct rte_eth_dev *dev,
 		    struct rte_eth_rss_reta_entry64 *reta_conf,
 		    uint16_t reta_size)
@@ -1404,7 +1381,7 @@ nfp_net_rss_reta_write(struct rte_eth_dev *dev,
 }
 
 /* Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device */
-static int
+int
 nfp_net_reta_update(struct rte_eth_dev *dev,
 		    struct rte_eth_rss_reta_entry64 *reta_conf,
 		    uint16_t reta_size)
@@ -1430,7 +1407,7 @@ nfp_net_reta_update(struct rte_eth_dev *dev,
 }
 
  /* Query Redirection Table(RETA) of Receive Side Scaling of Ethernet device. */
-static int
+int
 nfp_net_reta_query(struct rte_eth_dev *dev,
 		   struct rte_eth_rss_reta_entry64 *reta_conf,
 		   uint16_t reta_size)
@@ -1477,7 +1454,7 @@ nfp_net_reta_query(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf)
 {
@@ -1527,7 +1504,7 @@ nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf)
 {
@@ -1563,7 +1540,7 @@ nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
 			  struct rte_eth_rss_conf *rss_conf)
 {
@@ -1614,7 +1591,7 @@ nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_config_default(struct rte_eth_dev *dev)
 {
 	struct rte_eth_conf *dev_conf;
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index 9265496bf0..dc05e888df 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -349,6 +349,55 @@ nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
 		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
 }
 
+/* Prototypes for common NFP functions */
+int nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update);
+int nfp_net_configure(struct rte_eth_dev *dev);
+void nfp_net_enable_queues(struct rte_eth_dev *dev);
+void nfp_net_disable_queues(struct rte_eth_dev *dev);
+void nfp_net_params_setup(struct nfp_net_hw *hw);
+void nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src);
+void nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac);
+int nfp_set_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr);
+int nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
+			       struct rte_intr_handle *intr_handle);
+uint32_t nfp_check_offloads(struct rte_eth_dev *dev);
+int nfp_net_promisc_enable(struct rte_eth_dev *dev);
+int nfp_net_promisc_disable(struct rte_eth_dev *dev);
+int nfp_net_link_update(struct rte_eth_dev *dev,
+			__rte_unused int wait_to_complete);
+int nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats);
+int nfp_net_stats_reset(struct rte_eth_dev *dev);
+int nfp_net_infos_get(struct rte_eth_dev *dev,
+		      struct rte_eth_dev_info *dev_info);
+const uint32_t *nfp_net_supported_ptypes_get(struct rte_eth_dev *dev);
+int nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
+int nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
+void nfp_net_params_setup(struct nfp_net_hw *hw);
+void nfp_net_cfg_queue_setup(struct nfp_net_hw *hw);
+void nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src);
+void nfp_net_dev_link_status_print(struct rte_eth_dev *dev);
+void nfp_net_irq_unmask(struct rte_eth_dev *dev);
+void nfp_net_dev_interrupt_handler(void *param);
+void nfp_net_dev_interrupt_delayed_handler(void *param);
+int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
+int nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask);
+int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
+			   struct rte_eth_rss_reta_entry64 *reta_conf,
+			   uint16_t reta_size);
+int nfp_net_reta_update(struct rte_eth_dev *dev,
+			struct rte_eth_rss_reta_entry64 *reta_conf,
+			uint16_t reta_size);
+int nfp_net_reta_query(struct rte_eth_dev *dev,
+		       struct rte_eth_rss_reta_entry64 *reta_conf,
+		       uint16_t reta_size);
+int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
+			   struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
+			    struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
+			      struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_config_default(struct rte_eth_dev *dev);
+
 #define NFP_NET_DEV_PRIVATE_TO_HW(adapter)\
 	(&((struct nfp_net_adapter *)adapter)->hw)
 
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH 5/7] net/nfp: move VF functions into new file
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
                   ` (3 preceding siblings ...)
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
@ 2021-07-16  8:23 ` Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 6/7] net/nfp: move PF " Heinrich Kuhn
                   ` (2 subsequent siblings)
  7 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

Move any ethdev functionality specific to VF devices into a new file
called nfp_ethdev_vf.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
---
 drivers/net/nfp/meson.build     |   1 +
 drivers/net/nfp/nfp_ethdev_vf.c | 504 ++++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_net.c       |  42 +--
 3 files changed, 506 insertions(+), 41 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_ethdev_vf.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index b46ac2d40f..34f4054b3c 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -21,4 +21,5 @@ sources = files(
         'nfp_net.c',
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
+        'nfp_ethdev_vf.c',
 )
diff --git a/drivers/net/nfp/nfp_ethdev_vf.c b/drivers/net/nfp/nfp_ethdev_vf.c
new file mode 100644
index 0000000000..223142c0ed
--- /dev/null
+++ b/drivers/net/nfp/nfp_ethdev_vf.c
@@ -0,0 +1,504 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_ethdev_vf.c
+ *
+ * Netronome vNIC  VF DPDK Poll-Mode Driver: Main entry point
+ */
+
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_rtsym.h"
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+
+static void nfp_netvf_read_mac(struct nfp_net_hw *hw);
+static int nfp_netvf_start(struct rte_eth_dev *dev);
+static int nfp_netvf_stop(struct rte_eth_dev *dev);
+static int nfp_netvf_set_link_up(struct rte_eth_dev *dev);
+static int nfp_netvf_set_link_down(struct rte_eth_dev *dev);
+static int nfp_netvf_close(struct rte_eth_dev *dev);
+static int nfp_netvf_init(struct rte_eth_dev *eth_dev);
+static int nfp_vf_pci_uninit(struct rte_eth_dev *eth_dev);
+static int eth_nfp_vf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+	struct rte_pci_device *pci_dev);
+static int eth_nfp_vf_pci_remove(struct rte_pci_device *pci_dev);
+
+static void
+nfp_netvf_read_mac(struct nfp_net_hw *hw)
+{
+	uint32_t tmp;
+
+	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR));
+	memcpy(&hw->mac_addr[0], &tmp, 4);
+
+	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR + 4));
+	memcpy(&hw->mac_addr[4], &tmp, 2);
+}
+
+static int
+nfp_netvf_start(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+	uint32_t new_ctrl, update = 0;
+	struct nfp_net_hw *hw;
+	struct rte_eth_conf *dev_conf;
+	struct rte_eth_rxmode *rxmode;
+	uint32_t intr_vector;
+	int ret;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_LOG(DEBUG, "Start");
+
+	/* Disabling queues just in case... */
+	nfp_net_disable_queues(dev);
+
+	/* Enabling the required queues in the device */
+	nfp_net_enable_queues(dev);
+
+	/* check and configure queue intr-vector mapping */
+	if (dev->data->dev_conf.intr_conf.rxq != 0) {
+		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+			/*
+			 * Better not to share LSC with RX interrupts.
+			 * Unregistering LSC interrupt handler
+			 */
+			rte_intr_callback_unregister(&pci_dev->intr_handle,
+				nfp_net_dev_interrupt_handler, (void *)dev);
+
+			if (dev->data->nb_rx_queues > 1) {
+				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+					     "supports 1 queue with UIO");
+				return -EIO;
+			}
+		}
+		intr_vector = dev->data->nb_rx_queues;
+		if (rte_intr_efd_enable(intr_handle, intr_vector))
+			return -1;
+
+		nfp_configure_rx_interrupt(dev, intr_handle);
+		update = NFP_NET_CFG_UPDATE_MSIX;
+	}
+
+	rte_intr_enable(intr_handle);
+
+	new_ctrl = nfp_check_offloads(dev);
+
+	/* Writing configuration parameters in the device */
+	nfp_net_params_setup(hw);
+
+	dev_conf = &dev->data->dev_conf;
+	rxmode = &dev_conf->rxmode;
+
+	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
+		nfp_net_rss_config_default(dev);
+		update |= NFP_NET_CFG_UPDATE_RSS;
+		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
+	}
+
+	/* Enable device */
+	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
+
+	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+
+	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
+		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
+
+	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
+	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+		return -EIO;
+
+	/*
+	 * Allocating rte mbufs for configured rx queues.
+	 * This requires queues being enabled before
+	 */
+	if (nfp_net_rx_freelist_setup(dev) < 0) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	hw->ctrl = new_ctrl;
+
+	return 0;
+
+error:
+	/*
+	 * An error returned by this function should mean the app
+	 * exiting and then the system releasing all the memory
+	 * allocated even memory coming from hugepages.
+	 *
+	 * The device could be enabled at this point with some queues
+	 * ready for getting packets. This is true if the call to
+	 * nfp_net_rx_freelist_setup() succeeds for some queues but
+	 * fails for subsequent queues.
+	 *
+	 * This should make the app exiting but better if we tell the
+	 * device first.
+	 */
+	nfp_net_disable_queues(dev);
+
+	return ret;
+}
+
+static int
+nfp_netvf_stop(struct rte_eth_dev *dev)
+{
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	PMD_INIT_LOG(DEBUG, "Stop");
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	return 0;
+}
+
+static int
+nfp_netvf_set_link_up(struct rte_eth_dev *dev __rte_unused)
+{
+	return -ENOTSUP;
+}
+
+/* Set the link down. */
+static int
+nfp_netvf_set_link_down(struct rte_eth_dev *dev __rte_unused)
+{
+	return -ENOTSUP;
+}
+
+/* Reset and stop device. The device can not be restarted. */
+static int
+nfp_netvf_close(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	PMD_INIT_LOG(DEBUG, "Close");
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+	/*
+	 * We assume that the DPDK application is stopping all the
+	 * threads/queues before calling the device close function.
+	 */
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q =  (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q =  (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	rte_intr_disable(&pci_dev->intr_handle);
+
+	/* unregister callback func from eal lib */
+	rte_intr_callback_unregister(&pci_dev->intr_handle,
+				     nfp_net_dev_interrupt_handler,
+				     (void *)dev);
+
+	/*
+	 * The ixgbe PMD driver disables the pcie master on the
+	 * device. The i40e does not...
+	 */
+
+	return 0;
+}
+
+/* Initialise and register VF driver with DPDK Application */
+static const struct eth_dev_ops nfp_netvf_eth_dev_ops = {
+	.dev_configure		= nfp_net_configure,
+	.dev_start		= nfp_netvf_start,
+	.dev_stop		= nfp_netvf_stop,
+	.dev_set_link_up	= nfp_netvf_set_link_up,
+	.dev_set_link_down	= nfp_netvf_set_link_down,
+	.dev_close		= nfp_netvf_close,
+	.promiscuous_enable	= nfp_net_promisc_enable,
+	.promiscuous_disable	= nfp_net_promisc_disable,
+	.link_update		= nfp_net_link_update,
+	.stats_get		= nfp_net_stats_get,
+	.stats_reset		= nfp_net_stats_reset,
+	.dev_infos_get		= nfp_net_infos_get,
+	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
+	.mtu_set		= nfp_net_dev_mtu_set,
+	.mac_addr_set           = nfp_set_mac_addr,
+	.vlan_offload_set	= nfp_net_vlan_offload_set,
+	.reta_update		= nfp_net_reta_update,
+	.reta_query		= nfp_net_reta_query,
+	.rss_hash_update	= nfp_net_rss_hash_update,
+	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
+	.rx_queue_setup		= nfp_net_rx_queue_setup,
+	.rx_queue_release	= nfp_net_rx_queue_release,
+	.tx_queue_setup		= nfp_net_tx_queue_setup,
+	.tx_queue_release	= nfp_net_tx_queue_release,
+	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
+	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
+};
+
+static int
+nfp_netvf_init(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_net_hw *hw;
+	struct rte_ether_addr *tmp_ether_addr;
+
+	uint64_t tx_bar_off = 0, rx_bar_off = 0;
+	uint32_t start_q;
+	int stride = 4;
+	int port = 0;
+	int err;
+
+	PMD_INIT_FUNC_TRACE();
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* NFP can not handle DMA addresses requiring more than 40 bits */
+	if (rte_mem_check_dma_mask(40)) {
+		RTE_LOG(ERR, PMD, "device %s can not be used:",
+				   pci_dev->device.name);
+		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
+		return -ENODEV;
+	};
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+
+	eth_dev->dev_ops = &nfp_netvf_eth_dev_ops;
+	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	rte_eth_copy_pci_info(eth_dev, pci_dev);
+
+	hw->device_id = pci_dev->id.device_id;
+	hw->vendor_id = pci_dev->id.vendor_id;
+	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
+	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
+
+	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
+		     pci_dev->id.vendor_id, pci_dev->id.device_id,
+		     pci_dev->addr.domain, pci_dev->addr.bus,
+		     pci_dev->addr.devid, pci_dev->addr.function);
+
+	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
+	if (hw->ctrl_bar == NULL) {
+		PMD_DRV_LOG(ERR,
+			"hw->ctrl_bar is NULL. BAR0 not configured");
+		return -ENODEV;
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+
+	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
+	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
+
+	/* Work out where in the BAR the queues start. */
+	switch (pci_dev->id.device_id) {
+	case PCI_DEVICE_ID_NFP6000_VF_NIC:
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
+		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
+		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
+		err = -ENODEV;
+		goto dev_err_ctrl_map;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
+	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
+
+	hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+		     tx_bar_off;
+	hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+		     rx_bar_off;
+
+	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
+		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
+
+	nfp_net_cfg_queue_setup(hw);
+
+	/* Get some of the read-only fields from the config BAR */
+	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
+	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
+	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
+	hw->mtu = RTE_ETHER_MTU;
+
+	/* VLAN insertion is incompatible with LSOv2 */
+	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
+		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
+
+	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
+		hw->rx_offset = NFP_NET_RX_OFFSET;
+	else
+		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
+
+	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
+			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
+			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
+
+	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
+		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
+
+	hw->ctrl = 0;
+
+	hw->stride_rx = stride;
+	hw->stride_tx = stride;
+
+	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
+		     hw->max_rx_queues, hw->max_tx_queues);
+
+	/* Initializing spinlock for reconfigs */
+	rte_spinlock_init(&hw->reconfig_lock);
+
+	/* Allocating memory for mac addr */
+	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
+					       RTE_ETHER_ADDR_LEN, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
+		err = -ENOMEM;
+		goto dev_err_queues_map;
+	}
+
+	nfp_netvf_read_mac(hw);
+
+	tmp_ether_addr = (struct rte_ether_addr *)&hw->mac_addr;
+	if (!rte_is_valid_assigned_ether_addr(tmp_ether_addr)) {
+		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
+				   port);
+		/* Using random mac addresses for VFs */
+		rte_eth_random_addr(&hw->mac_addr[0]);
+		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+	}
+
+	/* Copying mac address to DPDK eth_dev struct */
+	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
+			&eth_dev->data->mac_addrs[0]);
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
+
+	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
+
+	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
+		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
+		     eth_dev->data->port_id, pci_dev->id.vendor_id,
+		     pci_dev->id.device_id,
+		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
+		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		/* Registering LSC interrupt handler */
+		rte_intr_callback_register(&pci_dev->intr_handle,
+					   nfp_net_dev_interrupt_handler,
+					   (void *)eth_dev);
+		/* Telling the firmware about the LSC interrupt entry */
+		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
+		/* Recording current stats counters values */
+		nfp_net_stats_reset(eth_dev);
+	}
+
+	return 0;
+
+dev_err_queues_map:
+		nfp_cpp_area_free(hw->hwqueues_area);
+dev_err_ctrl_map:
+		nfp_cpp_area_free(hw->ctrl_area);
+
+	return err;
+}
+
+static const struct rte_pci_id pci_id_nfp_vf_net_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP6000_VF_NIC)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static int nfp_vf_pci_uninit(struct rte_eth_dev *eth_dev)
+{
+	/* VF cleanup, just free private port data */
+	return nfp_netvf_close(eth_dev);
+}
+
+static int eth_nfp_vf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+	struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_probe(pci_dev,
+		sizeof(struct nfp_net_adapter), nfp_netvf_init);
+}
+
+static int eth_nfp_vf_pci_remove(struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_vf_pci_uninit);
+}
+
+static struct rte_pci_driver rte_nfp_net_vf_pmd = {
+	.id_table = pci_id_nfp_vf_net_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	.probe = eth_nfp_vf_pci_probe,
+	.remove = eth_nfp_vf_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_nfp_vf, rte_nfp_net_vf_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_vf, pci_id_nfp_vf_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_vf, "* igb_uio | uio_pci_generic | vfio");
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index da35bba4ef..a1722886ba 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -56,6 +56,7 @@ static int nfp_net_close(struct rte_eth_dev *dev);
 static int nfp_net_init(struct rte_eth_dev *eth_dev);
 static int nfp_pf_init(struct rte_pci_device *pci_dev);
 static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
+static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
 static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
@@ -283,18 +284,6 @@ nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
 	return 0;
 }
 
-static void
-nfp_net_vf_read_mac(struct nfp_net_hw *hw)
-{
-	uint32_t tmp;
-
-	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR));
-	memcpy(&hw->mac_addr[0], &tmp, 4);
-
-	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR + 4));
-	memcpy(&hw->mac_addr[4], &tmp, 2);
-}
-
 void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
@@ -1852,8 +1841,6 @@ nfp_net_init(struct rte_eth_dev *eth_dev)
 	if (hw->is_phyport) {
 		nfp_net_pf_read_mac(pf_dev, port);
 		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	} else {
-		nfp_net_vf_read_mac(hw);
 	}
 
 	if (!rte_is_valid_assigned_ether_addr(
@@ -2369,16 +2356,6 @@ static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
 	},
 };
 
-static const struct rte_pci_id pci_id_nfp_vf_net_map[] = {
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP6000_VF_NIC)
-	},
-	{
-		.vendor_id = 0,
-	},
-};
-
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
 {
 	struct rte_pci_device *pci_dev;
@@ -2402,13 +2379,6 @@ static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
 	return nfp_net_close(eth_dev);
 }
 
-static int eth_nfp_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
-	struct rte_pci_device *pci_dev)
-{
-	return rte_eth_dev_pci_generic_probe(pci_dev,
-		sizeof(struct nfp_net_adapter), nfp_net_init);
-}
-
 static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
 {
 	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
@@ -2421,19 +2391,9 @@ static struct rte_pci_driver rte_nfp_net_pf_pmd = {
 	.remove = eth_nfp_pci_remove,
 };
 
-static struct rte_pci_driver rte_nfp_net_vf_pmd = {
-	.id_table = pci_id_nfp_vf_net_map,
-	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
-	.probe = eth_nfp_pci_probe,
-	.remove = eth_nfp_pci_remove,
-};
-
 RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
-RTE_PMD_REGISTER_PCI(net_nfp_vf, rte_nfp_net_vf_pmd);
 RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
-RTE_PMD_REGISTER_PCI_TABLE(net_nfp_vf, pci_id_nfp_vf_net_map);
 RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
-RTE_PMD_REGISTER_KMOD_DEP(net_nfp_vf, "* igb_uio | uio_pci_generic | vfio");
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_init, init, NOTICE);
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_driver, driver, NOTICE);
 /*
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH 6/7] net/nfp: move PF functions into new file
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
                   ` (4 preceding siblings ...)
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
@ 2021-07-16  8:23 ` Heinrich Kuhn
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
  7 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

Similar to the last commit, this changeset moves all the PF specific
functions to a new file called nfp_ethdev.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
---
 drivers/net/nfp/meson.build  |    1 +
 drivers/net/nfp/nfp_ethdev.c | 1099 ++++++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_net.c    | 1088 +--------------------------------
 3 files changed, 1103 insertions(+), 1085 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_ethdev.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 34f4054b3c..ab64d0cac3 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -22,4 +22,5 @@ sources = files(
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
         'nfp_ethdev_vf.c',
+        'nfp_ethdev.c',
 )
diff --git a/drivers/net/nfp/nfp_ethdev.c b/drivers/net/nfp/nfp_ethdev.c
new file mode 100644
index 0000000000..ab08906704
--- /dev/null
+++ b/drivers/net/nfp/nfp_ethdev.c
@@ -0,0 +1,1099 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_ethdev.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: Main entry point
+ */
+
+#include <rte_common.h>
+#include <ethdev_driver.h>
+#include <ethdev_pci.h>
+#include <rte_dev.h>
+#include <rte_ether.h>
+#include <rte_malloc.h>
+#include <rte_memzone.h>
+#include <rte_mempool.h>
+#include <rte_service_component.h>
+
+#include "nfpcore/nfp_cpp.h"
+#include "nfpcore/nfp_nffw.h"
+#include "nfpcore/nfp_hwinfo.h"
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_rtsym.h"
+#include "nfpcore/nfp_nsp.h"
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+#include "nfp_cpp_bridge.h"
+
+
+static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
+static int nfp_net_start(struct rte_eth_dev *dev);
+static int nfp_net_stop(struct rte_eth_dev *dev);
+static int nfp_net_set_link_up(struct rte_eth_dev *dev);
+static int nfp_net_set_link_down(struct rte_eth_dev *dev);
+static int nfp_net_close(struct rte_eth_dev *dev);
+static int nfp_net_init(struct rte_eth_dev *eth_dev);
+static int nfp_fw_upload(struct rte_pci_device *dev,
+			 struct nfp_nsp *nsp, char *card);
+static int nfp_fw_setup(struct rte_pci_device *dev,
+			struct nfp_cpp *cpp,
+			struct nfp_eth_table *nfp_eth_table,
+			struct nfp_hwinfo *hwinfo);
+static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
+static int nfp_pf_init(struct rte_pci_device *pci_dev);
+static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
+static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+			    struct rte_pci_device *dev);
+static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
+static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev);
+
+static int
+nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
+{
+	struct nfp_eth_table *nfp_eth_table;
+	struct nfp_net_hw *hw = NULL;
+
+	/* Grab a pointer to the correct physical port */
+	hw = pf_dev->ports[port];
+
+	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
+
+	nfp_eth_copy_mac((uint8_t *)&hw->mac_addr,
+			 (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
+
+	free(nfp_eth_table);
+	return 0;
+}
+
+static int
+nfp_net_start(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+	uint32_t new_ctrl, update = 0;
+	struct nfp_net_hw *hw;
+	struct nfp_pf_dev *pf_dev;
+	struct rte_eth_conf *dev_conf;
+	struct rte_eth_rxmode *rxmode;
+	uint32_t intr_vector;
+	int ret;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+
+	PMD_INIT_LOG(DEBUG, "Start");
+
+	/* Disabling queues just in case... */
+	nfp_net_disable_queues(dev);
+
+	/* Enabling the required queues in the device */
+	nfp_net_enable_queues(dev);
+
+	/* check and configure queue intr-vector mapping */
+	if (dev->data->dev_conf.intr_conf.rxq != 0) {
+		if (pf_dev->multiport) {
+			PMD_INIT_LOG(ERR, "PMD rx interrupt is not supported "
+					  "with NFP multiport PF");
+				return -EINVAL;
+		}
+		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+			/*
+			 * Better not to share LSC with RX interrupts.
+			 * Unregistering LSC interrupt handler
+			 */
+			rte_intr_callback_unregister(&pci_dev->intr_handle,
+				nfp_net_dev_interrupt_handler, (void *)dev);
+
+			if (dev->data->nb_rx_queues > 1) {
+				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+					     "supports 1 queue with UIO");
+				return -EIO;
+			}
+		}
+		intr_vector = dev->data->nb_rx_queues;
+		if (rte_intr_efd_enable(intr_handle, intr_vector))
+			return -1;
+
+		nfp_configure_rx_interrupt(dev, intr_handle);
+		update = NFP_NET_CFG_UPDATE_MSIX;
+	}
+
+	rte_intr_enable(intr_handle);
+
+	new_ctrl = nfp_check_offloads(dev);
+
+	/* Writing configuration parameters in the device */
+	nfp_net_params_setup(hw);
+
+	dev_conf = &dev->data->dev_conf;
+	rxmode = &dev_conf->rxmode;
+
+	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
+		nfp_net_rss_config_default(dev);
+		update |= NFP_NET_CFG_UPDATE_RSS;
+		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
+	}
+
+	/* Enable device */
+	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
+
+	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+
+	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
+		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
+
+	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
+	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+		return -EIO;
+
+	/*
+	 * Allocating rte mbufs for configured rx queues.
+	 * This requires queues being enabled before
+	 */
+	if (nfp_net_rx_freelist_setup(dev) < 0) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port up */
+		nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
+	else
+		nfp_eth_set_configured(dev->process_private,
+				       hw->nfp_idx, 1);
+
+	hw->ctrl = new_ctrl;
+
+	return 0;
+
+error:
+	/*
+	 * An error returned by this function should mean the app
+	 * exiting and then the system releasing all the memory
+	 * allocated even memory coming from hugepages.
+	 *
+	 * The device could be enabled at this point with some queues
+	 * ready for getting packets. This is true if the call to
+	 * nfp_net_rx_freelist_setup() succeeds for some queues but
+	 * fails for subsequent queues.
+	 *
+	 * This should make the app exiting but better if we tell the
+	 * device first.
+	 */
+	nfp_net_disable_queues(dev);
+
+	return ret;
+}
+
+/* Stop device: disable rx and tx functions to allow for reconfiguring. */
+static int
+nfp_net_stop(struct rte_eth_dev *dev)
+{
+	int i;
+	struct nfp_net_hw *hw;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+
+	PMD_INIT_LOG(DEBUG, "Stop");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
+	else
+		nfp_eth_set_configured(dev->process_private,
+				       hw->nfp_idx, 0);
+
+	return 0;
+}
+
+/* Set the link up. */
+static int
+nfp_net_set_link_up(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+
+	PMD_DRV_LOG(DEBUG, "Set link up");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
+	else
+		return nfp_eth_set_configured(dev->process_private,
+					      hw->nfp_idx, 1);
+}
+
+/* Set the link down. */
+static int
+nfp_net_set_link_down(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+
+	PMD_DRV_LOG(DEBUG, "Set link down");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
+	else
+		return nfp_eth_set_configured(dev->process_private,
+					      hw->nfp_idx, 0);
+}
+
+/* Reset and stop device. The device can not be restarted. */
+static int
+nfp_net_close(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+	struct rte_pci_device *pci_dev;
+	struct nfp_pf_dev *pf_dev;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	PMD_INIT_LOG(DEBUG, "Close");
+
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+	/*
+	 * We assume that the DPDK application is stopping all the
+	 * threads/queues before calling the device close function.
+	 */
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	/* Only free PF resources after all physical ports have been closed */
+	/* Mark this port as unused and free device priv resources*/
+	nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
+	pf_dev->ports[hw->idx] = NULL;
+	rte_eth_dev_release_port(dev);
+
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		/* Check to see if ports are still in use */
+		if (pf_dev->ports[i])
+			return 0;
+	}
+
+	/* Now it is safe to free all PF resources */
+	PMD_INIT_LOG(INFO, "Freeing PF resources");
+	nfp_cpp_area_free(pf_dev->ctrl_area);
+	nfp_cpp_area_free(pf_dev->hwqueues_area);
+	free(pf_dev->hwinfo);
+	free(pf_dev->sym_tbl);
+	nfp_cpp_free(pf_dev->cpp);
+	rte_free(pf_dev);
+
+	rte_intr_disable(&pci_dev->intr_handle);
+
+	/* unregister callback func from eal lib */
+	rte_intr_callback_unregister(&pci_dev->intr_handle,
+				     nfp_net_dev_interrupt_handler,
+				     (void *)dev);
+
+	/*
+	 * The ixgbe PMD driver disables the pcie master on the
+	 * device. The i40e does not...
+	 */
+
+	return 0;
+}
+
+/* Initialise and register driver with DPDK Application */
+static const struct eth_dev_ops nfp_net_eth_dev_ops = {
+	.dev_configure		= nfp_net_configure,
+	.dev_start		= nfp_net_start,
+	.dev_stop		= nfp_net_stop,
+	.dev_set_link_up	= nfp_net_set_link_up,
+	.dev_set_link_down	= nfp_net_set_link_down,
+	.dev_close		= nfp_net_close,
+	.promiscuous_enable	= nfp_net_promisc_enable,
+	.promiscuous_disable	= nfp_net_promisc_disable,
+	.link_update		= nfp_net_link_update,
+	.stats_get		= nfp_net_stats_get,
+	.stats_reset		= nfp_net_stats_reset,
+	.dev_infos_get		= nfp_net_infos_get,
+	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
+	.mtu_set		= nfp_net_dev_mtu_set,
+	.mac_addr_set           = nfp_set_mac_addr,
+	.vlan_offload_set	= nfp_net_vlan_offload_set,
+	.reta_update		= nfp_net_reta_update,
+	.reta_query		= nfp_net_reta_query,
+	.rss_hash_update	= nfp_net_rss_hash_update,
+	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
+	.rx_queue_setup		= nfp_net_rx_queue_setup,
+	.rx_queue_release	= nfp_net_rx_queue_release,
+	.tx_queue_setup		= nfp_net_tx_queue_setup,
+	.tx_queue_release	= nfp_net_tx_queue_release,
+	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
+	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
+};
+
+static int
+nfp_net_init(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_pf_dev *pf_dev;
+	struct nfp_net_hw *hw;
+	struct rte_ether_addr *tmp_ether_addr;
+
+	uint64_t tx_bar_off = 0, rx_bar_off = 0;
+	uint32_t start_q;
+	int stride = 4;
+	int port = 0;
+	int err;
+
+	PMD_INIT_FUNC_TRACE();
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* Use backpointer here to the PF of this eth_dev */
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(eth_dev->data->dev_private);
+
+	/* NFP can not handle DMA addresses requiring more than 40 bits */
+	if (rte_mem_check_dma_mask(40)) {
+		RTE_LOG(ERR, PMD, "device %s can not be used:",
+				   pci_dev->device.name);
+		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
+		return -ENODEV;
+	};
+
+	port = ((struct nfp_net_hw *)eth_dev->data->dev_private)->idx;
+	if (port < 0 || port > 7) {
+		PMD_DRV_LOG(ERR, "Port value is wrong");
+		return -ENODEV;
+	}
+
+	/* Use PF array of physical ports to get pointer to
+	 * this specific port
+	 */
+	hw = pf_dev->ports[port];
+
+	PMD_INIT_LOG(DEBUG, "Working with physical port number: %d, "
+			    "NFP internal port number: %d",
+			    port, hw->nfp_idx);
+
+	eth_dev->dev_ops = &nfp_net_eth_dev_ops;
+	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	rte_eth_copy_pci_info(eth_dev, pci_dev);
+
+	hw->device_id = pci_dev->id.device_id;
+	hw->vendor_id = pci_dev->id.vendor_id;
+	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
+	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
+
+	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
+		     pci_dev->id.vendor_id, pci_dev->id.device_id,
+		     pci_dev->addr.domain, pci_dev->addr.bus,
+		     pci_dev->addr.devid, pci_dev->addr.function);
+
+	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
+	if (hw->ctrl_bar == NULL) {
+		PMD_DRV_LOG(ERR,
+			"hw->ctrl_bar is NULL. BAR0 not configured");
+		return -ENODEV;
+	}
+
+	if (port == 0) {
+		hw->ctrl_bar = pf_dev->ctrl_bar;
+	} else {
+		if (!pf_dev->ctrl_bar)
+			return -ENODEV;
+		/* Use port offset in pf ctrl_bar for this
+		 * ports control bar
+		 */
+		hw->ctrl_bar = pf_dev->ctrl_bar +
+			       (port * NFP_PF_CSR_SLICE_SIZE);
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+
+	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
+	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
+
+	/* Work out where in the BAR the queues start. */
+	switch (pci_dev->id.device_id) {
+	case PCI_DEVICE_ID_NFP4000_PF_NIC:
+	case PCI_DEVICE_ID_NFP6000_PF_NIC:
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
+		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
+		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
+		err = -ENODEV;
+		goto dev_err_ctrl_map;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
+	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
+
+	hw->tx_bar = pf_dev->hw_queues + tx_bar_off;
+	hw->rx_bar = pf_dev->hw_queues + rx_bar_off;
+	eth_dev->data->dev_private = hw;
+
+	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
+		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
+
+	nfp_net_cfg_queue_setup(hw);
+
+	/* Get some of the read-only fields from the config BAR */
+	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
+	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
+	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
+	hw->mtu = RTE_ETHER_MTU;
+
+	/* VLAN insertion is incompatible with LSOv2 */
+	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
+		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
+
+	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
+		hw->rx_offset = NFP_NET_RX_OFFSET;
+	else
+		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
+
+	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
+			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
+			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
+
+	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
+		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
+
+	hw->ctrl = 0;
+
+	hw->stride_rx = stride;
+	hw->stride_tx = stride;
+
+	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
+		     hw->max_rx_queues, hw->max_tx_queues);
+
+	/* Initializing spinlock for reconfigs */
+	rte_spinlock_init(&hw->reconfig_lock);
+
+	/* Allocating memory for mac addr */
+	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
+					       RTE_ETHER_ADDR_LEN, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
+		err = -ENOMEM;
+		goto dev_err_queues_map;
+	}
+
+	nfp_net_pf_read_mac(pf_dev, port);
+	nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+
+	tmp_ether_addr = (struct rte_ether_addr *)&hw->mac_addr;
+	if (!rte_is_valid_assigned_ether_addr(tmp_ether_addr)) {
+		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
+				   port);
+		/* Using random mac addresses for VFs */
+		rte_eth_random_addr(&hw->mac_addr[0]);
+		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+	}
+
+	/* Copying mac address to DPDK eth_dev struct */
+	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
+			&eth_dev->data->mac_addrs[0]);
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
+
+	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
+
+	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
+		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
+		     eth_dev->data->port_id, pci_dev->id.vendor_id,
+		     pci_dev->id.device_id,
+		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
+		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		/* Registering LSC interrupt handler */
+		rte_intr_callback_register(&pci_dev->intr_handle,
+					   nfp_net_dev_interrupt_handler,
+					   (void *)eth_dev);
+		/* Telling the firmware about the LSC interrupt entry */
+		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
+		/* Recording current stats counters values */
+		nfp_net_stats_reset(eth_dev);
+	}
+
+	return 0;
+
+dev_err_queues_map:
+		nfp_cpp_area_free(hw->hwqueues_area);
+dev_err_ctrl_map:
+		nfp_cpp_area_free(hw->ctrl_area);
+
+	return err;
+}
+
+#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
+
+static int
+nfp_fw_upload(struct rte_pci_device *dev, struct nfp_nsp *nsp, char *card)
+{
+	struct nfp_cpp *cpp = nsp->cpp;
+	int fw_f;
+	char *fw_buf;
+	char fw_name[125];
+	char serial[40];
+	struct stat file_stat;
+	off_t fsize, bytes;
+
+	/* Looking for firmware file in order of priority */
+
+	/* First try to find a firmware image specific for this device */
+	snprintf(serial, sizeof(serial),
+			"serial-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x",
+		cpp->serial[0], cpp->serial[1], cpp->serial[2], cpp->serial[3],
+		cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
+		cpp->interface & 0xff);
+
+	snprintf(fw_name, sizeof(fw_name), "%s/%s.nffw", DEFAULT_FW_PATH,
+			serial);
+
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f >= 0)
+		goto read_fw;
+
+	/* Then try the PCI name */
+	snprintf(fw_name, sizeof(fw_name), "%s/pci-%s.nffw", DEFAULT_FW_PATH,
+			dev->device.name);
+
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f >= 0)
+		goto read_fw;
+
+	/* Finally try the card type and media */
+	snprintf(fw_name, sizeof(fw_name), "%s/%s", DEFAULT_FW_PATH, card);
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f < 0) {
+		PMD_DRV_LOG(INFO, "Firmware file %s not found.", fw_name);
+		return -ENOENT;
+	}
+
+read_fw:
+	if (fstat(fw_f, &file_stat) < 0) {
+		PMD_DRV_LOG(INFO, "Firmware file %s size is unknown", fw_name);
+		close(fw_f);
+		return -ENOENT;
+	}
+
+	fsize = file_stat.st_size;
+	PMD_DRV_LOG(INFO, "Firmware file found at %s with size: %" PRIu64 "",
+			    fw_name, (uint64_t)fsize);
+
+	fw_buf = malloc((size_t)fsize);
+	if (!fw_buf) {
+		PMD_DRV_LOG(INFO, "malloc failed for fw buffer");
+		close(fw_f);
+		return -ENOMEM;
+	}
+	memset(fw_buf, 0, fsize);
+
+	bytes = read(fw_f, fw_buf, fsize);
+	if (bytes != fsize) {
+		PMD_DRV_LOG(INFO, "Reading fw to buffer failed."
+				   "Just %" PRIu64 " of %" PRIu64 " bytes read",
+				   (uint64_t)bytes, (uint64_t)fsize);
+		free(fw_buf);
+		close(fw_f);
+		return -EIO;
+	}
+
+	PMD_DRV_LOG(INFO, "Uploading the firmware ...");
+	nfp_nsp_load_fw(nsp, fw_buf, bytes);
+	PMD_DRV_LOG(INFO, "Done");
+
+	free(fw_buf);
+	close(fw_f);
+
+	return 0;
+}
+
+static int
+nfp_fw_setup(struct rte_pci_device *dev, struct nfp_cpp *cpp,
+	     struct nfp_eth_table *nfp_eth_table, struct nfp_hwinfo *hwinfo)
+{
+	struct nfp_nsp *nsp;
+	const char *nfp_fw_model;
+	char card_desc[100];
+	int err = 0;
+
+	nfp_fw_model = nfp_hwinfo_lookup(hwinfo, "assembly.partno");
+
+	if (nfp_fw_model) {
+		PMD_DRV_LOG(INFO, "firmware model found: %s", nfp_fw_model);
+	} else {
+		PMD_DRV_LOG(ERR, "firmware model NOT found");
+		return -EIO;
+	}
+
+	if (nfp_eth_table->count == 0 || nfp_eth_table->count > 8) {
+		PMD_DRV_LOG(ERR, "NFP ethernet table reports wrong ports: %u",
+		       nfp_eth_table->count);
+		return -EIO;
+	}
+
+	PMD_DRV_LOG(INFO, "NFP ethernet port table reports %u ports",
+			   nfp_eth_table->count);
+
+	PMD_DRV_LOG(INFO, "Port speed: %u", nfp_eth_table->ports[0].speed);
+
+	snprintf(card_desc, sizeof(card_desc), "nic_%s_%dx%d.nffw",
+			nfp_fw_model, nfp_eth_table->count,
+			nfp_eth_table->ports[0].speed / 1000);
+
+	nsp = nfp_nsp_open(cpp);
+	if (!nsp) {
+		PMD_DRV_LOG(ERR, "NFP error when obtaining NSP handle");
+		return -EIO;
+	}
+
+	nfp_nsp_device_soft_reset(nsp);
+	err = nfp_fw_upload(dev, nsp, card_desc);
+
+	nfp_nsp_close(nsp);
+	return err;
+}
+
+static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
+{
+	struct nfp_net_hw *hw;
+	struct rte_eth_dev *eth_dev;
+	struct nfp_eth_table *nfp_eth_table = NULL;
+	int ret = 0;
+	int i;
+
+	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
+	if (!nfp_eth_table) {
+		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
+		ret = -EIO;
+		goto error;
+	}
+
+	/* Loop through all physical ports on PF */
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		const unsigned int numa_node = rte_socket_id();
+		char port_name[RTE_ETH_NAME_MAX_LEN];
+
+		snprintf(port_name, sizeof(port_name), "%s_port%d",
+			 pf_dev->pci_dev->device.name, i);
+
+		/* Allocate a eth_dev for this phyport */
+		eth_dev = rte_eth_dev_allocate(port_name);
+		if (!eth_dev) {
+			ret = -ENODEV;
+			goto port_cleanup;
+		}
+
+		/* Allocate memory for this phyport */
+		eth_dev->data->dev_private =
+			rte_zmalloc_socket(port_name, sizeof(struct nfp_net_hw),
+					   RTE_CACHE_LINE_SIZE, numa_node);
+		if (!eth_dev->data->dev_private) {
+			ret = -ENOMEM;
+			rte_eth_dev_release_port(eth_dev);
+			goto port_cleanup;
+		}
+
+		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+
+		/* Add this device to the PF's array of physical ports */
+		pf_dev->ports[i] = hw;
+
+		hw->pf_dev = pf_dev;
+		hw->cpp = pf_dev->cpp;
+		hw->eth_dev = eth_dev;
+		hw->idx = i;
+		hw->nfp_idx = nfp_eth_table->ports[i].index;
+		hw->is_phyport = true;
+
+		eth_dev->device = &pf_dev->pci_dev->device;
+
+		/* ctrl/tx/rx BAR mappings and remaining init happens in
+		 * nfp_net_init
+		 */
+		ret = nfp_net_init(eth_dev);
+
+		if (ret) {
+			ret = -ENODEV;
+			goto port_cleanup;
+		}
+
+		rte_eth_dev_probing_finish(eth_dev);
+
+	} /* End loop, all ports on this PF */
+	ret = 0;
+	goto eth_table_cleanup;
+
+port_cleanup:
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		if (pf_dev->ports[i] && pf_dev->ports[i]->eth_dev) {
+			struct rte_eth_dev *tmp_dev;
+			tmp_dev = pf_dev->ports[i]->eth_dev;
+			rte_eth_dev_release_port(tmp_dev);
+			pf_dev->ports[i] = NULL;
+		}
+	}
+eth_table_cleanup:
+	free(nfp_eth_table);
+error:
+	return ret;
+}
+
+static int nfp_pf_init(struct rte_pci_device *pci_dev)
+{
+	struct nfp_pf_dev *pf_dev = NULL;
+	struct nfp_cpp *cpp;
+	struct nfp_hwinfo *hwinfo;
+	struct nfp_rtsym_table *sym_tbl;
+	struct nfp_eth_table *nfp_eth_table = NULL;
+	char name[RTE_ETH_NAME_MAX_LEN];
+	int total_ports;
+	int ret = -ENODEV;
+	int err;
+
+	if (!pci_dev)
+		return ret;
+
+	/*
+	 * When device bound to UIO, the device could be used, by mistake,
+	 * by two DPDK apps, and the UIO driver does not avoid it. This
+	 * could lead to a serious problem when configuring the NFP CPP
+	 * interface. Here we avoid this telling to the CPP init code to
+	 * use a lock file if UIO is being used.
+	 */
+	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
+		cpp = nfp_cpp_from_device_name(pci_dev, 0);
+	else
+		cpp = nfp_cpp_from_device_name(pci_dev, 1);
+
+	if (!cpp) {
+		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
+		ret = -EIO;
+		goto error;
+	}
+
+	hwinfo = nfp_hwinfo_read(cpp);
+	if (!hwinfo) {
+		PMD_INIT_LOG(ERR, "Error reading hwinfo table");
+		ret = -EIO;
+		goto error;
+	}
+
+	nfp_eth_table = nfp_eth_read_ports(cpp);
+	if (!nfp_eth_table) {
+		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
+		ret = -EIO;
+		goto hwinfo_cleanup;
+	}
+
+	if (nfp_fw_setup(pci_dev, cpp, nfp_eth_table, hwinfo)) {
+		PMD_INIT_LOG(ERR, "Error when uploading firmware");
+		ret = -EIO;
+		goto eth_table_cleanup;
+	}
+
+	/* Now the symbol table should be there */
+	sym_tbl = nfp_rtsym_table_read(cpp);
+	if (!sym_tbl) {
+		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
+				" symbol table");
+		ret = -EIO;
+		goto eth_table_cleanup;
+	}
+
+	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
+	if (total_ports != (int)nfp_eth_table->count) {
+		PMD_DRV_LOG(ERR, "Inconsistent number of ports");
+		ret = -EIO;
+		goto sym_tbl_cleanup;
+	}
+
+	PMD_INIT_LOG(INFO, "Total physical ports: %d", total_ports);
+
+	if (total_ports <= 0 || total_ports > 8) {
+		PMD_INIT_LOG(ERR, "nfd_cfg_pf0_num_ports symbol with wrong value");
+		ret = -ENODEV;
+		goto sym_tbl_cleanup;
+	}
+	/* Allocate memory for the PF "device" */
+	snprintf(name, sizeof(name), "nfp_pf%d", 0);
+	pf_dev = rte_zmalloc(name, sizeof(*pf_dev), 0);
+	if (!pf_dev) {
+		ret = -ENOMEM;
+		goto sym_tbl_cleanup;
+	}
+
+	/* Populate the newly created PF device */
+	pf_dev->cpp = cpp;
+	pf_dev->hwinfo = hwinfo;
+	pf_dev->sym_tbl = sym_tbl;
+	pf_dev->total_phyports = total_ports;
+
+	if (total_ports > 1)
+		pf_dev->multiport = true;
+
+	pf_dev->pci_dev = pci_dev;
+
+	/* Map the symbol table */
+	pf_dev->ctrl_bar = nfp_rtsym_map(pf_dev->sym_tbl, "_pf0_net_bar0",
+				     pf_dev->total_phyports * 32768,
+				     &pf_dev->ctrl_area);
+	if (!pf_dev->ctrl_bar) {
+		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for _pf0_net_ctrl_bar");
+		ret = -EIO;
+		goto pf_cleanup;
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", pf_dev->ctrl_bar);
+
+	/* configure access to tx/rx vNIC BARs */
+	pf_dev->hw_queues = nfp_cpp_map_area(pf_dev->cpp, 0, 0,
+					      NFP_PCIE_QUEUE(0),
+					      NFP_QCP_QUEUE_AREA_SZ,
+					      &pf_dev->hwqueues_area);
+	if (!pf_dev->hw_queues) {
+		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for net.qc");
+		ret = -EIO;
+		goto ctrl_area_cleanup;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p", pf_dev->hw_queues);
+
+	/* Initialize and prep physical ports now
+	 * This will loop through all physical ports
+	 */
+	ret = nfp_init_phyports(pf_dev);
+	if (ret) {
+		PMD_INIT_LOG(ERR, "Could not create physical ports");
+		goto hwqueues_cleanup;
+	}
+
+	/* register the CPP bridge service here for primary use */
+	nfp_register_cpp_service(pf_dev->cpp);
+
+	return 0;
+
+hwqueues_cleanup:
+	nfp_cpp_area_free(pf_dev->hwqueues_area);
+ctrl_area_cleanup:
+	nfp_cpp_area_free(pf_dev->ctrl_area);
+pf_cleanup:
+	rte_free(pf_dev);
+sym_tbl_cleanup:
+	free(sym_tbl);
+eth_table_cleanup:
+	free(nfp_eth_table);
+hwinfo_cleanup:
+	free(hwinfo);
+error:
+	return ret;
+}
+
+/*
+ * When attaching to the NFP4000/6000 PF on a secondary process there
+ * is no need to initilise the PF again. Only minimal work is required
+ * here
+ */
+static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev)
+{
+	struct nfp_cpp *cpp;
+	struct nfp_rtsym_table *sym_tbl;
+	int total_ports;
+	int i;
+	int err;
+
+	if (!pci_dev)
+		return -ENODEV;
+
+	/*
+	 * When device bound to UIO, the device could be used, by mistake,
+	 * by two DPDK apps, and the UIO driver does not avoid it. This
+	 * could lead to a serious problem when configuring the NFP CPP
+	 * interface. Here we avoid this telling to the CPP init code to
+	 * use a lock file if UIO is being used.
+	 */
+	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
+		cpp = nfp_cpp_from_device_name(pci_dev, 0);
+	else
+		cpp = nfp_cpp_from_device_name(pci_dev, 1);
+
+	if (!cpp) {
+		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
+		return -EIO;
+	}
+
+	/*
+	 * We don't have access to the PF created in the primary process
+	 * here so we have to read the number of ports from firmware
+	 */
+	sym_tbl = nfp_rtsym_table_read(cpp);
+	if (!sym_tbl) {
+		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
+				" symbol table");
+		return -EIO;
+	}
+
+	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
+
+	for (i = 0; i < total_ports; i++) {
+		struct rte_eth_dev *eth_dev;
+		char port_name[RTE_ETH_NAME_MAX_LEN];
+
+		snprintf(port_name, sizeof(port_name), "%s_port%d",
+			 pci_dev->device.name, i);
+
+		PMD_DRV_LOG(DEBUG, "Secondary attaching to port %s",
+		    port_name);
+		eth_dev = rte_eth_dev_attach_secondary(port_name);
+		if (!eth_dev) {
+			RTE_LOG(ERR, EAL,
+			"secondary process attach failed, "
+			"ethdev doesn't exist");
+			return -ENODEV;
+		}
+		eth_dev->process_private = cpp;
+		eth_dev->dev_ops = &nfp_net_eth_dev_ops;
+		eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+		eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+		eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+		rte_eth_dev_probing_finish(eth_dev);
+	}
+
+	/* Register the CPP bridge service for the secondary too */
+	nfp_register_cpp_service(cpp);
+
+	return 0;
+}
+
+static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+			    struct rte_pci_device *dev)
+{
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		return nfp_pf_init(dev);
+	else
+		return nfp_pf_secondary_init(dev);
+}
+
+static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP4000_PF_NIC)
+	},
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP6000_PF_NIC)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	uint16_t port_id;
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* Free up all physical ports under PF */
+	RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
+		rte_eth_dev_close(port_id);
+	/*
+	 * Ports can be closed and freed but hotplugging is not
+	 * currently supported
+	 */
+	return -ENOTSUP;
+}
+
+static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
+}
+
+static struct rte_pci_driver rte_nfp_net_pf_pmd = {
+	.id_table = pci_id_nfp_pf_net_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	.probe = nfp_pf_pci_probe,
+	.remove = eth_nfp_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index a1722886ba..a6097eaab0 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -37,10 +37,10 @@
 #include "nfpcore/nfp_rtsym.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 #include "nfp_cpp_bridge.h"
 
 #include <sys/types.h>
@@ -51,20 +51,6 @@
 #include <sys/ioctl.h>
 #include <errno.h>
 
-/* Prototypes */
-static int nfp_net_close(struct rte_eth_dev *dev);
-static int nfp_net_init(struct rte_eth_dev *eth_dev);
-static int nfp_pf_init(struct rte_pci_device *pci_dev);
-static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
-static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
-static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
-static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
-static int nfp_net_stop(struct rte_eth_dev *dev);
-static int nfp_fw_setup(struct rte_pci_device *dev,
-			struct nfp_cpp *cpp,
-			struct nfp_eth_table *nfp_eth_table,
-			struct nfp_hwinfo *hwinfo);
-
 static int
 __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
 {
@@ -266,24 +252,6 @@ nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
 		dst[i] = src[i];
 }
 
-static int
-nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
-{
-	struct nfp_eth_table *nfp_eth_table;
-	struct nfp_net_hw *hw = NULL;
-
-	/* Grab a pointer to the correct physical port */
-	hw = pf_dev->ports[port];
-
-	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
-
-	nfp_eth_copy_mac((uint8_t *)&hw->mac_addr,
-			 (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
-
-	free(nfp_eth_table);
-	return 0;
-}
-
 void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
@@ -436,282 +404,6 @@ nfp_check_offloads(struct rte_eth_dev *dev)
 	return ctrl;
 }
 
-static int
-nfp_net_start(struct rte_eth_dev *dev)
-{
-	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
-	uint32_t new_ctrl, update = 0;
-	struct nfp_net_hw *hw;
-	struct nfp_pf_dev *pf_dev;
-	struct rte_eth_conf *dev_conf;
-	struct rte_eth_rxmode *rxmode;
-	uint32_t intr_vector;
-	int ret;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
-
-	PMD_INIT_LOG(DEBUG, "Start");
-
-	/* Disabling queues just in case... */
-	nfp_net_disable_queues(dev);
-
-	/* Enabling the required queues in the device */
-	nfp_net_enable_queues(dev);
-
-	/* check and configure queue intr-vector mapping */
-	if (dev->data->dev_conf.intr_conf.rxq != 0) {
-		if (pf_dev->multiport) {
-			PMD_INIT_LOG(ERR, "PMD rx interrupt is not supported "
-					  "with NFP multiport PF");
-				return -EINVAL;
-		}
-		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
-			/*
-			 * Better not to share LSC with RX interrupts.
-			 * Unregistering LSC interrupt handler
-			 */
-			rte_intr_callback_unregister(&pci_dev->intr_handle,
-				nfp_net_dev_interrupt_handler, (void *)dev);
-
-			if (dev->data->nb_rx_queues > 1) {
-				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
-					     "supports 1 queue with UIO");
-				return -EIO;
-			}
-		}
-		intr_vector = dev->data->nb_rx_queues;
-		if (rte_intr_efd_enable(intr_handle, intr_vector))
-			return -1;
-
-		nfp_configure_rx_interrupt(dev, intr_handle);
-		update = NFP_NET_CFG_UPDATE_MSIX;
-	}
-
-	rte_intr_enable(intr_handle);
-
-	new_ctrl = nfp_check_offloads(dev);
-
-	/* Writing configuration parameters in the device */
-	nfp_net_params_setup(hw);
-
-	dev_conf = &dev->data->dev_conf;
-	rxmode = &dev_conf->rxmode;
-
-	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
-		nfp_net_rss_config_default(dev);
-		update |= NFP_NET_CFG_UPDATE_RSS;
-		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
-	}
-
-	/* Enable device */
-	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
-
-	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
-
-	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
-		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
-
-	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
-	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
-		return -EIO;
-
-	/*
-	 * Allocating rte mbufs for configured rx queues.
-	 * This requires queues being enabled before
-	 */
-	if (nfp_net_rx_freelist_setup(dev) < 0) {
-		ret = -ENOMEM;
-		goto error;
-	}
-
-	if (hw->is_phyport) {
-		if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-			/* Configure the physical port up */
-			nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
-		else
-			nfp_eth_set_configured(dev->process_private,
-					       hw->nfp_idx, 1);
-	}
-
-	hw->ctrl = new_ctrl;
-
-	return 0;
-
-error:
-	/*
-	 * An error returned by this function should mean the app
-	 * exiting and then the system releasing all the memory
-	 * allocated even memory coming from hugepages.
-	 *
-	 * The device could be enabled at this point with some queues
-	 * ready for getting packets. This is true if the call to
-	 * nfp_net_rx_freelist_setup() succeeds for some queues but
-	 * fails for subsequent queues.
-	 *
-	 * This should make the app exiting but better if we tell the
-	 * device first.
-	 */
-	nfp_net_disable_queues(dev);
-
-	return ret;
-}
-
-/* Stop device: disable rx and tx functions to allow for reconfiguring. */
-static int
-nfp_net_stop(struct rte_eth_dev *dev)
-{
-	int i;
-	struct nfp_net_hw *hw;
-
-	PMD_INIT_LOG(DEBUG, "Stop");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	nfp_net_disable_queues(dev);
-
-	/* Clear queues */
-	for (i = 0; i < dev->data->nb_tx_queues; i++) {
-		nfp_net_reset_tx_queue(
-			(struct nfp_net_txq *)dev->data->tx_queues[i]);
-	}
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		nfp_net_reset_rx_queue(
-			(struct nfp_net_rxq *)dev->data->rx_queues[i]);
-	}
-
-	if (hw->is_phyport) {
-		if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-			/* Configure the physical port down */
-			nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
-		else
-			nfp_eth_set_configured(dev->process_private,
-					       hw->nfp_idx, 0);
-	}
-
-	return 0;
-}
-
-/* Set the link up. */
-static int
-nfp_net_set_link_up(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-
-	PMD_DRV_LOG(DEBUG, "Set link up");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	if (!hw->is_phyport)
-		return -ENOTSUP;
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		/* Configure the physical port down */
-		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
-	else
-		return nfp_eth_set_configured(dev->process_private,
-					      hw->nfp_idx, 1);
-}
-
-/* Set the link down. */
-static int
-nfp_net_set_link_down(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-
-	PMD_DRV_LOG(DEBUG, "Set link down");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	if (!hw->is_phyport)
-		return -ENOTSUP;
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		/* Configure the physical port down */
-		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
-	else
-		return nfp_eth_set_configured(dev->process_private,
-					      hw->nfp_idx, 0);
-}
-
-/* Reset and stop device. The device can not be restarted. */
-static int
-nfp_net_close(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-	struct rte_pci_device *pci_dev;
-	int i;
-
-	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
-		return 0;
-
-	PMD_INIT_LOG(DEBUG, "Close");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-
-	/*
-	 * We assume that the DPDK application is stopping all the
-	 * threads/queues before calling the device close function.
-	 */
-
-	nfp_net_disable_queues(dev);
-
-	/* Clear queues */
-	for (i = 0; i < dev->data->nb_tx_queues; i++) {
-		nfp_net_reset_tx_queue(
-			(struct nfp_net_txq *)dev->data->tx_queues[i]);
-	}
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		nfp_net_reset_rx_queue(
-			(struct nfp_net_rxq *)dev->data->rx_queues[i]);
-	}
-
-	/* Only free PF resources after all physical ports have been closed */
-	if (pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC ||
-	    pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC) {
-		struct nfp_pf_dev *pf_dev;
-		pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
-
-		/* Mark this port as unused and free device priv resources*/
-		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
-		pf_dev->ports[hw->idx] = NULL;
-		rte_eth_dev_release_port(dev);
-
-		for (i = 0; i < pf_dev->total_phyports; i++) {
-			/* Check to see if ports are still in use */
-			if (pf_dev->ports[i])
-				return 0;
-		}
-
-		/* Now it is safe to free all PF resources */
-		PMD_INIT_LOG(INFO, "Freeing PF resources");
-		nfp_cpp_area_free(pf_dev->ctrl_area);
-		nfp_cpp_area_free(pf_dev->hwqueues_area);
-		free(pf_dev->hwinfo);
-		free(pf_dev->sym_tbl);
-		nfp_cpp_free(pf_dev->cpp);
-		rte_free(pf_dev);
-	}
-
-	rte_intr_disable(&pci_dev->intr_handle);
-
-	/* unregister callback func from eal lib */
-	rte_intr_callback_unregister(&pci_dev->intr_handle,
-				     nfp_net_dev_interrupt_handler,
-				     (void *)dev);
-
-	/*
-	 * The ixgbe PMD driver disables the pcie master on the
-	 * device. The i40e does not...
-	 */
-
-	return 0;
-}
-
 int
 nfp_net_promisc_enable(struct rte_eth_dev *dev)
 {
@@ -1620,780 +1312,6 @@ nfp_net_rss_config_default(struct rte_eth_dev *dev)
 	return ret;
 }
 
-
-/* Initialise and register driver with DPDK Application */
-static const struct eth_dev_ops nfp_net_eth_dev_ops = {
-	.dev_configure		= nfp_net_configure,
-	.dev_start		= nfp_net_start,
-	.dev_stop		= nfp_net_stop,
-	.dev_set_link_up	= nfp_net_set_link_up,
-	.dev_set_link_down	= nfp_net_set_link_down,
-	.dev_close		= nfp_net_close,
-	.promiscuous_enable	= nfp_net_promisc_enable,
-	.promiscuous_disable	= nfp_net_promisc_disable,
-	.link_update		= nfp_net_link_update,
-	.stats_get		= nfp_net_stats_get,
-	.stats_reset		= nfp_net_stats_reset,
-	.dev_infos_get		= nfp_net_infos_get,
-	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
-	.mtu_set		= nfp_net_dev_mtu_set,
-	.mac_addr_set           = nfp_set_mac_addr,
-	.vlan_offload_set	= nfp_net_vlan_offload_set,
-	.reta_update		= nfp_net_reta_update,
-	.reta_query		= nfp_net_reta_query,
-	.rss_hash_update	= nfp_net_rss_hash_update,
-	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
-	.rx_queue_setup		= nfp_net_rx_queue_setup,
-	.rx_queue_release	= nfp_net_rx_queue_release,
-	.tx_queue_setup		= nfp_net_tx_queue_setup,
-	.tx_queue_release	= nfp_net_tx_queue_release,
-	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
-	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
-};
-
-
-static int
-nfp_net_init(struct rte_eth_dev *eth_dev)
-{
-	struct rte_pci_device *pci_dev;
-	struct nfp_pf_dev *pf_dev;
-	struct nfp_net_hw *hw;
-
-	uint64_t tx_bar_off = 0, rx_bar_off = 0;
-	uint32_t start_q;
-	int stride = 4;
-	int port = 0;
-	int err;
-
-	PMD_INIT_FUNC_TRACE();
-
-	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
-
-	/* Use backpointer here to the PF of this eth_dev */
-	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(eth_dev->data->dev_private);
-
-	/* NFP can not handle DMA addresses requiring more than 40 bits */
-	if (rte_mem_check_dma_mask(40)) {
-		RTE_LOG(ERR, PMD, "device %s can not be used:",
-				   pci_dev->device.name);
-		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
-		return -ENODEV;
-	};
-
-	if ((pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC) ||
-	    (pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC)) {
-		port = ((struct nfp_net_hw *)eth_dev->data->dev_private)->idx;
-		if (port < 0 || port > 7) {
-			PMD_DRV_LOG(ERR, "Port value is wrong");
-			return -ENODEV;
-		}
-
-		/* Use PF array of physical ports to get pointer to
-		 * this specific port
-		 */
-		hw = pf_dev->ports[port];
-
-		PMD_INIT_LOG(DEBUG, "Working with physical port number: %d, "
-				    "NFP internal port number: %d",
-				    port, hw->nfp_idx);
-
-	} else {
-		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
-	}
-
-	eth_dev->dev_ops = &nfp_net_eth_dev_ops;
-	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
-	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
-	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
-
-	/* For secondary processes, the primary has done all the work */
-	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
-		return 0;
-
-	rte_eth_copy_pci_info(eth_dev, pci_dev);
-
-	hw->device_id = pci_dev->id.device_id;
-	hw->vendor_id = pci_dev->id.vendor_id;
-	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
-	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
-
-	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
-		     pci_dev->id.vendor_id, pci_dev->id.device_id,
-		     pci_dev->addr.domain, pci_dev->addr.bus,
-		     pci_dev->addr.devid, pci_dev->addr.function);
-
-	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
-	if (hw->ctrl_bar == NULL) {
-		PMD_DRV_LOG(ERR,
-			"hw->ctrl_bar is NULL. BAR0 not configured");
-		return -ENODEV;
-	}
-
-	if (hw->is_phyport) {
-		if (port == 0) {
-			hw->ctrl_bar = pf_dev->ctrl_bar;
-		} else {
-			if (!pf_dev->ctrl_bar)
-				return -ENODEV;
-			/* Use port offset in pf ctrl_bar for this
-			 * ports control bar
-			 */
-			hw->ctrl_bar = pf_dev->ctrl_bar +
-				       (port * NFP_PF_CSR_SLICE_SIZE);
-		}
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
-
-	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
-	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
-
-	/* Work out where in the BAR the queues start. */
-	switch (pci_dev->id.device_id) {
-	case PCI_DEVICE_ID_NFP4000_PF_NIC:
-	case PCI_DEVICE_ID_NFP6000_PF_NIC:
-	case PCI_DEVICE_ID_NFP6000_VF_NIC:
-		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
-		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
-		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
-		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
-		break;
-	default:
-		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
-		err = -ENODEV;
-		goto dev_err_ctrl_map;
-	}
-
-	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
-	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
-
-	if (hw->is_phyport) {
-		hw->tx_bar = pf_dev->hw_queues + tx_bar_off;
-		hw->rx_bar = pf_dev->hw_queues + rx_bar_off;
-		eth_dev->data->dev_private = hw;
-	} else {
-		hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
-			     tx_bar_off;
-		hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
-			     rx_bar_off;
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
-		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
-
-	nfp_net_cfg_queue_setup(hw);
-
-	/* Get some of the read-only fields from the config BAR */
-	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
-	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
-	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
-	hw->mtu = RTE_ETHER_MTU;
-
-	/* VLAN insertion is incompatible with LSOv2 */
-	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
-		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
-
-	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
-		hw->rx_offset = NFP_NET_RX_OFFSET;
-	else
-		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
-
-	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
-			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
-			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
-
-	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
-		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
-
-	hw->ctrl = 0;
-
-	hw->stride_rx = stride;
-	hw->stride_tx = stride;
-
-	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
-		     hw->max_rx_queues, hw->max_tx_queues);
-
-	/* Initializing spinlock for reconfigs */
-	rte_spinlock_init(&hw->reconfig_lock);
-
-	/* Allocating memory for mac addr */
-	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
-					       RTE_ETHER_ADDR_LEN, 0);
-	if (eth_dev->data->mac_addrs == NULL) {
-		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
-		err = -ENOMEM;
-		goto dev_err_queues_map;
-	}
-
-	if (hw->is_phyport) {
-		nfp_net_pf_read_mac(pf_dev, port);
-		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	}
-
-	if (!rte_is_valid_assigned_ether_addr(
-		    (struct rte_ether_addr *)&hw->mac_addr)) {
-		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
-				   port);
-		/* Using random mac addresses for VFs */
-		rte_eth_random_addr(&hw->mac_addr[0]);
-		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	}
-
-	/* Copying mac address to DPDK eth_dev struct */
-	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
-			&eth_dev->data->mac_addrs[0]);
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
-		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
-
-	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
-
-	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
-		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
-		     eth_dev->data->port_id, pci_dev->id.vendor_id,
-		     pci_dev->id.device_id,
-		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
-		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
-		/* Registering LSC interrupt handler */
-		rte_intr_callback_register(&pci_dev->intr_handle,
-					   nfp_net_dev_interrupt_handler,
-					   (void *)eth_dev);
-		/* Telling the firmware about the LSC interrupt entry */
-		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
-		/* Recording current stats counters values */
-		nfp_net_stats_reset(eth_dev);
-	}
-
-	return 0;
-
-dev_err_queues_map:
-		nfp_cpp_area_free(hw->hwqueues_area);
-dev_err_ctrl_map:
-		nfp_cpp_area_free(hw->ctrl_area);
-
-	return err;
-}
-
-#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
-
-static int
-nfp_fw_upload(struct rte_pci_device *dev, struct nfp_nsp *nsp, char *card)
-{
-	struct nfp_cpp *cpp = nsp->cpp;
-	int fw_f;
-	char *fw_buf;
-	char fw_name[125];
-	char serial[40];
-	struct stat file_stat;
-	off_t fsize, bytes;
-
-	/* Looking for firmware file in order of priority */
-
-	/* First try to find a firmware image specific for this device */
-	snprintf(serial, sizeof(serial),
-			"serial-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x",
-		cpp->serial[0], cpp->serial[1], cpp->serial[2], cpp->serial[3],
-		cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
-		cpp->interface & 0xff);
-
-	snprintf(fw_name, sizeof(fw_name), "%s/%s.nffw", DEFAULT_FW_PATH,
-			serial);
-
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	fw_f = open(fw_name, O_RDONLY);
-	if (fw_f >= 0)
-		goto read_fw;
-
-	/* Then try the PCI name */
-	snprintf(fw_name, sizeof(fw_name), "%s/pci-%s.nffw", DEFAULT_FW_PATH,
-			dev->device.name);
-
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	fw_f = open(fw_name, O_RDONLY);
-	if (fw_f >= 0)
-		goto read_fw;
-
-	/* Finally try the card type and media */
-	snprintf(fw_name, sizeof(fw_name), "%s/%s", DEFAULT_FW_PATH, card);
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	fw_f = open(fw_name, O_RDONLY);
-	if (fw_f < 0) {
-		PMD_DRV_LOG(INFO, "Firmware file %s not found.", fw_name);
-		return -ENOENT;
-	}
-
-read_fw:
-	if (fstat(fw_f, &file_stat) < 0) {
-		PMD_DRV_LOG(INFO, "Firmware file %s size is unknown", fw_name);
-		close(fw_f);
-		return -ENOENT;
-	}
-
-	fsize = file_stat.st_size;
-	PMD_DRV_LOG(INFO, "Firmware file found at %s with size: %" PRIu64 "",
-			    fw_name, (uint64_t)fsize);
-
-	fw_buf = malloc((size_t)fsize);
-	if (!fw_buf) {
-		PMD_DRV_LOG(INFO, "malloc failed for fw buffer");
-		close(fw_f);
-		return -ENOMEM;
-	}
-	memset(fw_buf, 0, fsize);
-
-	bytes = read(fw_f, fw_buf, fsize);
-	if (bytes != fsize) {
-		PMD_DRV_LOG(INFO, "Reading fw to buffer failed."
-				   "Just %" PRIu64 " of %" PRIu64 " bytes read",
-				   (uint64_t)bytes, (uint64_t)fsize);
-		free(fw_buf);
-		close(fw_f);
-		return -EIO;
-	}
-
-	PMD_DRV_LOG(INFO, "Uploading the firmware ...");
-	nfp_nsp_load_fw(nsp, fw_buf, bytes);
-	PMD_DRV_LOG(INFO, "Done");
-
-	free(fw_buf);
-	close(fw_f);
-
-	return 0;
-}
-
-static int
-nfp_fw_setup(struct rte_pci_device *dev, struct nfp_cpp *cpp,
-	     struct nfp_eth_table *nfp_eth_table, struct nfp_hwinfo *hwinfo)
-{
-	struct nfp_nsp *nsp;
-	const char *nfp_fw_model;
-	char card_desc[100];
-	int err = 0;
-
-	nfp_fw_model = nfp_hwinfo_lookup(hwinfo, "assembly.partno");
-
-	if (nfp_fw_model) {
-		PMD_DRV_LOG(INFO, "firmware model found: %s", nfp_fw_model);
-	} else {
-		PMD_DRV_LOG(ERR, "firmware model NOT found");
-		return -EIO;
-	}
-
-	if (nfp_eth_table->count == 0 || nfp_eth_table->count > 8) {
-		PMD_DRV_LOG(ERR, "NFP ethernet table reports wrong ports: %u",
-		       nfp_eth_table->count);
-		return -EIO;
-	}
-
-	PMD_DRV_LOG(INFO, "NFP ethernet port table reports %u ports",
-			   nfp_eth_table->count);
-
-	PMD_DRV_LOG(INFO, "Port speed: %u", nfp_eth_table->ports[0].speed);
-
-	snprintf(card_desc, sizeof(card_desc), "nic_%s_%dx%d.nffw",
-			nfp_fw_model, nfp_eth_table->count,
-			nfp_eth_table->ports[0].speed / 1000);
-
-	nsp = nfp_nsp_open(cpp);
-	if (!nsp) {
-		PMD_DRV_LOG(ERR, "NFP error when obtaining NSP handle");
-		return -EIO;
-	}
-
-	nfp_nsp_device_soft_reset(nsp);
-	err = nfp_fw_upload(dev, nsp, card_desc);
-
-	nfp_nsp_close(nsp);
-	return err;
-}
-
-static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
-{
-	struct nfp_net_hw *hw;
-	struct rte_eth_dev *eth_dev;
-	struct nfp_eth_table *nfp_eth_table = NULL;
-	int ret = 0;
-	int i;
-
-	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
-	if (!nfp_eth_table) {
-		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
-		ret = -EIO;
-		goto error;
-	}
-
-	/* Loop through all physical ports on PF */
-	for (i = 0; i < pf_dev->total_phyports; i++) {
-		const unsigned int numa_node = rte_socket_id();
-		char port_name[RTE_ETH_NAME_MAX_LEN];
-
-		snprintf(port_name, sizeof(port_name), "%s_port%d",
-			 pf_dev->pci_dev->device.name, i);
-
-		/* Allocate a eth_dev for this phyport */
-		eth_dev = rte_eth_dev_allocate(port_name);
-		if (!eth_dev) {
-			ret = -ENODEV;
-			goto port_cleanup;
-		}
-
-		/* Allocate memory for this phyport */
-		eth_dev->data->dev_private =
-			rte_zmalloc_socket(port_name, sizeof(struct nfp_net_hw),
-					   RTE_CACHE_LINE_SIZE, numa_node);
-		if (!eth_dev->data->dev_private) {
-			ret = -ENOMEM;
-			rte_eth_dev_release_port(eth_dev);
-			goto port_cleanup;
-		}
-
-		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
-
-		/* Add this device to the PF's array of physical ports */
-		pf_dev->ports[i] = hw;
-
-		hw->pf_dev = pf_dev;
-		hw->cpp = pf_dev->cpp;
-		hw->eth_dev = eth_dev;
-		hw->idx = i;
-		hw->nfp_idx = nfp_eth_table->ports[i].index;
-		hw->is_phyport = true;
-
-		eth_dev->device = &pf_dev->pci_dev->device;
-
-		/* ctrl/tx/rx BAR mappings and remaining init happens in
-		 * nfp_net_init
-		 */
-		ret = nfp_net_init(eth_dev);
-
-		if (ret) {
-			ret = -ENODEV;
-			goto port_cleanup;
-		}
-
-		rte_eth_dev_probing_finish(eth_dev);
-
-	} /* End loop, all ports on this PF */
-	ret = 0;
-	goto eth_table_cleanup;
-
-port_cleanup:
-	for (i = 0; i < pf_dev->total_phyports; i++) {
-		if (pf_dev->ports[i] && pf_dev->ports[i]->eth_dev) {
-			struct rte_eth_dev *tmp_dev;
-			tmp_dev = pf_dev->ports[i]->eth_dev;
-			rte_eth_dev_release_port(tmp_dev);
-			pf_dev->ports[i] = NULL;
-		}
-	}
-eth_table_cleanup:
-	free(nfp_eth_table);
-error:
-	return ret;
-}
-
-static int nfp_pf_init(struct rte_pci_device *pci_dev)
-{
-	struct nfp_pf_dev *pf_dev = NULL;
-	struct nfp_cpp *cpp;
-	struct nfp_hwinfo *hwinfo;
-	struct nfp_rtsym_table *sym_tbl;
-	struct nfp_eth_table *nfp_eth_table = NULL;
-	char name[RTE_ETH_NAME_MAX_LEN];
-	int total_ports;
-	int ret = -ENODEV;
-	int err;
-
-	if (!pci_dev)
-		return ret;
-
-	/*
-	 * When device bound to UIO, the device could be used, by mistake,
-	 * by two DPDK apps, and the UIO driver does not avoid it. This
-	 * could lead to a serious problem when configuring the NFP CPP
-	 * interface. Here we avoid this telling to the CPP init code to
-	 * use a lock file if UIO is being used.
-	 */
-	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
-		cpp = nfp_cpp_from_device_name(pci_dev, 0);
-	else
-		cpp = nfp_cpp_from_device_name(pci_dev, 1);
-
-	if (!cpp) {
-		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
-		ret = -EIO;
-		goto error;
-	}
-
-	hwinfo = nfp_hwinfo_read(cpp);
-	if (!hwinfo) {
-		PMD_INIT_LOG(ERR, "Error reading hwinfo table");
-		ret = -EIO;
-		goto error;
-	}
-
-	nfp_eth_table = nfp_eth_read_ports(cpp);
-	if (!nfp_eth_table) {
-		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
-		ret = -EIO;
-		goto hwinfo_cleanup;
-	}
-
-	if (nfp_fw_setup(pci_dev, cpp, nfp_eth_table, hwinfo)) {
-		PMD_INIT_LOG(ERR, "Error when uploading firmware");
-		ret = -EIO;
-		goto eth_table_cleanup;
-	}
-
-	/* Now the symbol table should be there */
-	sym_tbl = nfp_rtsym_table_read(cpp);
-	if (!sym_tbl) {
-		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
-				" symbol table");
-		ret = -EIO;
-		goto eth_table_cleanup;
-	}
-
-	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
-	if (total_ports != (int)nfp_eth_table->count) {
-		PMD_DRV_LOG(ERR, "Inconsistent number of ports");
-		ret = -EIO;
-		goto sym_tbl_cleanup;
-	}
-
-	PMD_INIT_LOG(INFO, "Total physical ports: %d", total_ports);
-
-	if (total_ports <= 0 || total_ports > 8) {
-		PMD_INIT_LOG(ERR, "nfd_cfg_pf0_num_ports symbol with wrong value");
-		ret = -ENODEV;
-		goto sym_tbl_cleanup;
-	}
-	/* Allocate memory for the PF "device" */
-	snprintf(name, sizeof(name), "nfp_pf%d", 0);
-	pf_dev = rte_zmalloc(name, sizeof(*pf_dev), 0);
-	if (!pf_dev) {
-		ret = -ENOMEM;
-		goto sym_tbl_cleanup;
-	}
-
-	/* Populate the newly created PF device */
-	pf_dev->cpp = cpp;
-	pf_dev->hwinfo = hwinfo;
-	pf_dev->sym_tbl = sym_tbl;
-	pf_dev->total_phyports = total_ports;
-
-	if (total_ports > 1)
-		pf_dev->multiport = true;
-
-	pf_dev->pci_dev = pci_dev;
-
-	/* Map the symbol table */
-	pf_dev->ctrl_bar = nfp_rtsym_map(pf_dev->sym_tbl, "_pf0_net_bar0",
-				     pf_dev->total_phyports * 32768,
-				     &pf_dev->ctrl_area);
-	if (!pf_dev->ctrl_bar) {
-		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for _pf0_net_ctrl_bar");
-		ret = -EIO;
-		goto pf_cleanup;
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", pf_dev->ctrl_bar);
-
-	/* configure access to tx/rx vNIC BARs */
-	pf_dev->hw_queues = nfp_cpp_map_area(pf_dev->cpp, 0, 0,
-					      NFP_PCIE_QUEUE(0),
-					      NFP_QCP_QUEUE_AREA_SZ,
-					      &pf_dev->hwqueues_area);
-	if (!pf_dev->hw_queues) {
-		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for net.qc");
-		ret = -EIO;
-		goto ctrl_area_cleanup;
-	}
-
-	PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p", pf_dev->hw_queues);
-
-	/* Initialize and prep physical ports now
-	 * This will loop through all physical ports
-	 */
-	ret = nfp_init_phyports(pf_dev);
-	if (ret) {
-		PMD_INIT_LOG(ERR, "Could not create physical ports");
-		goto hwqueues_cleanup;
-	}
-
-	/* register the CPP bridge service here for primary use */
-	nfp_register_cpp_service(pf_dev->cpp);
-
-	return 0;
-
-hwqueues_cleanup:
-	nfp_cpp_area_free(pf_dev->hwqueues_area);
-ctrl_area_cleanup:
-	nfp_cpp_area_free(pf_dev->ctrl_area);
-pf_cleanup:
-	rte_free(pf_dev);
-sym_tbl_cleanup:
-	free(sym_tbl);
-eth_table_cleanup:
-	free(nfp_eth_table);
-hwinfo_cleanup:
-	free(hwinfo);
-error:
-	return ret;
-}
-
-/*
- * When attaching to the NFP4000/6000 PF on a secondary process there
- * is no need to initialize the PF again. Only minimal work is required
- * here
- */
-static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev)
-{
-	struct nfp_cpp *cpp;
-	struct nfp_rtsym_table *sym_tbl;
-	int total_ports;
-	int i;
-	int err;
-
-	if (!pci_dev)
-		return -ENODEV;
-
-	/*
-	 * When device bound to UIO, the device could be used, by mistake,
-	 * by two DPDK apps, and the UIO driver does not avoid it. This
-	 * could lead to a serious problem when configuring the NFP CPP
-	 * interface. Here we avoid this telling to the CPP init code to
-	 * use a lock file if UIO is being used.
-	 */
-	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
-		cpp = nfp_cpp_from_device_name(pci_dev, 0);
-	else
-		cpp = nfp_cpp_from_device_name(pci_dev, 1);
-
-	if (!cpp) {
-		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
-		return -EIO;
-	}
-
-	/*
-	 * We don't have access to the PF created in the primary process
-	 * here so we have to read the number of ports from firmware
-	 */
-	sym_tbl = nfp_rtsym_table_read(cpp);
-	if (!sym_tbl) {
-		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
-				" symbol table");
-		return -EIO;
-	}
-
-	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
-
-	for (i = 0; i < total_ports; i++) {
-		struct rte_eth_dev *eth_dev;
-		char port_name[RTE_ETH_NAME_MAX_LEN];
-
-		snprintf(port_name, sizeof(port_name), "%s_port%d",
-			 pci_dev->device.name, i);
-
-		PMD_DRV_LOG(DEBUG, "Secondary attaching to port %s",
-		    port_name);
-		eth_dev = rte_eth_dev_attach_secondary(port_name);
-		if (!eth_dev) {
-			RTE_LOG(ERR, EAL,
-			"secondary process attach failed, "
-			"ethdev doesn't exist");
-			return -ENODEV;
-		}
-		eth_dev->process_private = cpp;
-		eth_dev->dev_ops = &nfp_net_eth_dev_ops;
-		eth_dev->rx_queue_count = nfp_net_rx_queue_count;
-		eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
-		eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
-		rte_eth_dev_probing_finish(eth_dev);
-	}
-
-	/* Register the CPP bridge service for the secondary too */
-	nfp_register_cpp_service(cpp);
-
-	return 0;
-}
-
-static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
-			    struct rte_pci_device *dev)
-{
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		return nfp_pf_init(dev);
-	else
-		return nfp_pf_secondary_init(dev);
-}
-
-static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP4000_PF_NIC)
-	},
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP6000_PF_NIC)
-	},
-	{
-		.vendor_id = 0,
-	},
-};
-
-static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
-{
-	struct rte_pci_device *pci_dev;
-	uint16_t port_id;
-
-	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
-
-	if (pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC ||
-	    pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC) {
-		/* Free up all physical ports under PF */
-		RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
-			rte_eth_dev_close(port_id);
-		/*
-		 * Ports can be closed and freed but hotplugging is not
-		 * currently supported
-		 */
-		return -ENOTSUP;
-	}
-
-	/* VF cleanup, just free private port data */
-	return nfp_net_close(eth_dev);
-}
-
-static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
-{
-	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
-}
-
-static struct rte_pci_driver rte_nfp_net_pf_pmd = {
-	.id_table = pci_id_nfp_pf_net_map,
-	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
-	.probe = nfp_pf_pci_probe,
-	.remove = eth_nfp_pci_remove,
-};
-
-RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
-RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
-RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_init, init, NOTICE);
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_driver, driver, NOTICE);
 /*
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH 7/7] net/nfp: batch file rename for consistency
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
                   ` (5 preceding siblings ...)
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 6/7] net/nfp: move PF " Heinrich Kuhn
@ 2021-07-16  8:23 ` Heinrich Kuhn
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
  7 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:23 UTC (permalink / raw)
  To: dev; +Cc: heinrich.kuhn, Heinrich Kuhn

Rename the nfp_net.c file to nfp_common as it now contains functions
common to VF and PF functionality. Rename the header file too to be
consistent. Also remove the "net" naming from the _ctrl and _logs files
for consistency across the PMD.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
---
 drivers/net/nfp/meson.build                     | 2 +-
 drivers/net/nfp/{nfp_net.c => nfp_common.c}     | 4 ++--
 drivers/net/nfp/{nfp_net_pmd.h => nfp_common.h} | 6 +++---
 drivers/net/nfp/nfp_cpp_bridge.c                | 2 +-
 drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h}  | 6 +++---
 drivers/net/nfp/nfp_ethdev.c                    | 6 +++---
 drivers/net/nfp/nfp_ethdev_vf.c                 | 6 +++---
 drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h}  | 6 +++---
 drivers/net/nfp/nfp_rxtx.c                      | 6 +++---
 9 files changed, 22 insertions(+), 22 deletions(-)
 rename drivers/net/nfp/{nfp_net.c => nfp_common.c} (99%)
 rename drivers/net/nfp/{nfp_net_pmd.h => nfp_common.h} (99%)
 rename drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h} (99%)
 rename drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h} (94%)

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index ab64d0cac3..810f02ae5b 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -18,7 +18,7 @@ sources = files(
         'nfpcore/nfp_mutex.c',
         'nfpcore/nfp_nsp_eth.c',
         'nfpcore/nfp_hwinfo.c',
-        'nfp_net.c',
+        'nfp_common.c',
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
         'nfp_ethdev_vf.c',
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_common.c
similarity index 99%
rename from drivers/net/nfp/nfp_net.c
rename to drivers/net/nfp/nfp_common.c
index a6097eaab0..87e8f5f333 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_common.c
@@ -8,9 +8,9 @@
 /*
  * vim:shiftwidth=8:noexpandtab
  *
- * @file dpdk/pmd/nfp_net.c
+ * @file dpdk/pmd/nfp_common.c
  *
- * Netronome vNIC DPDK Poll-Mode Driver: Main entry point
+ * Netronome vNIC DPDK Poll-Mode Driver: Common files
  */
 
 #include <rte_byteorder.h>
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_common.h
similarity index 99%
rename from drivers/net/nfp/nfp_net_pmd.h
rename to drivers/net/nfp/nfp_common.h
index dc05e888df..54ac937bd2 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_common.h
@@ -11,8 +11,8 @@
  * Netronome NFP_NET PMD driver
  */
 
-#ifndef _NFP_NET_PMD_H_
-#define _NFP_NET_PMD_H_
+#ifndef _NFP_COMMON_H_
+#define _NFP_COMMON_H_
 
 #define NFP_NET_PMD_VERSION "0.1"
 #define PCI_VENDOR_ID_NETRONOME         0x19ee
@@ -404,7 +404,7 @@ int nfp_net_rss_config_default(struct rte_eth_dev *dev);
 #define NFP_NET_DEV_PRIVATE_TO_PF(dev_priv)\
 	(((struct nfp_net_hw *)dev_priv)->pf_dev)
 
-#endif /* _NFP_NET_PMD_H_ */
+#endif /* _NFP_COMMON_H_ */
 /*
  * Local variables:
  * c-file-style: "Linux"
diff --git a/drivers/net/nfp/nfp_cpp_bridge.c b/drivers/net/nfp/nfp_cpp_bridge.c
index d916793338..74a0eacb3f 100644
--- a/drivers/net/nfp/nfp_cpp_bridge.c
+++ b/drivers/net/nfp/nfp_cpp_bridge.c
@@ -19,7 +19,7 @@
 #include "nfpcore/nfp_mip.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_logs.h"
+#include "nfp_logs.h"
 #include "nfp_cpp_bridge.h"
 
 #include <sys/ioctl.h>
diff --git a/drivers/net/nfp/nfp_net_ctrl.h b/drivers/net/nfp/nfp_ctrl.h
similarity index 99%
rename from drivers/net/nfp/nfp_net_ctrl.h
rename to drivers/net/nfp/nfp_ctrl.h
index 4f26ccf483..4dd62ef194 100644
--- a/drivers/net/nfp/nfp_net_ctrl.h
+++ b/drivers/net/nfp/nfp_ctrl.h
@@ -8,8 +8,8 @@
  *
  * Netronome network device driver: Control BAR layout
  */
-#ifndef _NFP_NET_CTRL_H_
-#define _NFP_NET_CTRL_H_
+#ifndef _NFP_CTRL_H_
+#define _NFP_CTRL_H_
 
 /*
  * Configuration BAR size.
@@ -317,7 +317,7 @@
 /* PF multiport offset */
 #define NFP_PF_CSR_SLICE_SIZE	(32 * 1024)
 
-#endif /* _NFP_NET_CTRL_H_ */
+#endif /* _NFP_CTRL_H_ */
 /*
  * Local variables:
  * c-file-style: "Linux"
diff --git a/drivers/net/nfp/nfp_ethdev.c b/drivers/net/nfp/nfp_ethdev.c
index ab08906704..29e6bb128a 100644
--- a/drivers/net/nfp/nfp_ethdev.c
+++ b/drivers/net/nfp/nfp_ethdev.c
@@ -30,10 +30,10 @@
 #include "nfpcore/nfp_rtsym.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 #include "nfp_cpp_bridge.h"
 
 
diff --git a/drivers/net/nfp/nfp_ethdev_vf.c b/drivers/net/nfp/nfp_ethdev_vf.c
index 223142c0ed..b697b55865 100644
--- a/drivers/net/nfp/nfp_ethdev_vf.c
+++ b/drivers/net/nfp/nfp_ethdev_vf.c
@@ -16,10 +16,10 @@
 #include "nfpcore/nfp_mip.h"
 #include "nfpcore/nfp_rtsym.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 
 static void nfp_netvf_read_mac(struct nfp_net_hw *hw);
 static int nfp_netvf_start(struct rte_eth_dev *dev);
diff --git a/drivers/net/nfp/nfp_net_logs.h b/drivers/net/nfp/nfp_logs.h
similarity index 94%
rename from drivers/net/nfp/nfp_net_logs.h
rename to drivers/net/nfp/nfp_logs.h
index 27dd87611b..bd5a5e1ec5 100644
--- a/drivers/net/nfp/nfp_net_logs.h
+++ b/drivers/net/nfp/nfp_logs.h
@@ -3,8 +3,8 @@
  * All rights reserved.
  */
 
-#ifndef _NFP_NET_LOGS_H_
-#define _NFP_NET_LOGS_H_
+#ifndef _NFP_LOGS_H_
+#define _NFP_LOGS_H_
 
 #include <rte_log.h>
 
@@ -44,4 +44,4 @@ extern int nfp_logtype_driver;
 	rte_log(RTE_LOG_ ## level, nfp_logtype_driver, \
 		"%s(): " fmt "\n", __func__, ## args)
 
-#endif /* _NFP_NET_LOGS_H_ */
+#endif /* _NFP_LOGS_H_ */
diff --git a/drivers/net/nfp/nfp_rxtx.c b/drivers/net/nfp/nfp_rxtx.c
index 9ee9e5c9a3..1402c5f84a 100644
--- a/drivers/net/nfp/nfp_rxtx.c
+++ b/drivers/net/nfp/nfp_rxtx.c
@@ -16,10 +16,10 @@
 #include <ethdev_driver.h>
 #include <ethdev_pci.h>
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 
 /* Prototypes */
 static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD
  2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
                   ` (6 preceding siblings ...)
  2021-07-16  8:23 ` [dpdk-dev] [PATCH 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
@ 2021-07-16  8:35 ` Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
                     ` (8 more replies)
  7 siblings, 9 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn

This patch set restructures the NFP PMD, aligning it more with the
common layout adopted by most other PMD's. Although the changes look
fairly large, functionally nothing is added or removed from the driver
and the existing code is mostly just reorganized into the familiar
structure seen in other PMD's. Apart form adopting the common PMD layout
this change should also aid in future feature development to the NFP
PMD. The previous layout where most of the logic resided in a single
file (nfp_net.c) would have become tedious to support going forward.

v2:
* Added missing sign-off's

Heinrich Kuhn (7):
  net/nfp: split rxtx headers into separate file
  net/nfp: move rxtx functions to their own file
  net/nfp: move CPP bridge to a separate file
  net/nfp: prototype common functions in header file
  net/nfp: move VF functions into new file
  net/nfp: move PF functions into new file
  net/nfp: batch file rename for consistency

 drivers/net/nfp/meson.build                   |    6 +-
 drivers/net/nfp/nfp_common.c                  | 1322 ++++++
 drivers/net/nfp/nfp_common.h                  |  413 ++
 drivers/net/nfp/nfp_cpp_bridge.c              |  392 ++
 drivers/net/nfp/nfp_cpp_bridge.h              |   36 +
 .../net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h}    |    6 +-
 drivers/net/nfp/nfp_ethdev.c                  | 1099 +++++
 drivers/net/nfp/nfp_ethdev_vf.c               |  504 +++
 .../net/nfp/{nfp_net_logs.h => nfp_logs.h}    |    6 +-
 drivers/net/nfp/nfp_net.c                     | 3921 -----------------
 drivers/net/nfp/nfp_rxtx.c                    | 1002 +++++
 drivers/net/nfp/{nfp_net_pmd.h => nfp_rxtx.h} |  279 +-
 12 files changed, 4815 insertions(+), 4171 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_common.c
 create mode 100644 drivers/net/nfp/nfp_common.h
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.c
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.h
 rename drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h} (99%)
 create mode 100644 drivers/net/nfp/nfp_ethdev.c
 create mode 100644 drivers/net/nfp/nfp_ethdev_vf.c
 rename drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h} (94%)
 delete mode 100644 drivers/net/nfp/nfp_net.c
 create mode 100644 drivers/net/nfp/nfp_rxtx.c
 rename drivers/net/nfp/{nfp_net_pmd.h => nfp_rxtx.h} (54%)

-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 1/7] net/nfp: split rxtx headers into separate file
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
@ 2021-07-16  8:35   ` Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
                     ` (7 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

This change splits out the rx/tx specific structs and defines from the
main nfp_net_pmd header file and into their own header file.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/nfp_net.c     |   1 +
 drivers/net/nfp/nfp_net_pmd.h | 248 ------------------------------
 drivers/net/nfp/nfp_rxtx.h    | 276 ++++++++++++++++++++++++++++++++++
 3 files changed, 277 insertions(+), 248 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_rxtx.h

diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index b18edd8c7b..67288abeff 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -38,6 +38,7 @@
 #include "nfpcore/nfp_nsp.h"
 
 #include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
 #include "nfp_net_logs.h"
 #include "nfp_net_ctrl.h"
 
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index 212f9ef162..a3a3ba32d6 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -23,19 +23,6 @@
 /* Forward declaration */
 struct nfp_net_adapter;
 
-/*
- * The maximum number of descriptors is limited by design as
- * DPDK uses uint16_t variables for these values
- */
-#define NFP_NET_MAX_TX_DESC (32 * 1024)
-#define NFP_NET_MIN_TX_DESC 64
-
-#define NFP_NET_MAX_RX_DESC (32 * 1024)
-#define NFP_NET_MIN_RX_DESC 64
-
-/* Descriptor alignment */
-#define NFP_ALIGN_RING_DESC 128
-
 #define NFP_TX_MAX_SEG     UINT8_MAX
 #define NFP_TX_MAX_MTU_SEG 8
 
@@ -150,241 +137,6 @@ static inline void nn_writeq(uint64_t val, volatile void *addr)
 	nn_writel(val, addr);
 }
 
-/* TX descriptor format */
-#define PCIE_DESC_TX_EOP                (1 << 7)
-#define PCIE_DESC_TX_OFFSET_MASK        (0x7f)
-
-/* Flags in the host TX descriptor */
-#define PCIE_DESC_TX_CSUM               (1 << 7)
-#define PCIE_DESC_TX_IP4_CSUM           (1 << 6)
-#define PCIE_DESC_TX_TCP_CSUM           (1 << 5)
-#define PCIE_DESC_TX_UDP_CSUM           (1 << 4)
-#define PCIE_DESC_TX_VLAN               (1 << 3)
-#define PCIE_DESC_TX_LSO                (1 << 2)
-#define PCIE_DESC_TX_ENCAP_NONE         (0)
-#define PCIE_DESC_TX_ENCAP_VXLAN        (1 << 1)
-#define PCIE_DESC_TX_ENCAP_GRE          (1 << 0)
-
-struct nfp_net_tx_desc {
-	union {
-		struct {
-			uint8_t dma_addr_hi; /* High bits of host buf address */
-			__le16 dma_len;     /* Length to DMA for this desc */
-			uint8_t offset_eop; /* Offset in buf where pkt starts +
-					     * highest bit is eop flag.
-					     */
-			__le32 dma_addr_lo; /* Low 32bit of host buf addr */
-
-			__le16 mss;         /* MSS to be used for LSO */
-			uint8_t lso_hdrlen; /* LSO, where the data starts */
-			uint8_t flags;      /* TX Flags, see @PCIE_DESC_TX_* */
-
-			union {
-				struct {
-					/*
-					 * L3 and L4 header offsets required
-					 * for TSOv2
-					 */
-					uint8_t l3_offset;
-					uint8_t l4_offset;
-				};
-				__le16 vlan; /* VLAN tag to add if indicated */
-			};
-			__le16 data_len;    /* Length of frame + meta data */
-		} __rte_packed;
-		__le32 vals[4];
-	};
-};
-
-struct nfp_net_txq {
-	struct nfp_net_hw *hw; /* Backpointer to nfp_net structure */
-
-	/*
-	 * Queue information: @qidx is the queue index from Linux's
-	 * perspective.  @tx_qcidx is the index of the Queue
-	 * Controller Peripheral queue relative to the TX queue BAR.
-	 * @cnt is the size of the queue in number of
-	 * descriptors. @qcp_q is a pointer to the base of the queue
-	 * structure on the NFP
-	 */
-	uint8_t *qcp_q;
-
-	/*
-	 * Read and Write pointers.  @wr_p and @rd_p are host side pointer,
-	 * they are free running and have little relation to the QCP pointers *
-	 * @qcp_rd_p is a local copy queue controller peripheral read pointer
-	 */
-
-	uint32_t wr_p;
-	uint32_t rd_p;
-
-	uint32_t tx_count;
-
-	uint32_t tx_free_thresh;
-
-	/*
-	 * For each descriptor keep a reference to the mbuf and
-	 * DMA address used until completion is signalled.
-	 */
-	struct {
-		struct rte_mbuf *mbuf;
-	} *txbufs;
-
-	/*
-	 * Information about the host side queue location. @txds is
-	 * the virtual address for the queue, @dma is the DMA address
-	 * of the queue and @size is the size in bytes for the queue
-	 * (needed for free)
-	 */
-	struct nfp_net_tx_desc *txds;
-
-	/*
-	 * At this point 48 bytes have been used for all the fields in the
-	 * TX critical path. We have room for 8 bytes and still all placed
-	 * in a cache line. We are not using the threshold values below but
-	 * if we need to, we can add the most used in the remaining bytes.
-	 */
-	uint32_t tx_rs_thresh; /* not used by now. Future? */
-	uint32_t tx_pthresh;   /* not used by now. Future? */
-	uint32_t tx_hthresh;   /* not used by now. Future? */
-	uint32_t tx_wthresh;   /* not used by now. Future? */
-	uint16_t port_id;
-	int qidx;
-	int tx_qcidx;
-	__le64 dma;
-} __rte_aligned(64);
-
-/* RX and freelist descriptor format */
-#define PCIE_DESC_RX_DD                 (1 << 7)
-#define PCIE_DESC_RX_META_LEN_MASK      (0x7f)
-
-/* Flags in the RX descriptor */
-#define PCIE_DESC_RX_RSS                (1 << 15)
-#define PCIE_DESC_RX_I_IP4_CSUM         (1 << 14)
-#define PCIE_DESC_RX_I_IP4_CSUM_OK      (1 << 13)
-#define PCIE_DESC_RX_I_TCP_CSUM         (1 << 12)
-#define PCIE_DESC_RX_I_TCP_CSUM_OK      (1 << 11)
-#define PCIE_DESC_RX_I_UDP_CSUM         (1 << 10)
-#define PCIE_DESC_RX_I_UDP_CSUM_OK      (1 <<  9)
-#define PCIE_DESC_RX_SPARE              (1 <<  8)
-#define PCIE_DESC_RX_EOP                (1 <<  7)
-#define PCIE_DESC_RX_IP4_CSUM           (1 <<  6)
-#define PCIE_DESC_RX_IP4_CSUM_OK        (1 <<  5)
-#define PCIE_DESC_RX_TCP_CSUM           (1 <<  4)
-#define PCIE_DESC_RX_TCP_CSUM_OK        (1 <<  3)
-#define PCIE_DESC_RX_UDP_CSUM           (1 <<  2)
-#define PCIE_DESC_RX_UDP_CSUM_OK        (1 <<  1)
-#define PCIE_DESC_RX_VLAN               (1 <<  0)
-
-#define PCIE_DESC_RX_L4_CSUM_OK         (PCIE_DESC_RX_TCP_CSUM_OK | \
-					 PCIE_DESC_RX_UDP_CSUM_OK)
-struct nfp_net_rx_desc {
-	union {
-		/* Freelist descriptor */
-		struct {
-			uint8_t dma_addr_hi;
-			__le16 spare;
-			uint8_t dd;
-
-			__le32 dma_addr_lo;
-		} __rte_packed fld;
-
-		/* RX descriptor */
-		struct {
-			__le16 data_len;
-			uint8_t reserved;
-			uint8_t meta_len_dd;
-
-			__le16 flags;
-			__le16 vlan;
-		} __rte_packed rxd;
-
-		__le32 vals[2];
-	};
-};
-
-struct nfp_net_rx_buff {
-	struct rte_mbuf *mbuf;
-};
-
-struct nfp_net_rxq {
-	struct nfp_net_hw *hw;	/* Backpointer to nfp_net structure */
-
-	 /*
-	  * @qcp_fl and @qcp_rx are pointers to the base addresses of the
-	  * freelist and RX queue controller peripheral queue structures on the
-	  * NFP
-	  */
-	uint8_t *qcp_fl;
-	uint8_t *qcp_rx;
-
-	/*
-	 * Read and Write pointers.  @wr_p and @rd_p are host side
-	 * pointer, they are free running and have little relation to
-	 * the QCP pointers. @wr_p is where the driver adds new
-	 * freelist descriptors and @rd_p is where the driver start
-	 * reading descriptors for newly arrive packets from.
-	 */
-	uint32_t rd_p;
-
-	/*
-	 * For each buffer placed on the freelist, record the
-	 * associated SKB
-	 */
-	struct nfp_net_rx_buff *rxbufs;
-
-	/*
-	 * Information about the host side queue location.  @rxds is
-	 * the virtual address for the queue
-	 */
-	struct nfp_net_rx_desc *rxds;
-
-	/*
-	 * The mempool is created by the user specifying a mbuf size.
-	 * We save here the reference of the mempool needed in the RX
-	 * path and the mbuf size for checking received packets can be
-	 * safely copied to the mbuf using the NFP_NET_RX_OFFSET
-	 */
-	struct rte_mempool *mem_pool;
-	uint16_t mbuf_size;
-
-	/*
-	 * Next two fields are used for giving more free descriptors
-	 * to the NFP
-	 */
-	uint16_t rx_free_thresh;
-	uint16_t nb_rx_hold;
-
-	 /* the size of the queue in number of descriptors */
-	uint16_t rx_count;
-
-	/*
-	 * Fields above this point fit in a single cache line and are all used
-	 * in the RX critical path. Fields below this point are just used
-	 * during queue configuration or not used at all (yet)
-	 */
-
-	/* referencing dev->data->port_id */
-	uint16_t port_id;
-
-	uint8_t  crc_len; /* Not used by now */
-	uint8_t  drop_en; /* Not used by now */
-
-	/* DMA address of the queue */
-	__le64 dma;
-
-	/*
-	 * Queue information: @qidx is the queue index from Linux's
-	 * perspective.  @fl_qcidx is the index of the Queue
-	 * Controller peripheral queue relative to the RX queue BAR
-	 * used for the freelist and @rx_qcidx is the Queue Controller
-	 * Peripheral index for the RX queue.
-	 */
-	int qidx;
-	int fl_qcidx;
-	int rx_qcidx;
-} __rte_aligned(64);
-
 struct nfp_pf_dev {
 	/* Backpointer to associated pci device */
 	struct rte_pci_device *pci_dev;
diff --git a/drivers/net/nfp/nfp_rxtx.h b/drivers/net/nfp/nfp_rxtx.h
new file mode 100644
index 0000000000..41a3a4b4e7
--- /dev/null
+++ b/drivers/net/nfp/nfp_rxtx.h
@@ -0,0 +1,276 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_rxtx.h
+ *
+ * Netronome NFP Rx/Tx specific header file
+ */
+
+#ifndef _NFP_RXTX_H_
+#define _NFP_RXTX_H_
+
+#include <linux/types.h>
+#include <rte_io.h>
+
+/*
+ * The maximum number of descriptors is limited by design as
+ * DPDK uses uint16_t variables for these values
+ */
+#define NFP_NET_MAX_TX_DESC (32 * 1024)
+#define NFP_NET_MIN_TX_DESC 64
+
+#define NFP_NET_MAX_RX_DESC (32 * 1024)
+#define NFP_NET_MIN_RX_DESC 64
+
+/* Descriptor alignment */
+#define NFP_ALIGN_RING_DESC 128
+
+/* TX descriptor format */
+#define PCIE_DESC_TX_EOP                (1 << 7)
+#define PCIE_DESC_TX_OFFSET_MASK        (0x7f)
+
+/* Flags in the host TX descriptor */
+#define PCIE_DESC_TX_CSUM               (1 << 7)
+#define PCIE_DESC_TX_IP4_CSUM           (1 << 6)
+#define PCIE_DESC_TX_TCP_CSUM           (1 << 5)
+#define PCIE_DESC_TX_UDP_CSUM           (1 << 4)
+#define PCIE_DESC_TX_VLAN               (1 << 3)
+#define PCIE_DESC_TX_LSO                (1 << 2)
+#define PCIE_DESC_TX_ENCAP_NONE         (0)
+#define PCIE_DESC_TX_ENCAP_VXLAN        (1 << 1)
+#define PCIE_DESC_TX_ENCAP_GRE          (1 << 0)
+
+struct nfp_net_tx_desc {
+	union {
+		struct {
+			uint8_t dma_addr_hi; /* High bits of host buf address */
+			__le16 dma_len;     /* Length to DMA for this desc */
+			uint8_t offset_eop; /* Offset in buf where pkt starts +
+					     * highest bit is eop flag.
+					     */
+			__le32 dma_addr_lo; /* Low 32bit of host buf addr */
+
+			__le16 mss;         /* MSS to be used for LSO */
+			uint8_t lso_hdrlen; /* LSO, where the data starts */
+			uint8_t flags;      /* TX Flags, see @PCIE_DESC_TX_* */
+
+			union {
+				struct {
+					/*
+					 * L3 and L4 header offsets required
+					 * for TSOv2
+					 */
+					uint8_t l3_offset;
+					uint8_t l4_offset;
+				};
+				__le16 vlan; /* VLAN tag to add if indicated */
+			};
+			__le16 data_len;    /* Length of frame + meta data */
+		} __rte_packed;
+		__le32 vals[4];
+	};
+};
+
+struct nfp_net_txq {
+	struct nfp_net_hw *hw; /* Backpointer to nfp_net structure */
+
+	/*
+	 * Queue information: @qidx is the queue index from Linux's
+	 * perspective.  @tx_qcidx is the index of the Queue
+	 * Controller Peripheral queue relative to the TX queue BAR.
+	 * @cnt is the size of the queue in number of
+	 * descriptors. @qcp_q is a pointer to the base of the queue
+	 * structure on the NFP
+	 */
+	uint8_t *qcp_q;
+
+	/*
+	 * Read and Write pointers.  @wr_p and @rd_p are host side pointer,
+	 * they are free running and have little relation to the QCP pointers *
+	 * @qcp_rd_p is a local copy queue controller peripheral read pointer
+	 */
+
+	uint32_t wr_p;
+	uint32_t rd_p;
+
+	uint32_t tx_count;
+
+	uint32_t tx_free_thresh;
+
+	/*
+	 * For each descriptor keep a reference to the mbuf and
+	 * DMA address used until completion is signalled.
+	 */
+	struct {
+		struct rte_mbuf *mbuf;
+	} *txbufs;
+
+	/*
+	 * Information about the host side queue location. @txds is
+	 * the virtual address for the queue, @dma is the DMA address
+	 * of the queue and @size is the size in bytes for the queue
+	 * (needed for free)
+	 */
+	struct nfp_net_tx_desc *txds;
+
+	/*
+	 * At this point 48 bytes have been used for all the fields in the
+	 * TX critical path. We have room for 8 bytes and still all placed
+	 * in a cache line. We are not using the threshold values below but
+	 * if we need to, we can add the most used in the remaining bytes.
+	 */
+	uint32_t tx_rs_thresh; /* not used by now. Future? */
+	uint32_t tx_pthresh;   /* not used by now. Future? */
+	uint32_t tx_hthresh;   /* not used by now. Future? */
+	uint32_t tx_wthresh;   /* not used by now. Future? */
+	uint16_t port_id;
+	int qidx;
+	int tx_qcidx;
+	__le64 dma;
+} __rte_aligned(64);
+
+/* RX and freelist descriptor format */
+#define PCIE_DESC_RX_DD                 (1 << 7)
+#define PCIE_DESC_RX_META_LEN_MASK      (0x7f)
+
+/* Flags in the RX descriptor */
+#define PCIE_DESC_RX_RSS                (1 << 15)
+#define PCIE_DESC_RX_I_IP4_CSUM         (1 << 14)
+#define PCIE_DESC_RX_I_IP4_CSUM_OK      (1 << 13)
+#define PCIE_DESC_RX_I_TCP_CSUM         (1 << 12)
+#define PCIE_DESC_RX_I_TCP_CSUM_OK      (1 << 11)
+#define PCIE_DESC_RX_I_UDP_CSUM         (1 << 10)
+#define PCIE_DESC_RX_I_UDP_CSUM_OK      (1 <<  9)
+#define PCIE_DESC_RX_SPARE              (1 <<  8)
+#define PCIE_DESC_RX_EOP                (1 <<  7)
+#define PCIE_DESC_RX_IP4_CSUM           (1 <<  6)
+#define PCIE_DESC_RX_IP4_CSUM_OK        (1 <<  5)
+#define PCIE_DESC_RX_TCP_CSUM           (1 <<  4)
+#define PCIE_DESC_RX_TCP_CSUM_OK        (1 <<  3)
+#define PCIE_DESC_RX_UDP_CSUM           (1 <<  2)
+#define PCIE_DESC_RX_UDP_CSUM_OK        (1 <<  1)
+#define PCIE_DESC_RX_VLAN               (1 <<  0)
+
+#define PCIE_DESC_RX_L4_CSUM_OK         (PCIE_DESC_RX_TCP_CSUM_OK | \
+					 PCIE_DESC_RX_UDP_CSUM_OK)
+
+struct nfp_net_rx_desc {
+	union {
+		/* Freelist descriptor */
+		struct {
+			uint8_t dma_addr_hi;
+			__le16 spare;
+			uint8_t dd;
+
+			__le32 dma_addr_lo;
+		} __rte_packed fld;
+
+		/* RX descriptor */
+		struct {
+			__le16 data_len;
+			uint8_t reserved;
+			uint8_t meta_len_dd;
+
+			__le16 flags;
+			__le16 vlan;
+		} __rte_packed rxd;
+
+		__le32 vals[2];
+	};
+};
+
+struct nfp_net_rx_buff {
+	struct rte_mbuf *mbuf;
+};
+
+struct nfp_net_rxq {
+	struct nfp_net_hw *hw;	/* Backpointer to nfp_net structure */
+
+	 /*
+	  * @qcp_fl and @qcp_rx are pointers to the base addresses of the
+	  * freelist and RX queue controller peripheral queue structures on the
+	  * NFP
+	  */
+	uint8_t *qcp_fl;
+	uint8_t *qcp_rx;
+
+	/*
+	 * Read and Write pointers.  @wr_p and @rd_p are host side
+	 * pointer, they are free running and have little relation to
+	 * the QCP pointers. @wr_p is where the driver adds new
+	 * freelist descriptors and @rd_p is where the driver start
+	 * reading descriptors for newly arrive packets from.
+	 */
+	uint32_t rd_p;
+
+	/*
+	 * For each buffer placed on the freelist, record the
+	 * associated SKB
+	 */
+	struct nfp_net_rx_buff *rxbufs;
+
+	/*
+	 * Information about the host side queue location.  @rxds is
+	 * the virtual address for the queue
+	 */
+	struct nfp_net_rx_desc *rxds;
+
+	/*
+	 * The mempool is created by the user specifying a mbuf size.
+	 * We save here the reference of the mempool needed in the RX
+	 * path and the mbuf size for checking received packets can be
+	 * safely copied to the mbuf using the NFP_NET_RX_OFFSET
+	 */
+	struct rte_mempool *mem_pool;
+	uint16_t mbuf_size;
+
+	/*
+	 * Next two fields are used for giving more free descriptors
+	 * to the NFP
+	 */
+	uint16_t rx_free_thresh;
+	uint16_t nb_rx_hold;
+
+	 /* the size of the queue in number of descriptors */
+	uint16_t rx_count;
+
+	/*
+	 * Fields above this point fit in a single cache line and are all used
+	 * in the RX critical path. Fields below this point are just used
+	 * during queue configuration or not used at all (yet)
+	 */
+
+	/* referencing dev->data->port_id */
+	uint16_t port_id;
+
+	uint8_t  crc_len; /* Not used by now */
+	uint8_t  drop_en; /* Not used by now */
+
+	/* DMA address of the queue */
+	__le64 dma;
+
+	/*
+	 * Queue information: @qidx is the queue index from Linux's
+	 * perspective.  @fl_qcidx is the index of the Queue
+	 * Controller peripheral queue relative to the RX queue BAR
+	 * used for the freelist and @rx_qcidx is the Queue Controller
+	 * Peripheral index for the RX queue.
+	 */
+	int qidx;
+	int fl_qcidx;
+	int rx_qcidx;
+} __rte_aligned(64);
+
+#endif /* _NFP_RXTX_H_ */
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
+
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 2/7] net/nfp: move rxtx functions to their own file
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
@ 2021-07-16  8:35   ` Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
                     ` (6 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Create a new rxtx file and move the Rx/Tx functions to this file. This
commit will also move the needed shared functions to the nfp_net_pmd.h
file as needed.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build   |    1 +
 drivers/net/nfp/nfp_net.c     | 1090 ---------------------------------
 drivers/net/nfp/nfp_net_pmd.h |  184 ++++--
 drivers/net/nfp/nfp_rxtx.c    | 1002 ++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_rxtx.h    |   27 +
 5 files changed, 1173 insertions(+), 1131 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_rxtx.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index b51e2e5f20..1b289e2354 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -19,4 +19,5 @@ sources = files(
         'nfpcore/nfp_nsp_eth.c',
         'nfpcore/nfp_hwinfo.c',
         'nfp_net.c',
+        'nfp_rxtx.c',
 )
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 67288abeff..5bfc23ba04 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -66,29 +66,11 @@ static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
 static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
 static int nfp_net_promisc_enable(struct rte_eth_dev *dev);
 static int nfp_net_promisc_disable(struct rte_eth_dev *dev);
-static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
-static uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
-				       uint16_t queue_idx);
-static uint16_t nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
-				  uint16_t nb_pkts);
-static void nfp_net_rx_queue_release(void *rxq);
-static int nfp_net_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-				  uint16_t nb_desc, unsigned int socket_id,
-				  const struct rte_eth_rxconf *rx_conf,
-				  struct rte_mempool *mp);
-static int nfp_net_tx_free_bufs(struct nfp_net_txq *txq);
-static void nfp_net_tx_queue_release(void *txq);
-static int nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-				  uint16_t nb_desc, unsigned int socket_id,
-				  const struct rte_eth_txconf *tx_conf);
 static int nfp_net_start(struct rte_eth_dev *dev);
 static int nfp_net_stats_get(struct rte_eth_dev *dev,
 			      struct rte_eth_stats *stats);
 static int nfp_net_stats_reset(struct rte_eth_dev *dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
-static uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
-				  uint16_t nb_pkts);
-
 static int nfp_net_rss_config_default(struct rte_eth_dev *dev);
 static int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 				   struct rte_eth_rss_conf *rss_conf);
@@ -106,184 +88,6 @@ static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_eth_table *nfp_eth_table,
 			struct nfp_hwinfo *hwinfo);
 
-
-/* The offset of the queue controller queues in the PCIe Target */
-#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
-
-/* Maximum value which can be added to a queue with one transaction */
-#define NFP_QCP_MAX_ADD	0x7f
-
-#define RTE_MBUF_DMA_ADDR_DEFAULT(mb) \
-	(uint64_t)((mb)->buf_iova + RTE_PKTMBUF_HEADROOM)
-
-/* nfp_qcp_ptr - Read or Write Pointer of a queue */
-enum nfp_qcp_ptr {
-	NFP_QCP_READ_PTR = 0,
-	NFP_QCP_WRITE_PTR
-};
-
-/*
- * nfp_qcp_ptr_add - Add the value to the selected pointer of a queue
- * @q: Base address for queue structure
- * @ptr: Add to the Read or Write pointer
- * @val: Value to add to the queue pointer
- *
- * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed.
- */
-static inline void
-nfp_qcp_ptr_add(uint8_t *q, enum nfp_qcp_ptr ptr, uint32_t val)
-{
-	uint32_t off;
-
-	if (ptr == NFP_QCP_READ_PTR)
-		off = NFP_QCP_QUEUE_ADD_RPTR;
-	else
-		off = NFP_QCP_QUEUE_ADD_WPTR;
-
-	while (val > NFP_QCP_MAX_ADD) {
-		nn_writel(rte_cpu_to_le_32(NFP_QCP_MAX_ADD), q + off);
-		val -= NFP_QCP_MAX_ADD;
-	}
-
-	nn_writel(rte_cpu_to_le_32(val), q + off);
-}
-
-/*
- * nfp_qcp_read - Read the current Read/Write pointer value for a queue
- * @q:  Base address for queue structure
- * @ptr: Read or Write pointer
- */
-static inline uint32_t
-nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
-{
-	uint32_t off;
-	uint32_t val;
-
-	if (ptr == NFP_QCP_READ_PTR)
-		off = NFP_QCP_QUEUE_STS_LO;
-	else
-		off = NFP_QCP_QUEUE_STS_HI;
-
-	val = rte_cpu_to_le_32(nn_readl(q + off));
-
-	if (ptr == NFP_QCP_READ_PTR)
-		return val & NFP_QCP_QUEUE_STS_LO_READPTR_mask;
-	else
-		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
-}
-
-/*
- * Functions to read/write from/to Config BAR
- * Performs any endian conversion necessary.
- */
-static inline uint8_t
-nn_cfg_readb(struct nfp_net_hw *hw, int off)
-{
-	return nn_readb(hw->ctrl_bar + off);
-}
-
-static inline void
-nn_cfg_writeb(struct nfp_net_hw *hw, int off, uint8_t val)
-{
-	nn_writeb(val, hw->ctrl_bar + off);
-}
-
-static inline uint32_t
-nn_cfg_readl(struct nfp_net_hw *hw, int off)
-{
-	return rte_le_to_cpu_32(nn_readl(hw->ctrl_bar + off));
-}
-
-static inline void
-nn_cfg_writel(struct nfp_net_hw *hw, int off, uint32_t val)
-{
-	nn_writel(rte_cpu_to_le_32(val), hw->ctrl_bar + off);
-}
-
-static inline uint64_t
-nn_cfg_readq(struct nfp_net_hw *hw, int off)
-{
-	return rte_le_to_cpu_64(nn_readq(hw->ctrl_bar + off));
-}
-
-static inline void
-nn_cfg_writeq(struct nfp_net_hw *hw, int off, uint64_t val)
-{
-	nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
-}
-
-static void
-nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
-{
-	unsigned i;
-
-	if (rxq->rxbufs == NULL)
-		return;
-
-	for (i = 0; i < rxq->rx_count; i++) {
-		if (rxq->rxbufs[i].mbuf) {
-			rte_pktmbuf_free_seg(rxq->rxbufs[i].mbuf);
-			rxq->rxbufs[i].mbuf = NULL;
-		}
-	}
-}
-
-static void
-nfp_net_rx_queue_release(void *rx_queue)
-{
-	struct nfp_net_rxq *rxq = rx_queue;
-
-	if (rxq) {
-		nfp_net_rx_queue_release_mbufs(rxq);
-		rte_free(rxq->rxbufs);
-		rte_free(rxq);
-	}
-}
-
-static void
-nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
-{
-	nfp_net_rx_queue_release_mbufs(rxq);
-	rxq->rd_p = 0;
-	rxq->nb_rx_hold = 0;
-}
-
-static void
-nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq)
-{
-	unsigned i;
-
-	if (txq->txbufs == NULL)
-		return;
-
-	for (i = 0; i < txq->tx_count; i++) {
-		if (txq->txbufs[i].mbuf) {
-			rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
-			txq->txbufs[i].mbuf = NULL;
-		}
-	}
-}
-
-static void
-nfp_net_tx_queue_release(void *tx_queue)
-{
-	struct nfp_net_txq *txq = tx_queue;
-
-	if (txq) {
-		nfp_net_tx_queue_release_mbufs(txq);
-		rte_free(txq->txbufs);
-		rte_free(txq);
-	}
-}
-
-static void
-nfp_net_reset_tx_queue(struct nfp_net_txq *txq)
-{
-	nfp_net_tx_queue_release_mbufs(txq);
-	txq->wr_p = 0;
-	txq->rd_p = 0;
-}
-
 static int
 __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
 {
@@ -461,18 +265,6 @@ nfp_net_disable_queues(struct rte_eth_dev *dev)
 	hw->ctrl = new_ctrl;
 }
 
-static int
-nfp_net_rx_freelist_setup(struct rte_eth_dev *dev)
-{
-	int i;
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		if (nfp_net_rx_fill_freelist(dev->data->rx_queues[i]) < 0)
-			return -1;
-	}
-	return 0;
-}
-
 static void
 nfp_net_params_setup(struct nfp_net_hw *hw)
 {
@@ -1349,44 +1141,6 @@ nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 	return NULL;
 }
 
-static uint32_t
-nfp_net_rx_queue_count(struct rte_eth_dev *dev, uint16_t queue_idx)
-{
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_rx_desc *rxds;
-	uint32_t idx;
-	uint32_t count;
-
-	rxq = (struct nfp_net_rxq *)dev->data->rx_queues[queue_idx];
-
-	idx = rxq->rd_p;
-
-	count = 0;
-
-	/*
-	 * Other PMDs are just checking the DD bit in intervals of 4
-	 * descriptors and counting all four if the first has the DD
-	 * bit on. Of course, this is not accurate but can be good for
-	 * performance. But ideally that should be done in descriptors
-	 * chunks belonging to the same cache line
-	 */
-
-	while (count < rxq->rx_count) {
-		rxds = &rxq->rxds[idx];
-		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
-			break;
-
-		count++;
-		idx++;
-
-		/* Wrapping? */
-		if ((idx) == rxq->rx_count)
-			idx = 0;
-	}
-
-	return count;
-}
-
 static int
 nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
@@ -1568,850 +1322,6 @@ nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	return 0;
 }
 
-static int
-nfp_net_rx_queue_setup(struct rte_eth_dev *dev,
-		       uint16_t queue_idx, uint16_t nb_desc,
-		       unsigned int socket_id,
-		       const struct rte_eth_rxconf *rx_conf,
-		       struct rte_mempool *mp)
-{
-	const struct rte_memzone *tz;
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_hw *hw;
-	uint32_t rx_desc_sz;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	PMD_INIT_FUNC_TRACE();
-
-	/* Validating number of descriptors */
-	rx_desc_sz = nb_desc * sizeof(struct nfp_net_rx_desc);
-	if (rx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
-	    nb_desc > NFP_NET_MAX_RX_DESC ||
-	    nb_desc < NFP_NET_MIN_RX_DESC) {
-		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
-		return -EINVAL;
-	}
-
-	/*
-	 * Free memory prior to re-allocation if needed. This is the case after
-	 * calling nfp_net_stop
-	 */
-	if (dev->data->rx_queues[queue_idx]) {
-		nfp_net_rx_queue_release(dev->data->rx_queues[queue_idx]);
-		dev->data->rx_queues[queue_idx] = NULL;
-	}
-
-	/* Allocating rx queue data structure */
-	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct nfp_net_rxq),
-				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (rxq == NULL)
-		return -ENOMEM;
-
-	/* Hw queues mapping based on firmware configuration */
-	rxq->qidx = queue_idx;
-	rxq->fl_qcidx = queue_idx * hw->stride_rx;
-	rxq->rx_qcidx = rxq->fl_qcidx + (hw->stride_rx - 1);
-	rxq->qcp_fl = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->fl_qcidx);
-	rxq->qcp_rx = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->rx_qcidx);
-
-	/*
-	 * Tracking mbuf size for detecting a potential mbuf overflow due to
-	 * RX offset
-	 */
-	rxq->mem_pool = mp;
-	rxq->mbuf_size = rxq->mem_pool->elt_size;
-	rxq->mbuf_size -= (sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM);
-	hw->flbufsz = rxq->mbuf_size;
-
-	rxq->rx_count = nb_desc;
-	rxq->port_id = dev->data->port_id;
-	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
-	rxq->drop_en = rx_conf->rx_drop_en;
-
-	/*
-	 * Allocate RX ring hardware descriptors. A memzone large enough to
-	 * handle the maximum ring size is allocated in order to allow for
-	 * resizing in later calls to the queue setup function.
-	 */
-	tz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
-				   sizeof(struct nfp_net_rx_desc) *
-				   NFP_NET_MAX_RX_DESC, NFP_MEMZONE_ALIGN,
-				   socket_id);
-
-	if (tz == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating rx dma");
-		nfp_net_rx_queue_release(rxq);
-		return -ENOMEM;
-	}
-
-	/* Saving physical and virtual addresses for the RX ring */
-	rxq->dma = (uint64_t)tz->iova;
-	rxq->rxds = (struct nfp_net_rx_desc *)tz->addr;
-
-	/* mbuf pointers array for referencing mbufs linked to RX descriptors */
-	rxq->rxbufs = rte_zmalloc_socket("rxq->rxbufs",
-					 sizeof(*rxq->rxbufs) * nb_desc,
-					 RTE_CACHE_LINE_SIZE, socket_id);
-	if (rxq->rxbufs == NULL) {
-		nfp_net_rx_queue_release(rxq);
-		return -ENOMEM;
-	}
-
-	PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
-		   rxq->rxbufs, rxq->rxds, (unsigned long int)rxq->dma);
-
-	nfp_net_reset_rx_queue(rxq);
-
-	dev->data->rx_queues[queue_idx] = rxq;
-	rxq->hw = hw;
-
-	/*
-	 * Telling the HW about the physical address of the RX ring and number
-	 * of descriptors in log2 format
-	 */
-	nn_cfg_writeq(hw, NFP_NET_CFG_RXR_ADDR(queue_idx), rxq->dma);
-	nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), rte_log2_u32(nb_desc));
-
-	return 0;
-}
-
-static int
-nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq)
-{
-	struct nfp_net_rx_buff *rxe = rxq->rxbufs;
-	uint64_t dma_addr;
-	unsigned i;
-
-	PMD_RX_LOG(DEBUG, "nfp_net_rx_fill_freelist for %u descriptors",
-		   rxq->rx_count);
-
-	for (i = 0; i < rxq->rx_count; i++) {
-		struct nfp_net_rx_desc *rxd;
-		struct rte_mbuf *mbuf = rte_pktmbuf_alloc(rxq->mem_pool);
-
-		if (mbuf == NULL) {
-			PMD_DRV_LOG(ERR, "RX mbuf alloc failed queue_id=%u",
-				(unsigned)rxq->qidx);
-			return -ENOMEM;
-		}
-
-		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(mbuf));
-
-		rxd = &rxq->rxds[i];
-		rxd->fld.dd = 0;
-		rxd->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
-		rxd->fld.dma_addr_lo = dma_addr & 0xffffffff;
-		rxe[i].mbuf = mbuf;
-		PMD_RX_LOG(DEBUG, "[%d]: %" PRIx64, i, dma_addr);
-	}
-
-	/* Make sure all writes are flushed before telling the hardware */
-	rte_wmb();
-
-	/* Not advertising the whole ring as the firmware gets confused if so */
-	PMD_RX_LOG(DEBUG, "Increment FL write pointer in %u",
-		   rxq->rx_count - 1);
-
-	nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, rxq->rx_count - 1);
-
-	return 0;
-}
-
-static int
-nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-		       uint16_t nb_desc, unsigned int socket_id,
-		       const struct rte_eth_txconf *tx_conf)
-{
-	const struct rte_memzone *tz;
-	struct nfp_net_txq *txq;
-	uint16_t tx_free_thresh;
-	struct nfp_net_hw *hw;
-	uint32_t tx_desc_sz;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	PMD_INIT_FUNC_TRACE();
-
-	/* Validating number of descriptors */
-	tx_desc_sz = nb_desc * sizeof(struct nfp_net_tx_desc);
-	if (tx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
-	    nb_desc > NFP_NET_MAX_TX_DESC ||
-	    nb_desc < NFP_NET_MIN_TX_DESC) {
-		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
-		return -EINVAL;
-	}
-
-	tx_free_thresh = (uint16_t)((tx_conf->tx_free_thresh) ?
-				    tx_conf->tx_free_thresh :
-				    DEFAULT_TX_FREE_THRESH);
-
-	if (tx_free_thresh > (nb_desc)) {
-		PMD_DRV_LOG(ERR,
-			"tx_free_thresh must be less than the number of TX "
-			"descriptors. (tx_free_thresh=%u port=%d "
-			"queue=%d)", (unsigned int)tx_free_thresh,
-			dev->data->port_id, (int)queue_idx);
-		return -(EINVAL);
-	}
-
-	/*
-	 * Free memory prior to re-allocation if needed. This is the case after
-	 * calling nfp_net_stop
-	 */
-	if (dev->data->tx_queues[queue_idx]) {
-		PMD_TX_LOG(DEBUG, "Freeing memory prior to re-allocation %d",
-			   queue_idx);
-		nfp_net_tx_queue_release(dev->data->tx_queues[queue_idx]);
-		dev->data->tx_queues[queue_idx] = NULL;
-	}
-
-	/* Allocating tx queue data structure */
-	txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct nfp_net_txq),
-				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating tx dma");
-		return -ENOMEM;
-	}
-
-	/*
-	 * Allocate TX ring hardware descriptors. A memzone large enough to
-	 * handle the maximum ring size is allocated in order to allow for
-	 * resizing in later calls to the queue setup function.
-	 */
-	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
-				   sizeof(struct nfp_net_tx_desc) *
-				   NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN,
-				   socket_id);
-	if (tz == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating tx dma");
-		nfp_net_tx_queue_release(txq);
-		return -ENOMEM;
-	}
-
-	txq->tx_count = nb_desc;
-	txq->tx_free_thresh = tx_free_thresh;
-	txq->tx_pthresh = tx_conf->tx_thresh.pthresh;
-	txq->tx_hthresh = tx_conf->tx_thresh.hthresh;
-	txq->tx_wthresh = tx_conf->tx_thresh.wthresh;
-
-	/* queue mapping based on firmware configuration */
-	txq->qidx = queue_idx;
-	txq->tx_qcidx = queue_idx * hw->stride_tx;
-	txq->qcp_q = hw->tx_bar + NFP_QCP_QUEUE_OFF(txq->tx_qcidx);
-
-	txq->port_id = dev->data->port_id;
-
-	/* Saving physical and virtual addresses for the TX ring */
-	txq->dma = (uint64_t)tz->iova;
-	txq->txds = (struct nfp_net_tx_desc *)tz->addr;
-
-	/* mbuf pointers array for referencing mbufs linked to TX descriptors */
-	txq->txbufs = rte_zmalloc_socket("txq->txbufs",
-					 sizeof(*txq->txbufs) * nb_desc,
-					 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq->txbufs == NULL) {
-		nfp_net_tx_queue_release(txq);
-		return -ENOMEM;
-	}
-	PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
-		   txq->txbufs, txq->txds, (unsigned long int)txq->dma);
-
-	nfp_net_reset_tx_queue(txq);
-
-	dev->data->tx_queues[queue_idx] = txq;
-	txq->hw = hw;
-
-	/*
-	 * Telling the HW about the physical address of the TX ring and number
-	 * of descriptors in log2 format
-	 */
-	nn_cfg_writeq(hw, NFP_NET_CFG_TXR_ADDR(queue_idx), txq->dma);
-	nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), rte_log2_u32(nb_desc));
-
-	return 0;
-}
-
-/* nfp_net_tx_tso - Set TX descriptor for TSO */
-static inline void
-nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
-	       struct rte_mbuf *mb)
-{
-	uint64_t ol_flags;
-	struct nfp_net_hw *hw = txq->hw;
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY))
-		goto clean_txd;
-
-	ol_flags = mb->ol_flags;
-
-	if (!(ol_flags & PKT_TX_TCP_SEG))
-		goto clean_txd;
-
-	txd->l3_offset = mb->l2_len;
-	txd->l4_offset = mb->l2_len + mb->l3_len;
-	txd->lso_hdrlen = mb->l2_len + mb->l3_len + mb->l4_len;
-	txd->mss = rte_cpu_to_le_16(mb->tso_segsz);
-	txd->flags = PCIE_DESC_TX_LSO;
-	return;
-
-clean_txd:
-	txd->flags = 0;
-	txd->l3_offset = 0;
-	txd->l4_offset = 0;
-	txd->lso_hdrlen = 0;
-	txd->mss = 0;
-}
-
-/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
-static inline void
-nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
-		 struct rte_mbuf *mb)
-{
-	uint64_t ol_flags;
-	struct nfp_net_hw *hw = txq->hw;
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_TXCSUM))
-		return;
-
-	ol_flags = mb->ol_flags;
-
-	/* IPv6 does not need checksum */
-	if (ol_flags & PKT_TX_IP_CKSUM)
-		txd->flags |= PCIE_DESC_TX_IP4_CSUM;
-
-	switch (ol_flags & PKT_TX_L4_MASK) {
-	case PKT_TX_UDP_CKSUM:
-		txd->flags |= PCIE_DESC_TX_UDP_CSUM;
-		break;
-	case PKT_TX_TCP_CKSUM:
-		txd->flags |= PCIE_DESC_TX_TCP_CSUM;
-		break;
-	}
-
-	if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
-		txd->flags |= PCIE_DESC_TX_CSUM;
-}
-
-/* nfp_net_rx_cksum - set mbuf checksum flags based on RX descriptor flags */
-static inline void
-nfp_net_rx_cksum(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
-		 struct rte_mbuf *mb)
-{
-	struct nfp_net_hw *hw = rxq->hw;
-
-	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RXCSUM))
-		return;
-
-	/* If IPv4 and IP checksum error, fail */
-	if (unlikely((rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM) &&
-	    !(rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM_OK)))
-		mb->ol_flags |= PKT_RX_IP_CKSUM_BAD;
-	else
-		mb->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
-
-	/* If neither UDP nor TCP return */
-	if (!(rxd->rxd.flags & PCIE_DESC_RX_TCP_CSUM) &&
-	    !(rxd->rxd.flags & PCIE_DESC_RX_UDP_CSUM))
-		return;
-
-	if (likely(rxd->rxd.flags & PCIE_DESC_RX_L4_CSUM_OK))
-		mb->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
-	else
-		mb->ol_flags |= PKT_RX_L4_CKSUM_BAD;
-}
-
-#define NFP_HASH_OFFSET      ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 4)
-#define NFP_HASH_TYPE_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 8)
-
-#define NFP_DESC_META_LEN(d) (d->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
-
-/*
- * nfp_net_set_hash - Set mbuf hash data
- *
- * The RSS hash and hash-type are pre-pended to the packet data.
- * Extract and decode it and set the mbuf fields.
- */
-static inline void
-nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
-		 struct rte_mbuf *mbuf)
-{
-	struct nfp_net_hw *hw = rxq->hw;
-	uint8_t *meta_offset;
-	uint32_t meta_info;
-	uint32_t hash = 0;
-	uint32_t hash_type = 0;
-
-	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
-		return;
-
-	/* this is true for new firmwares */
-	if (likely(((hw->cap & NFP_NET_CFG_CTRL_RSS2) ||
-	    (NFD_CFG_MAJOR_VERSION_of(hw->ver) == 4)) &&
-	     NFP_DESC_META_LEN(rxd))) {
-		/*
-		 * new metadata api:
-		 * <----  32 bit  ----->
-		 * m    field type word
-		 * e     data field #2
-		 * t     data field #1
-		 * a     data field #0
-		 * ====================
-		 *    packet data
-		 *
-		 * Field type word contains up to 8 4bit field types
-		 * A 4bit field type refers to a data field word
-		 * A data field word can have several 4bit field types
-		 */
-		meta_offset = rte_pktmbuf_mtod(mbuf, uint8_t *);
-		meta_offset -= NFP_DESC_META_LEN(rxd);
-		meta_info = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
-		meta_offset += 4;
-		/* NFP PMD just supports metadata for hashing */
-		switch (meta_info & NFP_NET_META_FIELD_MASK) {
-		case NFP_NET_META_HASH:
-			/* next field type is about the hash type */
-			meta_info >>= NFP_NET_META_FIELD_SIZE;
-			/* hash value is in the data field */
-			hash = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
-			hash_type = meta_info & NFP_NET_META_FIELD_MASK;
-			break;
-		default:
-			/* Unsupported metadata can be a performance issue */
-			return;
-		}
-	} else {
-		if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
-			return;
-
-		hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
-		hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
-	}
-
-	mbuf->hash.rss = hash;
-	mbuf->ol_flags |= PKT_RX_RSS_HASH;
-
-	switch (hash_type) {
-	case NFP_NET_RSS_IPV4:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV4;
-		break;
-	case NFP_NET_RSS_IPV6:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6;
-		break;
-	case NFP_NET_RSS_IPV6_EX:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV4_TCP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV6_TCP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV4_UDP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV6_UDP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	default:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L4_MASK;
-	}
-}
-
-static inline void
-nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq)
-{
-	rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
-}
-
-#define NFP_DESC_META_LEN(d) (d->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
-
-/*
- * RX path design:
- *
- * There are some decisions to take:
- * 1) How to check DD RX descriptors bit
- * 2) How and when to allocate new mbufs
- *
- * Current implementation checks just one single DD bit each loop. As each
- * descriptor is 8 bytes, it is likely a good idea to check descriptors in
- * a single cache line instead. Tests with this change have not shown any
- * performance improvement but it requires further investigation. For example,
- * depending on which descriptor is next, the number of descriptors could be
- * less than 8 for just checking those in the same cache line. This implies
- * extra work which could be counterproductive by itself. Indeed, last firmware
- * changes are just doing this: writing several descriptors with the DD bit
- * for saving PCIe bandwidth and DMA operations from the NFP.
- *
- * Mbuf allocation is done when a new packet is received. Then the descriptor
- * is automatically linked with the new mbuf and the old one is given to the
- * user. The main drawback with this design is mbuf allocation is heavier than
- * using bulk allocations allowed by DPDK with rte_mempool_get_bulk. From the
- * cache point of view it does not seem allocating the mbuf early on as we are
- * doing now have any benefit at all. Again, tests with this change have not
- * shown any improvement. Also, rte_mempool_get_bulk returns all or nothing
- * so looking at the implications of this type of allocation should be studied
- * deeply
- */
-
-static uint16_t
-nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
-{
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_rx_desc *rxds;
-	struct nfp_net_rx_buff *rxb;
-	struct nfp_net_hw *hw;
-	struct rte_mbuf *mb;
-	struct rte_mbuf *new_mb;
-	uint16_t nb_hold;
-	uint64_t dma_addr;
-	int avail;
-
-	rxq = rx_queue;
-	if (unlikely(rxq == NULL)) {
-		/*
-		 * DPDK just checks the queue is lower than max queues
-		 * enabled. But the queue needs to be configured
-		 */
-		RTE_LOG_DP(ERR, PMD, "RX Bad queue\n");
-		return -EINVAL;
-	}
-
-	hw = rxq->hw;
-	avail = 0;
-	nb_hold = 0;
-
-	while (avail < nb_pkts) {
-		rxb = &rxq->rxbufs[rxq->rd_p];
-		if (unlikely(rxb == NULL)) {
-			RTE_LOG_DP(ERR, PMD, "rxb does not exist!\n");
-			break;
-		}
-
-		rxds = &rxq->rxds[rxq->rd_p];
-		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
-			break;
-
-		/*
-		 * Memory barrier to ensure that we won't do other
-		 * reads before the DD bit.
-		 */
-		rte_rmb();
-
-		/*
-		 * We got a packet. Let's alloc a new mbuf for refilling the
-		 * free descriptor ring as soon as possible
-		 */
-		new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
-		if (unlikely(new_mb == NULL)) {
-			RTE_LOG_DP(DEBUG, PMD,
-			"RX mbuf alloc failed port_id=%u queue_id=%u\n",
-				rxq->port_id, (unsigned int)rxq->qidx);
-			nfp_net_mbuf_alloc_failed(rxq);
-			break;
-		}
-
-		nb_hold++;
-
-		/*
-		 * Grab the mbuf and refill the descriptor with the
-		 * previously allocated mbuf
-		 */
-		mb = rxb->mbuf;
-		rxb->mbuf = new_mb;
-
-		PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u",
-			   rxds->rxd.data_len, rxq->mbuf_size);
-
-		/* Size of this segment */
-		mb->data_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
-		/* Size of the whole packet. We just support 1 segment */
-		mb->pkt_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
-
-		if (unlikely((mb->data_len + hw->rx_offset) >
-			     rxq->mbuf_size)) {
-			/*
-			 * This should not happen and the user has the
-			 * responsibility of avoiding it. But we have
-			 * to give some info about the error
-			 */
-			RTE_LOG_DP(ERR, PMD,
-				"mbuf overflow likely due to the RX offset.\n"
-				"\t\tYour mbuf size should have extra space for"
-				" RX offset=%u bytes.\n"
-				"\t\tCurrently you just have %u bytes available"
-				" but the received packet is %u bytes long",
-				hw->rx_offset,
-				rxq->mbuf_size - hw->rx_offset,
-				mb->data_len);
-			return -EINVAL;
-		}
-
-		/* Filling the received mbuf with packet info */
-		if (hw->rx_offset)
-			mb->data_off = RTE_PKTMBUF_HEADROOM + hw->rx_offset;
-		else
-			mb->data_off = RTE_PKTMBUF_HEADROOM +
-				       NFP_DESC_META_LEN(rxds);
-
-		/* No scatter mode supported */
-		mb->nb_segs = 1;
-		mb->next = NULL;
-
-		mb->port = rxq->port_id;
-
-		/* Checking the RSS flag */
-		nfp_net_set_hash(rxq, rxds, mb);
-
-		/* Checking the checksum flag */
-		nfp_net_rx_cksum(rxq, rxds, mb);
-
-		if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
-		    (hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
-			mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
-			mb->ol_flags |= PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
-		}
-
-		/* Adding the mbuf to the mbuf array passed by the app */
-		rx_pkts[avail++] = mb;
-
-		/* Now resetting and updating the descriptor */
-		rxds->vals[0] = 0;
-		rxds->vals[1] = 0;
-		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(new_mb));
-		rxds->fld.dd = 0;
-		rxds->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
-		rxds->fld.dma_addr_lo = dma_addr & 0xffffffff;
-
-		rxq->rd_p++;
-		if (unlikely(rxq->rd_p == rxq->rx_count)) /* wrapping?*/
-			rxq->rd_p = 0;
-	}
-
-	if (nb_hold == 0)
-		return nb_hold;
-
-	PMD_RX_LOG(DEBUG, "RX  port_id=%u queue_id=%u, %d packets received",
-		   rxq->port_id, (unsigned int)rxq->qidx, nb_hold);
-
-	nb_hold += rxq->nb_rx_hold;
-
-	/*
-	 * FL descriptors needs to be written before incrementing the
-	 * FL queue WR pointer
-	 */
-	rte_wmb();
-	if (nb_hold > rxq->rx_free_thresh) {
-		PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u",
-			   rxq->port_id, (unsigned int)rxq->qidx,
-			   (unsigned)nb_hold, (unsigned)avail);
-		nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, nb_hold);
-		nb_hold = 0;
-	}
-	rxq->nb_rx_hold = nb_hold;
-
-	return avail;
-}
-
-/*
- * nfp_net_tx_free_bufs - Check for descriptors with a complete
- * status
- * @txq: TX queue to work with
- * Returns number of descriptors freed
- */
-int
-nfp_net_tx_free_bufs(struct nfp_net_txq *txq)
-{
-	uint32_t qcp_rd_p;
-	int todo;
-
-	PMD_TX_LOG(DEBUG, "queue %u. Check for descriptor with a complete"
-		   " status", txq->qidx);
-
-	/* Work out how many packets have been sent */
-	qcp_rd_p = nfp_qcp_read(txq->qcp_q, NFP_QCP_READ_PTR);
-
-	if (qcp_rd_p == txq->rd_p) {
-		PMD_TX_LOG(DEBUG, "queue %u: It seems harrier is not sending "
-			   "packets (%u, %u)", txq->qidx,
-			   qcp_rd_p, txq->rd_p);
-		return 0;
-	}
-
-	if (qcp_rd_p > txq->rd_p)
-		todo = qcp_rd_p - txq->rd_p;
-	else
-		todo = qcp_rd_p + txq->tx_count - txq->rd_p;
-
-	PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->rd_p: %u, qcp->rd_p: %u",
-		   qcp_rd_p, txq->rd_p, txq->rd_p);
-
-	if (todo == 0)
-		return todo;
-
-	txq->rd_p += todo;
-	if (unlikely(txq->rd_p >= txq->tx_count))
-		txq->rd_p -= txq->tx_count;
-
-	return todo;
-}
-
-/* Leaving always free descriptors for avoiding wrapping confusion */
-static inline
-uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq)
-{
-	if (txq->wr_p >= txq->rd_p)
-		return txq->tx_count - (txq->wr_p - txq->rd_p) - 8;
-	else
-		return txq->rd_p - txq->wr_p - 8;
-}
-
-/*
- * nfp_net_txq_full - Check if the TX queue free descriptors
- * is below tx_free_threshold
- *
- * @txq: TX queue to check
- *
- * This function uses the host copy* of read/write pointers
- */
-static inline
-uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
-{
-	return (nfp_free_tx_desc(txq) < txq->tx_free_thresh);
-}
-
-static uint16_t
-nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
-{
-	struct nfp_net_txq *txq;
-	struct nfp_net_hw *hw;
-	struct nfp_net_tx_desc *txds, txd;
-	struct rte_mbuf *pkt;
-	uint64_t dma_addr;
-	int pkt_size, dma_size;
-	uint16_t free_descs, issued_descs;
-	struct rte_mbuf **lmbuf;
-	int i;
-
-	txq = tx_queue;
-	hw = txq->hw;
-	txds = &txq->txds[txq->wr_p];
-
-	PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets",
-		   txq->qidx, txq->wr_p, nb_pkts);
-
-	if ((nfp_free_tx_desc(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
-		nfp_net_tx_free_bufs(txq);
-
-	free_descs = (uint16_t)nfp_free_tx_desc(txq);
-	if (unlikely(free_descs == 0))
-		return 0;
-
-	pkt = *tx_pkts;
-
-	i = 0;
-	issued_descs = 0;
-	PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets",
-		   txq->qidx, nb_pkts);
-	/* Sending packets */
-	while ((i < nb_pkts) && free_descs) {
-		/* Grabbing the mbuf linked to the current descriptor */
-		lmbuf = &txq->txbufs[txq->wr_p].mbuf;
-		/* Warming the cache for releasing the mbuf later on */
-		RTE_MBUF_PREFETCH_TO_FREE(*lmbuf);
-
-		pkt = *(tx_pkts + i);
-
-		if (unlikely((pkt->nb_segs > 1) &&
-			     !(hw->cap & NFP_NET_CFG_CTRL_GATHER))) {
-			PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set");
-			rte_panic("Multisegment packet unsupported\n");
-		}
-
-		/* Checking if we have enough descriptors */
-		if (unlikely(pkt->nb_segs > free_descs))
-			goto xmit_end;
-
-		/*
-		 * Checksum and VLAN flags just in the first descriptor for a
-		 * multisegment packet, but TSO info needs to be in all of them.
-		 */
-		txd.data_len = pkt->pkt_len;
-		nfp_net_tx_tso(txq, &txd, pkt);
-		nfp_net_tx_cksum(txq, &txd, pkt);
-
-		if ((pkt->ol_flags & PKT_TX_VLAN_PKT) &&
-		    (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)) {
-			txd.flags |= PCIE_DESC_TX_VLAN;
-			txd.vlan = pkt->vlan_tci;
-		}
-
-		/*
-		 * mbuf data_len is the data in one segment and pkt_len data
-		 * in the whole packet. When the packet is just one segment,
-		 * then data_len = pkt_len
-		 */
-		pkt_size = pkt->pkt_len;
-
-		while (pkt) {
-			/* Copying TSO, VLAN and cksum info */
-			*txds = txd;
-
-			/* Releasing mbuf used by this descriptor previously*/
-			if (*lmbuf)
-				rte_pktmbuf_free_seg(*lmbuf);
-
-			/*
-			 * Linking mbuf with descriptor for being released
-			 * next time descriptor is used
-			 */
-			*lmbuf = pkt;
-
-			dma_size = pkt->data_len;
-			dma_addr = rte_mbuf_data_iova(pkt);
-			PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
-				   "%" PRIx64 "", dma_addr);
-
-			/* Filling descriptors fields */
-			txds->dma_len = dma_size;
-			txds->data_len = txd.data_len;
-			txds->dma_addr_hi = (dma_addr >> 32) & 0xff;
-			txds->dma_addr_lo = (dma_addr & 0xffffffff);
-			ASSERT(free_descs > 0);
-			free_descs--;
-
-			txq->wr_p++;
-			if (unlikely(txq->wr_p == txq->tx_count)) /* wrapping?*/
-				txq->wr_p = 0;
-
-			pkt_size -= dma_size;
-
-			/*
-			 * Making the EOP, packets with just one segment
-			 * the priority
-			 */
-			if (likely(!pkt_size))
-				txds->offset_eop = PCIE_DESC_TX_EOP;
-			else
-				txds->offset_eop = 0;
-
-			pkt = pkt->next;
-			/* Referencing next free TX descriptor */
-			txds = &txq->txds[txq->wr_p];
-			lmbuf = &txq->txbufs[txq->wr_p].mbuf;
-			issued_descs++;
-		}
-		i++;
-	}
-
-xmit_end:
-	/* Increment write pointers. Force memory write before we let HW know */
-	rte_wmb();
-	nfp_qcp_ptr_add(txq->qcp_q, NFP_QCP_WRITE_PTR, issued_descs);
-
-	return i;
-}
-
 static int
 nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 {
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index a3a3ba32d6..9265496bf0 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -41,6 +41,12 @@ struct nfp_net_adapter;
 #define NFP_QCP_QUEUE_STS_HI                    0x000c
 #define NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask    (0x3ffff)
 
+/* The offset of the queue controller queues in the PCIe Target */
+#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
+
+/* Maximum value which can be added to a queue with one transaction */
+#define NFP_QCP_MAX_ADD	0x7f
+
 /* Interrupt definitions */
 #define NFP_NET_IRQ_LSC_IDX             0
 
@@ -95,47 +101,11 @@ struct nfp_net_adapter;
 #include <linux/types.h>
 #include <rte_io.h>
 
-static inline uint8_t nn_readb(volatile const void *addr)
-{
-	return rte_read8(addr);
-}
-
-static inline void nn_writeb(uint8_t val, volatile void *addr)
-{
-	rte_write8(val, addr);
-}
-
-static inline uint32_t nn_readl(volatile const void *addr)
-{
-	return rte_read32(addr);
-}
-
-static inline void nn_writel(uint32_t val, volatile void *addr)
-{
-	rte_write32(val, addr);
-}
-
-static inline void nn_writew(uint16_t val, volatile void *addr)
-{
-	rte_write16(val, addr);
-}
-
-static inline uint64_t nn_readq(volatile void *addr)
-{
-	const volatile uint32_t *p = addr;
-	uint32_t low, high;
-
-	high = nn_readl((volatile const void *)(p + 1));
-	low = nn_readl((volatile const void *)p);
-
-	return low + ((uint64_t)high << 32);
-}
-
-static inline void nn_writeq(uint64_t val, volatile void *addr)
-{
-	nn_writel(val >> 32, (volatile char *)addr + 4);
-	nn_writel(val, addr);
-}
+/* nfp_qcp_ptr - Read or Write Pointer of a queue */
+enum nfp_qcp_ptr {
+	NFP_QCP_READ_PTR = 0,
+	NFP_QCP_WRITE_PTR
+};
 
 struct nfp_pf_dev {
 	/* Backpointer to associated pci device */
@@ -247,6 +217,138 @@ struct nfp_net_adapter {
 	struct nfp_net_hw hw;
 };
 
+static inline uint8_t nn_readb(volatile const void *addr)
+{
+	return rte_read8(addr);
+}
+
+static inline void nn_writeb(uint8_t val, volatile void *addr)
+{
+	rte_write8(val, addr);
+}
+
+static inline uint32_t nn_readl(volatile const void *addr)
+{
+	return rte_read32(addr);
+}
+
+static inline void nn_writel(uint32_t val, volatile void *addr)
+{
+	rte_write32(val, addr);
+}
+
+static inline void nn_writew(uint16_t val, volatile void *addr)
+{
+	rte_write16(val, addr);
+}
+
+static inline uint64_t nn_readq(volatile void *addr)
+{
+	const volatile uint32_t *p = addr;
+	uint32_t low, high;
+
+	high = nn_readl((volatile const void *)(p + 1));
+	low = nn_readl((volatile const void *)p);
+
+	return low + ((uint64_t)high << 32);
+}
+
+static inline void nn_writeq(uint64_t val, volatile void *addr)
+{
+	nn_writel(val >> 32, (volatile char *)addr + 4);
+	nn_writel(val, addr);
+}
+
+/*
+ * Functions to read/write from/to Config BAR
+ * Performs any endian conversion necessary.
+ */
+static inline uint8_t
+nn_cfg_readb(struct nfp_net_hw *hw, int off)
+{
+	return nn_readb(hw->ctrl_bar + off);
+}
+
+static inline void
+nn_cfg_writeb(struct nfp_net_hw *hw, int off, uint8_t val)
+{
+	nn_writeb(val, hw->ctrl_bar + off);
+}
+
+static inline uint32_t
+nn_cfg_readl(struct nfp_net_hw *hw, int off)
+{
+	return rte_le_to_cpu_32(nn_readl(hw->ctrl_bar + off));
+}
+
+static inline void
+nn_cfg_writel(struct nfp_net_hw *hw, int off, uint32_t val)
+{
+	nn_writel(rte_cpu_to_le_32(val), hw->ctrl_bar + off);
+}
+
+static inline uint64_t
+nn_cfg_readq(struct nfp_net_hw *hw, int off)
+{
+	return rte_le_to_cpu_64(nn_readq(hw->ctrl_bar + off));
+}
+
+static inline void
+nn_cfg_writeq(struct nfp_net_hw *hw, int off, uint64_t val)
+{
+	nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
+}
+
+/*
+ * nfp_qcp_ptr_add - Add the value to the selected pointer of a queue
+ * @q: Base address for queue structure
+ * @ptr: Add to the Read or Write pointer
+ * @val: Value to add to the queue pointer
+ *
+ * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed.
+ */
+static inline void
+nfp_qcp_ptr_add(uint8_t *q, enum nfp_qcp_ptr ptr, uint32_t val)
+{
+	uint32_t off;
+
+	if (ptr == NFP_QCP_READ_PTR)
+		off = NFP_QCP_QUEUE_ADD_RPTR;
+	else
+		off = NFP_QCP_QUEUE_ADD_WPTR;
+
+	while (val > NFP_QCP_MAX_ADD) {
+		nn_writel(rte_cpu_to_le_32(NFP_QCP_MAX_ADD), q + off);
+		val -= NFP_QCP_MAX_ADD;
+}
+
+nn_writel(rte_cpu_to_le_32(val), q + off);
+}
+
+/*
+ * nfp_qcp_read - Read the current Read/Write pointer value for a queue
+ * @q:  Base address for queue structure
+ * @ptr: Read or Write pointer
+ */
+static inline uint32_t
+nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
+{
+	uint32_t off;
+	uint32_t val;
+
+	if (ptr == NFP_QCP_READ_PTR)
+		off = NFP_QCP_QUEUE_STS_LO;
+	else
+		off = NFP_QCP_QUEUE_STS_HI;
+
+	val = rte_cpu_to_le_32(nn_readl(q + off));
+
+	if (ptr == NFP_QCP_READ_PTR)
+		return val & NFP_QCP_QUEUE_STS_LO_READPTR_mask;
+	else
+		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
+}
+
 #define NFP_NET_DEV_PRIVATE_TO_HW(adapter)\
 	(&((struct nfp_net_adapter *)adapter)->hw)
 
diff --git a/drivers/net/nfp/nfp_rxtx.c b/drivers/net/nfp/nfp_rxtx.c
new file mode 100644
index 0000000000..9ee9e5c9a3
--- /dev/null
+++ b/drivers/net/nfp/nfp_rxtx.c
@@ -0,0 +1,1002 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_rxtx.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: Rx/Tx functions
+ */
+
+#include <ethdev_driver.h>
+#include <ethdev_pci.h>
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+
+/* Prototypes */
+static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
+static inline void nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq);
+static inline void nfp_net_set_hash(struct nfp_net_rxq *rxq,
+				    struct nfp_net_rx_desc *rxd,
+				    struct rte_mbuf *mbuf);
+static inline void nfp_net_rx_cksum(struct nfp_net_rxq *rxq,
+				    struct nfp_net_rx_desc *rxd,
+				    struct rte_mbuf *mb);
+static void nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq);
+static int nfp_net_tx_free_bufs(struct nfp_net_txq *txq);
+static void nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq);
+static inline uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq);
+static inline uint32_t nfp_net_txq_full(struct nfp_net_txq *txq);
+static inline void nfp_net_tx_tso(struct nfp_net_txq *txq,
+				  struct nfp_net_tx_desc *txd,
+				  struct rte_mbuf *mb);
+static inline void nfp_net_tx_cksum(struct nfp_net_txq *txq,
+				    struct nfp_net_tx_desc *txd,
+				    struct rte_mbuf *mb);
+
+static int
+nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq)
+{
+	struct nfp_net_rx_buff *rxe = rxq->rxbufs;
+	uint64_t dma_addr;
+	unsigned int i;
+
+	PMD_RX_LOG(DEBUG, "Fill Rx Freelist for %u descriptors",
+		   rxq->rx_count);
+
+	for (i = 0; i < rxq->rx_count; i++) {
+		struct nfp_net_rx_desc *rxd;
+		struct rte_mbuf *mbuf = rte_pktmbuf_alloc(rxq->mem_pool);
+
+		if (mbuf == NULL) {
+			PMD_DRV_LOG(ERR, "RX mbuf alloc failed queue_id=%u",
+				(unsigned int)rxq->qidx);
+			return -ENOMEM;
+		}
+
+		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(mbuf));
+
+		rxd = &rxq->rxds[i];
+		rxd->fld.dd = 0;
+		rxd->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
+		rxd->fld.dma_addr_lo = dma_addr & 0xffffffff;
+		rxe[i].mbuf = mbuf;
+		PMD_RX_LOG(DEBUG, "[%d]: %" PRIx64, i, dma_addr);
+	}
+
+	/* Make sure all writes are flushed before telling the hardware */
+	rte_wmb();
+
+	/* Not advertising the whole ring as the firmware gets confused if so */
+	PMD_RX_LOG(DEBUG, "Increment FL write pointer in %u",
+		   rxq->rx_count - 1);
+
+	nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, rxq->rx_count - 1);
+
+	return 0;
+}
+
+int
+nfp_net_rx_freelist_setup(struct rte_eth_dev *dev)
+{
+	int i;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		if (nfp_net_rx_fill_freelist(dev->data->rx_queues[i]) < 0)
+			return -1;
+	}
+	return 0;
+}
+
+uint32_t
+nfp_net_rx_queue_count(struct rte_eth_dev *dev, uint16_t queue_idx)
+{
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_rx_desc *rxds;
+	uint32_t idx;
+	uint32_t count;
+
+	rxq = (struct nfp_net_rxq *)dev->data->rx_queues[queue_idx];
+
+	idx = rxq->rd_p;
+
+	count = 0;
+
+	/*
+	 * Other PMDs are just checking the DD bit in intervals of 4
+	 * descriptors and counting all four if the first has the DD
+	 * bit on. Of course, this is not accurate but can be good for
+	 * performance. But ideally that should be done in descriptors
+	 * chunks belonging to the same cache line
+	 */
+
+	while (count < rxq->rx_count) {
+		rxds = &rxq->rxds[idx];
+		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
+			break;
+
+		count++;
+		idx++;
+
+		/* Wrapping? */
+		if ((idx) == rxq->rx_count)
+			idx = 0;
+	}
+
+	return count;
+}
+
+static inline void
+nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq)
+{
+	rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
+}
+
+/*
+ * nfp_net_set_hash - Set mbuf hash data
+ *
+ * The RSS hash and hash-type are pre-pended to the packet data.
+ * Extract and decode it and set the mbuf fields.
+ */
+static inline void
+nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
+		 struct rte_mbuf *mbuf)
+{
+	struct nfp_net_hw *hw = rxq->hw;
+	uint8_t *meta_offset;
+	uint32_t meta_info;
+	uint32_t hash = 0;
+	uint32_t hash_type = 0;
+
+	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
+		return;
+
+	/* this is true for new firmwares */
+	if (likely(((hw->cap & NFP_NET_CFG_CTRL_RSS2) ||
+	    (NFD_CFG_MAJOR_VERSION_of(hw->ver) == 4)) &&
+	     NFP_DESC_META_LEN(rxd))) {
+		/*
+		 * new metadata api:
+		 * <----  32 bit  ----->
+		 * m    field type word
+		 * e     data field #2
+		 * t     data field #1
+		 * a     data field #0
+		 * ====================
+		 *    packet data
+		 *
+		 * Field type word contains up to 8 4bit field types
+		 * A 4bit field type refers to a data field word
+		 * A data field word can have several 4bit field types
+		 */
+		meta_offset = rte_pktmbuf_mtod(mbuf, uint8_t *);
+		meta_offset -= NFP_DESC_META_LEN(rxd);
+		meta_info = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+		meta_offset += 4;
+		/* NFP PMD just supports metadata for hashing */
+		switch (meta_info & NFP_NET_META_FIELD_MASK) {
+		case NFP_NET_META_HASH:
+			/* next field type is about the hash type */
+			meta_info >>= NFP_NET_META_FIELD_SIZE;
+			/* hash value is in the data field */
+			hash = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+			hash_type = meta_info & NFP_NET_META_FIELD_MASK;
+			break;
+		default:
+			/* Unsupported metadata can be a performance issue */
+			return;
+		}
+	} else {
+		if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
+			return;
+
+		hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
+		hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
+	}
+
+	mbuf->hash.rss = hash;
+	mbuf->ol_flags |= PKT_RX_RSS_HASH;
+
+	switch (hash_type) {
+	case NFP_NET_RSS_IPV4:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV4;
+		break;
+	case NFP_NET_RSS_IPV6:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6;
+		break;
+	case NFP_NET_RSS_IPV6_EX:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV4_TCP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV6_TCP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV4_UDP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV6_UDP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	default:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L4_MASK;
+	}
+}
+
+/* nfp_net_rx_cksum - set mbuf checksum flags based on RX descriptor flags */
+static inline void
+nfp_net_rx_cksum(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
+		 struct rte_mbuf *mb)
+{
+	struct nfp_net_hw *hw = rxq->hw;
+
+	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RXCSUM))
+		return;
+
+	/* If IPv4 and IP checksum error, fail */
+	if (unlikely((rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM) &&
+	    !(rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM_OK)))
+		mb->ol_flags |= PKT_RX_IP_CKSUM_BAD;
+	else
+		mb->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+
+	/* If neither UDP nor TCP return */
+	if (!(rxd->rxd.flags & PCIE_DESC_RX_TCP_CSUM) &&
+	    !(rxd->rxd.flags & PCIE_DESC_RX_UDP_CSUM))
+		return;
+
+	if (likely(rxd->rxd.flags & PCIE_DESC_RX_L4_CSUM_OK))
+		mb->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
+	else
+		mb->ol_flags |= PKT_RX_L4_CKSUM_BAD;
+}
+
+/*
+ * RX path design:
+ *
+ * There are some decisions to take:
+ * 1) How to check DD RX descriptors bit
+ * 2) How and when to allocate new mbufs
+ *
+ * Current implementation checks just one single DD bit each loop. As each
+ * descriptor is 8 bytes, it is likely a good idea to check descriptors in
+ * a single cache line instead. Tests with this change have not shown any
+ * performance improvement but it requires further investigation. For example,
+ * depending on which descriptor is next, the number of descriptors could be
+ * less than 8 for just checking those in the same cache line. This implies
+ * extra work which could be counterproductive by itself. Indeed, last firmware
+ * changes are just doing this: writing several descriptors with the DD bit
+ * for saving PCIe bandwidth and DMA operations from the NFP.
+ *
+ * Mbuf allocation is done when a new packet is received. Then the descriptor
+ * is automatically linked with the new mbuf and the old one is given to the
+ * user. The main drawback with this design is mbuf allocation is heavier than
+ * using bulk allocations allowed by DPDK with rte_mempool_get_bulk. From the
+ * cache point of view it does not seem allocating the mbuf early on as we are
+ * doing now have any benefit at all. Again, tests with this change have not
+ * shown any improvement. Also, rte_mempool_get_bulk returns all or nothing
+ * so looking at the implications of this type of allocation should be studied
+ * deeply
+ */
+
+uint16_t
+nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_rx_desc *rxds;
+	struct nfp_net_rx_buff *rxb;
+	struct nfp_net_hw *hw;
+	struct rte_mbuf *mb;
+	struct rte_mbuf *new_mb;
+	uint16_t nb_hold;
+	uint64_t dma_addr;
+	int avail;
+
+	rxq = rx_queue;
+	if (unlikely(rxq == NULL)) {
+		/*
+		 * DPDK just checks the queue is lower than max queues
+		 * enabled. But the queue needs to be configured
+		 */
+		RTE_LOG_DP(ERR, PMD, "RX Bad queue\n");
+		return -EINVAL;
+	}
+
+	hw = rxq->hw;
+	avail = 0;
+	nb_hold = 0;
+
+	while (avail < nb_pkts) {
+		rxb = &rxq->rxbufs[rxq->rd_p];
+		if (unlikely(rxb == NULL)) {
+			RTE_LOG_DP(ERR, PMD, "rxb does not exist!\n");
+			break;
+		}
+
+		rxds = &rxq->rxds[rxq->rd_p];
+		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
+			break;
+
+		/*
+		 * Memory barrier to ensure that we won't do other
+		 * reads before the DD bit.
+		 */
+		rte_rmb();
+
+		/*
+		 * We got a packet. Let's alloc a new mbuf for refilling the
+		 * free descriptor ring as soon as possible
+		 */
+		new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
+		if (unlikely(new_mb == NULL)) {
+			RTE_LOG_DP(DEBUG, PMD,
+			"RX mbuf alloc failed port_id=%u queue_id=%u\n",
+				rxq->port_id, (unsigned int)rxq->qidx);
+			nfp_net_mbuf_alloc_failed(rxq);
+			break;
+		}
+
+		nb_hold++;
+
+		/*
+		 * Grab the mbuf and refill the descriptor with the
+		 * previously allocated mbuf
+		 */
+		mb = rxb->mbuf;
+		rxb->mbuf = new_mb;
+
+		PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u",
+			   rxds->rxd.data_len, rxq->mbuf_size);
+
+		/* Size of this segment */
+		mb->data_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
+		/* Size of the whole packet. We just support 1 segment */
+		mb->pkt_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
+
+		if (unlikely((mb->data_len + hw->rx_offset) >
+			     rxq->mbuf_size)) {
+			/*
+			 * This should not happen and the user has the
+			 * responsibility of avoiding it. But we have
+			 * to give some info about the error
+			 */
+			RTE_LOG_DP(ERR, PMD,
+				"mbuf overflow likely due to the RX offset.\n"
+				"\t\tYour mbuf size should have extra space for"
+				" RX offset=%u bytes.\n"
+				"\t\tCurrently you just have %u bytes available"
+				" but the received packet is %u bytes long",
+				hw->rx_offset,
+				rxq->mbuf_size - hw->rx_offset,
+				mb->data_len);
+			return -EINVAL;
+		}
+
+		/* Filling the received mbuf with packet info */
+		if (hw->rx_offset)
+			mb->data_off = RTE_PKTMBUF_HEADROOM + hw->rx_offset;
+		else
+			mb->data_off = RTE_PKTMBUF_HEADROOM +
+				       NFP_DESC_META_LEN(rxds);
+
+		/* No scatter mode supported */
+		mb->nb_segs = 1;
+		mb->next = NULL;
+
+		mb->port = rxq->port_id;
+
+		/* Checking the RSS flag */
+		nfp_net_set_hash(rxq, rxds, mb);
+
+		/* Checking the checksum flag */
+		nfp_net_rx_cksum(rxq, rxds, mb);
+
+		if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
+		    (hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
+			mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
+			mb->ol_flags |= PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
+		}
+
+		/* Adding the mbuf to the mbuf array passed by the app */
+		rx_pkts[avail++] = mb;
+
+		/* Now resetting and updating the descriptor */
+		rxds->vals[0] = 0;
+		rxds->vals[1] = 0;
+		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(new_mb));
+		rxds->fld.dd = 0;
+		rxds->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
+		rxds->fld.dma_addr_lo = dma_addr & 0xffffffff;
+
+		rxq->rd_p++;
+		if (unlikely(rxq->rd_p == rxq->rx_count)) /* wrapping?*/
+			rxq->rd_p = 0;
+	}
+
+	if (nb_hold == 0)
+		return nb_hold;
+
+	PMD_RX_LOG(DEBUG, "RX  port_id=%u queue_id=%u, %d packets received",
+		   rxq->port_id, (unsigned int)rxq->qidx, nb_hold);
+
+	nb_hold += rxq->nb_rx_hold;
+
+	/*
+	 * FL descriptors needs to be written before incrementing the
+	 * FL queue WR pointer
+	 */
+	rte_wmb();
+	if (nb_hold > rxq->rx_free_thresh) {
+		PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u",
+			   rxq->port_id, (unsigned int)rxq->qidx,
+			   (unsigned int)nb_hold, (unsigned int)avail);
+		nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, nb_hold);
+		nb_hold = 0;
+	}
+	rxq->nb_rx_hold = nb_hold;
+
+	return avail;
+}
+
+static void
+nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
+{
+	unsigned int i;
+
+	if (rxq->rxbufs == NULL)
+		return;
+
+	for (i = 0; i < rxq->rx_count; i++) {
+		if (rxq->rxbufs[i].mbuf) {
+			rte_pktmbuf_free_seg(rxq->rxbufs[i].mbuf);
+			rxq->rxbufs[i].mbuf = NULL;
+		}
+	}
+}
+
+void
+nfp_net_rx_queue_release(void *rx_queue)
+{
+	struct nfp_net_rxq *rxq = rx_queue;
+
+	if (rxq) {
+		nfp_net_rx_queue_release_mbufs(rxq);
+		rte_free(rxq->rxbufs);
+		rte_free(rxq);
+	}
+}
+
+void
+nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
+{
+	nfp_net_rx_queue_release_mbufs(rxq);
+	rxq->rd_p = 0;
+	rxq->nb_rx_hold = 0;
+}
+
+int
+nfp_net_rx_queue_setup(struct rte_eth_dev *dev,
+		       uint16_t queue_idx, uint16_t nb_desc,
+		       unsigned int socket_id,
+		       const struct rte_eth_rxconf *rx_conf,
+		       struct rte_mempool *mp)
+{
+	const struct rte_memzone *tz;
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_hw *hw;
+	uint32_t rx_desc_sz;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* Validating number of descriptors */
+	rx_desc_sz = nb_desc * sizeof(struct nfp_net_rx_desc);
+	if (rx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
+	    nb_desc > NFP_NET_MAX_RX_DESC ||
+	    nb_desc < NFP_NET_MIN_RX_DESC) {
+		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
+		return -EINVAL;
+	}
+
+	/*
+	 * Free memory prior to re-allocation if needed. This is the case after
+	 * calling nfp_net_stop
+	 */
+	if (dev->data->rx_queues[queue_idx]) {
+		nfp_net_rx_queue_release(dev->data->rx_queues[queue_idx]);
+		dev->data->rx_queues[queue_idx] = NULL;
+	}
+
+	/* Allocating rx queue data structure */
+	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct nfp_net_rxq),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxq == NULL)
+		return -ENOMEM;
+
+	/* Hw queues mapping based on firmware configuration */
+	rxq->qidx = queue_idx;
+	rxq->fl_qcidx = queue_idx * hw->stride_rx;
+	rxq->rx_qcidx = rxq->fl_qcidx + (hw->stride_rx - 1);
+	rxq->qcp_fl = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->fl_qcidx);
+	rxq->qcp_rx = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->rx_qcidx);
+
+	/*
+	 * Tracking mbuf size for detecting a potential mbuf overflow due to
+	 * RX offset
+	 */
+	rxq->mem_pool = mp;
+	rxq->mbuf_size = rxq->mem_pool->elt_size;
+	rxq->mbuf_size -= (sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM);
+	hw->flbufsz = rxq->mbuf_size;
+
+	rxq->rx_count = nb_desc;
+	rxq->port_id = dev->data->port_id;
+	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
+	rxq->drop_en = rx_conf->rx_drop_en;
+
+	/*
+	 * Allocate RX ring hardware descriptors. A memzone large enough to
+	 * handle the maximum ring size is allocated in order to allow for
+	 * resizing in later calls to the queue setup function.
+	 */
+	tz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
+				   sizeof(struct nfp_net_rx_desc) *
+				   NFP_NET_MAX_RX_DESC, NFP_MEMZONE_ALIGN,
+				   socket_id);
+
+	if (tz == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating rx dma");
+		nfp_net_rx_queue_release(rxq);
+		return -ENOMEM;
+	}
+
+	/* Saving physical and virtual addresses for the RX ring */
+	rxq->dma = (uint64_t)tz->iova;
+	rxq->rxds = (struct nfp_net_rx_desc *)tz->addr;
+
+	/* mbuf pointers array for referencing mbufs linked to RX descriptors */
+	rxq->rxbufs = rte_zmalloc_socket("rxq->rxbufs",
+					 sizeof(*rxq->rxbufs) * nb_desc,
+					 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxq->rxbufs == NULL) {
+		nfp_net_rx_queue_release(rxq);
+		return -ENOMEM;
+	}
+
+	PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
+		   rxq->rxbufs, rxq->rxds, (unsigned long)rxq->dma);
+
+	nfp_net_reset_rx_queue(rxq);
+
+	dev->data->rx_queues[queue_idx] = rxq;
+	rxq->hw = hw;
+
+	/*
+	 * Telling the HW about the physical address of the RX ring and number
+	 * of descriptors in log2 format
+	 */
+	nn_cfg_writeq(hw, NFP_NET_CFG_RXR_ADDR(queue_idx), rxq->dma);
+	nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), rte_log2_u32(nb_desc));
+
+	return 0;
+}
+
+/*
+ * nfp_net_tx_free_bufs - Check for descriptors with a complete
+ * status
+ * @txq: TX queue to work with
+ * Returns number of descriptors freed
+ */
+static int
+nfp_net_tx_free_bufs(struct nfp_net_txq *txq)
+{
+	uint32_t qcp_rd_p;
+	int todo;
+
+	PMD_TX_LOG(DEBUG, "queue %u. Check for descriptor with a complete"
+		   " status", txq->qidx);
+
+	/* Work out how many packets have been sent */
+	qcp_rd_p = nfp_qcp_read(txq->qcp_q, NFP_QCP_READ_PTR);
+
+	if (qcp_rd_p == txq->rd_p) {
+		PMD_TX_LOG(DEBUG, "queue %u: It seems harrier is not sending "
+			   "packets (%u, %u)", txq->qidx,
+			   qcp_rd_p, txq->rd_p);
+		return 0;
+	}
+
+	if (qcp_rd_p > txq->rd_p)
+		todo = qcp_rd_p - txq->rd_p;
+	else
+		todo = qcp_rd_p + txq->tx_count - txq->rd_p;
+
+	PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->rd_p: %u, qcp->rd_p: %u",
+		   qcp_rd_p, txq->rd_p, txq->rd_p);
+
+	if (todo == 0)
+		return todo;
+
+	txq->rd_p += todo;
+	if (unlikely(txq->rd_p >= txq->tx_count))
+		txq->rd_p -= txq->tx_count;
+
+	return todo;
+}
+
+static void
+nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq)
+{
+	unsigned int i;
+
+	if (txq->txbufs == NULL)
+		return;
+
+	for (i = 0; i < txq->tx_count; i++) {
+		if (txq->txbufs[i].mbuf) {
+			rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
+			txq->txbufs[i].mbuf = NULL;
+		}
+	}
+}
+
+void
+nfp_net_tx_queue_release(void *tx_queue)
+{
+	struct nfp_net_txq *txq = tx_queue;
+
+	if (txq) {
+		nfp_net_tx_queue_release_mbufs(txq);
+		rte_free(txq->txbufs);
+		rte_free(txq);
+	}
+}
+
+void
+nfp_net_reset_tx_queue(struct nfp_net_txq *txq)
+{
+	nfp_net_tx_queue_release_mbufs(txq);
+	txq->wr_p = 0;
+	txq->rd_p = 0;
+}
+
+int
+nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+		       uint16_t nb_desc, unsigned int socket_id,
+		       const struct rte_eth_txconf *tx_conf)
+{
+	const struct rte_memzone *tz;
+	struct nfp_net_txq *txq;
+	uint16_t tx_free_thresh;
+	struct nfp_net_hw *hw;
+	uint32_t tx_desc_sz;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* Validating number of descriptors */
+	tx_desc_sz = nb_desc * sizeof(struct nfp_net_tx_desc);
+	if (tx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
+	    nb_desc > NFP_NET_MAX_TX_DESC ||
+	    nb_desc < NFP_NET_MIN_TX_DESC) {
+		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
+		return -EINVAL;
+	}
+
+	tx_free_thresh = (uint16_t)((tx_conf->tx_free_thresh) ?
+				    tx_conf->tx_free_thresh :
+				    DEFAULT_TX_FREE_THRESH);
+
+	if (tx_free_thresh > (nb_desc)) {
+		PMD_DRV_LOG(ERR,
+			"tx_free_thresh must be less than the number of TX "
+			"descriptors. (tx_free_thresh=%u port=%d "
+			"queue=%d)", (unsigned int)tx_free_thresh,
+			dev->data->port_id, (int)queue_idx);
+		return -(EINVAL);
+	}
+
+	/*
+	 * Free memory prior to re-allocation if needed. This is the case after
+	 * calling nfp_net_stop
+	 */
+	if (dev->data->tx_queues[queue_idx]) {
+		PMD_TX_LOG(DEBUG, "Freeing memory prior to re-allocation %d",
+			   queue_idx);
+		nfp_net_tx_queue_release(dev->data->tx_queues[queue_idx]);
+		dev->data->tx_queues[queue_idx] = NULL;
+	}
+
+	/* Allocating tx queue data structure */
+	txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct nfp_net_txq),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating tx dma");
+		return -ENOMEM;
+	}
+
+	/*
+	 * Allocate TX ring hardware descriptors. A memzone large enough to
+	 * handle the maximum ring size is allocated in order to allow for
+	 * resizing in later calls to the queue setup function.
+	 */
+	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
+				   sizeof(struct nfp_net_tx_desc) *
+				   NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN,
+				   socket_id);
+	if (tz == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating tx dma");
+		nfp_net_tx_queue_release(txq);
+		return -ENOMEM;
+	}
+
+	txq->tx_count = nb_desc;
+	txq->tx_free_thresh = tx_free_thresh;
+	txq->tx_pthresh = tx_conf->tx_thresh.pthresh;
+	txq->tx_hthresh = tx_conf->tx_thresh.hthresh;
+	txq->tx_wthresh = tx_conf->tx_thresh.wthresh;
+
+	/* queue mapping based on firmware configuration */
+	txq->qidx = queue_idx;
+	txq->tx_qcidx = queue_idx * hw->stride_tx;
+	txq->qcp_q = hw->tx_bar + NFP_QCP_QUEUE_OFF(txq->tx_qcidx);
+
+	txq->port_id = dev->data->port_id;
+
+	/* Saving physical and virtual addresses for the TX ring */
+	txq->dma = (uint64_t)tz->iova;
+	txq->txds = (struct nfp_net_tx_desc *)tz->addr;
+
+	/* mbuf pointers array for referencing mbufs linked to TX descriptors */
+	txq->txbufs = rte_zmalloc_socket("txq->txbufs",
+					 sizeof(*txq->txbufs) * nb_desc,
+					 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq->txbufs == NULL) {
+		nfp_net_tx_queue_release(txq);
+		return -ENOMEM;
+	}
+	PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
+		   txq->txbufs, txq->txds, (unsigned long)txq->dma);
+
+	nfp_net_reset_tx_queue(txq);
+
+	dev->data->tx_queues[queue_idx] = txq;
+	txq->hw = hw;
+
+	/*
+	 * Telling the HW about the physical address of the TX ring and number
+	 * of descriptors in log2 format
+	 */
+	nn_cfg_writeq(hw, NFP_NET_CFG_TXR_ADDR(queue_idx), txq->dma);
+	nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), rte_log2_u32(nb_desc));
+
+	return 0;
+}
+
+/* Leaving always free descriptors for avoiding wrapping confusion */
+static inline
+uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq)
+{
+	if (txq->wr_p >= txq->rd_p)
+		return txq->tx_count - (txq->wr_p - txq->rd_p) - 8;
+	else
+		return txq->rd_p - txq->wr_p - 8;
+}
+
+/*
+ * nfp_net_txq_full - Check if the TX queue free descriptors
+ * is below tx_free_threshold
+ *
+ * @txq: TX queue to check
+ *
+ * This function uses the host copy* of read/write pointers
+ */
+static inline
+uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
+{
+	return (nfp_free_tx_desc(txq) < txq->tx_free_thresh);
+}
+
+/* nfp_net_tx_tso - Set TX descriptor for TSO */
+static inline void
+nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+	       struct rte_mbuf *mb)
+{
+	uint64_t ol_flags;
+	struct nfp_net_hw *hw = txq->hw;
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY))
+		goto clean_txd;
+
+	ol_flags = mb->ol_flags;
+
+	if (!(ol_flags & PKT_TX_TCP_SEG))
+		goto clean_txd;
+
+	txd->l3_offset = mb->l2_len;
+	txd->l4_offset = mb->l2_len + mb->l3_len;
+	txd->lso_hdrlen = mb->l2_len + mb->l3_len + mb->l4_len;
+	txd->mss = rte_cpu_to_le_16(mb->tso_segsz);
+	txd->flags = PCIE_DESC_TX_LSO;
+	return;
+
+clean_txd:
+	txd->flags = 0;
+	txd->l3_offset = 0;
+	txd->l4_offset = 0;
+	txd->lso_hdrlen = 0;
+	txd->mss = 0;
+}
+
+/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
+static inline void
+nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+		 struct rte_mbuf *mb)
+{
+	uint64_t ol_flags;
+	struct nfp_net_hw *hw = txq->hw;
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_TXCSUM))
+		return;
+
+	ol_flags = mb->ol_flags;
+
+	/* IPv6 does not need checksum */
+	if (ol_flags & PKT_TX_IP_CKSUM)
+		txd->flags |= PCIE_DESC_TX_IP4_CSUM;
+
+	switch (ol_flags & PKT_TX_L4_MASK) {
+	case PKT_TX_UDP_CKSUM:
+		txd->flags |= PCIE_DESC_TX_UDP_CSUM;
+		break;
+	case PKT_TX_TCP_CKSUM:
+		txd->flags |= PCIE_DESC_TX_TCP_CSUM;
+		break;
+	}
+
+	if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
+		txd->flags |= PCIE_DESC_TX_CSUM;
+}
+
+uint16_t
+nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+	struct nfp_net_txq *txq;
+	struct nfp_net_hw *hw;
+	struct nfp_net_tx_desc *txds, txd;
+	struct rte_mbuf *pkt;
+	uint64_t dma_addr;
+	int pkt_size, dma_size;
+	uint16_t free_descs, issued_descs;
+	struct rte_mbuf **lmbuf;
+	int i;
+
+	txq = tx_queue;
+	hw = txq->hw;
+	txds = &txq->txds[txq->wr_p];
+
+	PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets",
+		   txq->qidx, txq->wr_p, nb_pkts);
+
+	if ((nfp_free_tx_desc(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
+		nfp_net_tx_free_bufs(txq);
+
+	free_descs = (uint16_t)nfp_free_tx_desc(txq);
+	if (unlikely(free_descs == 0))
+		return 0;
+
+	pkt = *tx_pkts;
+
+	i = 0;
+	issued_descs = 0;
+	PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets",
+		   txq->qidx, nb_pkts);
+	/* Sending packets */
+	while ((i < nb_pkts) && free_descs) {
+		/* Grabbing the mbuf linked to the current descriptor */
+		lmbuf = &txq->txbufs[txq->wr_p].mbuf;
+		/* Warming the cache for releasing the mbuf later on */
+		RTE_MBUF_PREFETCH_TO_FREE(*lmbuf);
+
+		pkt = *(tx_pkts + i);
+
+		if (unlikely(pkt->nb_segs > 1 &&
+			     !(hw->cap & NFP_NET_CFG_CTRL_GATHER))) {
+			PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set");
+			rte_panic("Multisegment packet unsupported\n");
+		}
+
+		/* Checking if we have enough descriptors */
+		if (unlikely(pkt->nb_segs > free_descs))
+			goto xmit_end;
+
+		/*
+		 * Checksum and VLAN flags just in the first descriptor for a
+		 * multisegment packet, but TSO info needs to be in all of them.
+		 */
+		txd.data_len = pkt->pkt_len;
+		nfp_net_tx_tso(txq, &txd, pkt);
+		nfp_net_tx_cksum(txq, &txd, pkt);
+
+		if ((pkt->ol_flags & PKT_TX_VLAN_PKT) &&
+		    (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)) {
+			txd.flags |= PCIE_DESC_TX_VLAN;
+			txd.vlan = pkt->vlan_tci;
+		}
+
+		/*
+		 * mbuf data_len is the data in one segment and pkt_len data
+		 * in the whole packet. When the packet is just one segment,
+		 * then data_len = pkt_len
+		 */
+		pkt_size = pkt->pkt_len;
+
+		while (pkt) {
+			/* Copying TSO, VLAN and cksum info */
+			*txds = txd;
+
+			/* Releasing mbuf used by this descriptor previously*/
+			if (*lmbuf)
+				rte_pktmbuf_free_seg(*lmbuf);
+
+			/*
+			 * Linking mbuf with descriptor for being released
+			 * next time descriptor is used
+			 */
+			*lmbuf = pkt;
+
+			dma_size = pkt->data_len;
+			dma_addr = rte_mbuf_data_iova(pkt);
+			PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
+				   "%" PRIx64 "", dma_addr);
+
+			/* Filling descriptors fields */
+			txds->dma_len = dma_size;
+			txds->data_len = txd.data_len;
+			txds->dma_addr_hi = (dma_addr >> 32) & 0xff;
+			txds->dma_addr_lo = (dma_addr & 0xffffffff);
+			ASSERT(free_descs > 0);
+			free_descs--;
+
+			txq->wr_p++;
+			if (unlikely(txq->wr_p == txq->tx_count)) /* wrapping?*/
+				txq->wr_p = 0;
+
+			pkt_size -= dma_size;
+
+			/*
+			 * Making the EOP, packets with just one segment
+			 * the priority
+			 */
+			if (likely(!pkt_size))
+				txds->offset_eop = PCIE_DESC_TX_EOP;
+			else
+				txds->offset_eop = 0;
+
+			pkt = pkt->next;
+			/* Referencing next free TX descriptor */
+			txds = &txq->txds[txq->wr_p];
+			lmbuf = &txq->txbufs[txq->wr_p].mbuf;
+			issued_descs++;
+		}
+		i++;
+	}
+
+xmit_end:
+	/* Increment write pointers. Force memory write before we let HW know */
+	rte_wmb();
+	nfp_qcp_ptr_add(txq->qcp_q, NFP_QCP_WRITE_PTR, issued_descs);
+
+	return i;
+}
diff --git a/drivers/net/nfp/nfp_rxtx.h b/drivers/net/nfp/nfp_rxtx.h
index 41a3a4b4e7..d2d0f3f175 100644
--- a/drivers/net/nfp/nfp_rxtx.h
+++ b/drivers/net/nfp/nfp_rxtx.h
@@ -17,6 +17,14 @@
 #include <linux/types.h>
 #include <rte_io.h>
 
+#define NFP_DESC_META_LEN(d) ((d)->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
+
+#define NFP_HASH_OFFSET      ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 4)
+#define NFP_HASH_TYPE_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 8)
+
+#define RTE_MBUF_DMA_ADDR_DEFAULT(mb) \
+	((uint64_t)((mb)->buf_iova + RTE_PKTMBUF_HEADROOM))
+
 /*
  * The maximum number of descriptors is limited by design as
  * DPDK uses uint16_t variables for these values
@@ -266,6 +274,25 @@ struct nfp_net_rxq {
 	int rx_qcidx;
 } __rte_aligned(64);
 
+int nfp_net_rx_freelist_setup(struct rte_eth_dev *dev);
+uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
+				       uint16_t queue_idx);
+uint16_t nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts);
+void nfp_net_rx_queue_release(void *rxq);
+void nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq);
+int nfp_net_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+				  uint16_t nb_desc, unsigned int socket_id,
+				  const struct rte_eth_rxconf *rx_conf,
+				  struct rte_mempool *mp);
+void nfp_net_tx_queue_release(void *txq);
+void nfp_net_reset_tx_queue(struct nfp_net_txq *txq);
+int nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+				  uint16_t nb_desc, unsigned int socket_id,
+				  const struct rte_eth_txconf *tx_conf);
+uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+				  uint16_t nb_pkts);
+
 #endif /* _NFP_RXTX_H_ */
 /*
  * Local variables:
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 3/7] net/nfp: move CPP bridge to a separate file
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
@ 2021-07-16  8:35   ` Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
                     ` (5 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

This commit moves the CPP bridge logic to a separate file. A new
corresponding header file is also created.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build      |   1 +
 drivers/net/nfp/nfp_cpp_bridge.c | 392 +++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_cpp_bridge.h |  36 +++
 drivers/net/nfp/nfp_net.c        | 367 +----------------------------
 4 files changed, 430 insertions(+), 366 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.c
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.h

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 1b289e2354..b46ac2d40f 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -20,4 +20,5 @@ sources = files(
         'nfpcore/nfp_hwinfo.c',
         'nfp_net.c',
         'nfp_rxtx.c',
+        'nfp_cpp_bridge.c',
 )
diff --git a/drivers/net/nfp/nfp_cpp_bridge.c b/drivers/net/nfp/nfp_cpp_bridge.c
new file mode 100644
index 0000000000..d916793338
--- /dev/null
+++ b/drivers/net/nfp/nfp_cpp_bridge.c
@@ -0,0 +1,392 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_cpp_bridge.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: CPP Bridge
+ */
+
+#include <rte_service_component.h>
+
+#include "nfpcore/nfp_cpp.h"
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_nsp.h"
+
+#include "nfp_net_logs.h"
+#include "nfp_cpp_bridge.h"
+
+#include <sys/ioctl.h>
+
+/* Prototypes */
+static int nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp);
+static int nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp);
+static int nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp);
+
+void nfp_register_cpp_service(struct nfp_cpp *cpp)
+{
+	uint32_t *cpp_service_id = NULL;
+	struct rte_service_spec service;
+
+	memset(&service, 0, sizeof(struct rte_service_spec));
+	snprintf(service.name, sizeof(service.name), "nfp_cpp_service");
+	service.callback = nfp_cpp_bridge_service_func;
+	service.callback_userdata = (void *)cpp;
+
+	if (rte_service_component_register(&service,
+					   cpp_service_id))
+		RTE_LOG(WARNING, PMD, "NFP CPP bridge service register() failed");
+	else
+		RTE_LOG(DEBUG, PMD, "NFP CPP bridge service registered");
+}
+
+/*
+ * Serving a write request to NFP from host programs. The request
+ * sends the write size and the CPP target. The bridge makes use
+ * of CPP interface handler configured by the PMD setup.
+ */
+static int
+nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp)
+{
+	struct nfp_cpp_area *area;
+	off_t offset, nfp_offset;
+	uint32_t cpp_id, pos, len;
+	uint32_t tmpbuf[16];
+	size_t count, curlen, totlen = 0;
+	int err = 0;
+
+	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
+		sizeof(off_t), sizeof(size_t));
+
+	/* Reading the count param */
+	err = recv(sockfd, &count, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	curlen = count;
+
+	/* Reading the offset param */
+	err = recv(sockfd, &offset, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	/* Obtain target's CPP ID and offset in target */
+	cpp_id = (offset >> 40) << 8;
+	nfp_offset = offset & ((1ull << 40) - 1);
+
+	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
+		offset);
+	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
+		cpp_id, nfp_offset);
+
+	/* Adjust length if not aligned */
+	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
+	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
+		curlen = NFP_CPP_MEMIO_BOUNDARY -
+			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
+	}
+
+	while (count > 0) {
+		/* configure a CPP PCIe2CPP BAR for mapping the CPP target */
+		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
+						    nfp_offset, curlen);
+		if (!area) {
+			RTE_LOG(ERR, PMD, "%s: area alloc fail\n", __func__);
+			return -EIO;
+		}
+
+		/* mapping the target */
+		err = nfp_cpp_area_acquire(area);
+		if (err < 0) {
+			RTE_LOG(ERR, PMD, "area acquire failed\n");
+			nfp_cpp_area_free(area);
+			return -EIO;
+		}
+
+		for (pos = 0; pos < curlen; pos += len) {
+			len = curlen - pos;
+			if (len > sizeof(tmpbuf))
+				len = sizeof(tmpbuf);
+
+			PMD_CPP_LOG(DEBUG, "%s: Receive %u of %zu\n", __func__,
+					   len, count);
+			err = recv(sockfd, tmpbuf, len, MSG_WAITALL);
+			if (err != (int)len) {
+				RTE_LOG(ERR, PMD,
+					"%s: error when receiving, %d of %zu\n",
+					__func__, err, count);
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+			err = nfp_cpp_area_write(area, pos, tmpbuf, len);
+			if (err < 0) {
+				RTE_LOG(ERR, PMD, "nfp_cpp_area_write error\n");
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+		}
+
+		nfp_offset += pos;
+		totlen += pos;
+		nfp_cpp_area_release(area);
+		nfp_cpp_area_free(area);
+
+		count -= pos;
+		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
+			 NFP_CPP_MEMIO_BOUNDARY : count;
+	}
+
+	return 0;
+}
+
+/*
+ * Serving a read request to NFP from host programs. The request
+ * sends the read size and the CPP target. The bridge makes use
+ * of CPP interface handler configured by the PMD setup. The read
+ * data is sent to the requester using the same socket.
+ */
+static int
+nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp)
+{
+	struct nfp_cpp_area *area;
+	off_t offset, nfp_offset;
+	uint32_t cpp_id, pos, len;
+	uint32_t tmpbuf[16];
+	size_t count, curlen, totlen = 0;
+	int err = 0;
+
+	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
+		sizeof(off_t), sizeof(size_t));
+
+	/* Reading the count param */
+	err = recv(sockfd, &count, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	curlen = count;
+
+	/* Reading the offset param */
+	err = recv(sockfd, &offset, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	/* Obtain target's CPP ID and offset in target */
+	cpp_id = (offset >> 40) << 8;
+	nfp_offset = offset & ((1ull << 40) - 1);
+
+	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
+			   offset);
+	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
+			   cpp_id, nfp_offset);
+
+	/* Adjust length if not aligned */
+	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
+	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
+		curlen = NFP_CPP_MEMIO_BOUNDARY -
+			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
+	}
+
+	while (count > 0) {
+		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
+						    nfp_offset, curlen);
+		if (!area) {
+			RTE_LOG(ERR, PMD, "%s: area alloc failed\n", __func__);
+			return -EIO;
+		}
+
+		err = nfp_cpp_area_acquire(area);
+		if (err < 0) {
+			RTE_LOG(ERR, PMD, "area acquire failed\n");
+			nfp_cpp_area_free(area);
+			return -EIO;
+		}
+
+		for (pos = 0; pos < curlen; pos += len) {
+			len = curlen - pos;
+			if (len > sizeof(tmpbuf))
+				len = sizeof(tmpbuf);
+
+			err = nfp_cpp_area_read(area, pos, tmpbuf, len);
+			if (err < 0) {
+				RTE_LOG(ERR, PMD, "nfp_cpp_area_read error\n");
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+			PMD_CPP_LOG(DEBUG, "%s: sending %u of %zu\n", __func__,
+					   len, count);
+
+			err = send(sockfd, tmpbuf, len, 0);
+			if (err != (int)len) {
+				RTE_LOG(ERR, PMD,
+					"%s: error when sending: %d of %zu\n",
+					__func__, err, count);
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+		}
+
+		nfp_offset += pos;
+		totlen += pos;
+		nfp_cpp_area_release(area);
+		nfp_cpp_area_free(area);
+
+		count -= pos;
+		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
+			NFP_CPP_MEMIO_BOUNDARY : count;
+	}
+	return 0;
+}
+
+/*
+ * Serving a ioctl command from host NFP tools. This usually goes to
+ * a kernel driver char driver but it is not available when the PF is
+ * bound to the PMD. Currently just one ioctl command is served and it
+ * does not require any CPP access at all.
+ */
+static int
+nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp)
+{
+	uint32_t cmd, ident_size, tmp;
+	int err;
+
+	/* Reading now the IOCTL command */
+	err = recv(sockfd, &cmd, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
+		return -EIO;
+	}
+
+	/* Only supporting NFP_IOCTL_CPP_IDENTIFICATION */
+	if (cmd != NFP_IOCTL_CPP_IDENTIFICATION) {
+		RTE_LOG(ERR, PMD, "%s: unknown cmd %d\n", __func__, cmd);
+		return -EINVAL;
+	}
+
+	err = recv(sockfd, &ident_size, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
+		return -EIO;
+	}
+
+	tmp = nfp_cpp_model(cpp);
+
+	PMD_CPP_LOG(DEBUG, "%s: sending NFP model %08x\n", __func__, tmp);
+
+	err = send(sockfd, &tmp, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
+		return -EIO;
+	}
+
+	tmp = cpp->interface;
+
+	PMD_CPP_LOG(DEBUG, "%s: sending NFP interface %08x\n", __func__, tmp);
+
+	err = send(sockfd, &tmp, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
+		return -EIO;
+	}
+
+	return 0;
+}
+
+/*
+ * This is the code to be executed by a service core. The CPP bridge interface
+ * is based on a unix socket and requests usually received by a kernel char
+ * driver, read, write and ioctl, are handled by the CPP bridge. NFP host tools
+ * can be executed with a wrapper library and LD_LIBRARY being completely
+ * unaware of the CPP bridge performing the NFP kernel char driver for CPP
+ * accesses.
+ */
+int32_t
+nfp_cpp_bridge_service_func(void *args)
+{
+	struct sockaddr address;
+	struct nfp_cpp *cpp = args;
+	int sockfd, datafd, op, ret;
+
+	unlink("/tmp/nfp_cpp");
+	sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
+	if (sockfd < 0) {
+		RTE_LOG(ERR, PMD, "%s: socket creation error. Service failed\n",
+			__func__);
+		return -EIO;
+	}
+
+	memset(&address, 0, sizeof(struct sockaddr));
+
+	address.sa_family = AF_UNIX;
+	strcpy(address.sa_data, "/tmp/nfp_cpp");
+
+	ret = bind(sockfd, (const struct sockaddr *)&address,
+		   sizeof(struct sockaddr));
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "%s: bind error (%d). Service failed\n",
+				  __func__, errno);
+		close(sockfd);
+		return ret;
+	}
+
+	ret = listen(sockfd, 20);
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "%s: listen error(%d). Service failed\n",
+				  __func__, errno);
+		close(sockfd);
+		return ret;
+	}
+
+	for (;;) {
+		datafd = accept(sockfd, NULL, NULL);
+		if (datafd < 0) {
+			RTE_LOG(ERR, PMD, "%s: accept call error (%d)\n",
+					  __func__, errno);
+			RTE_LOG(ERR, PMD, "%s: service failed\n", __func__);
+			close(sockfd);
+			return -EIO;
+		}
+
+		while (1) {
+			ret = recv(datafd, &op, 4, 0);
+			if (ret <= 0) {
+				PMD_CPP_LOG(DEBUG, "%s: socket close\n",
+						   __func__);
+				break;
+			}
+
+			PMD_CPP_LOG(DEBUG, "%s: getting op %u\n", __func__, op);
+
+			if (op == NFP_BRIDGE_OP_READ)
+				nfp_cpp_bridge_serve_read(datafd, cpp);
+
+			if (op == NFP_BRIDGE_OP_WRITE)
+				nfp_cpp_bridge_serve_write(datafd, cpp);
+
+			if (op == NFP_BRIDGE_OP_IOCTL)
+				nfp_cpp_bridge_serve_ioctl(datafd, cpp);
+
+			if (op == 0)
+				break;
+		}
+		close(datafd);
+	}
+	close(sockfd);
+
+	return 0;
+}
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_cpp_bridge.h b/drivers/net/nfp/nfp_cpp_bridge.h
new file mode 100644
index 0000000000..aea5fdc784
--- /dev/null
+++ b/drivers/net/nfp/nfp_cpp_bridge.h
@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_cpp_bridge.h
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: CPP Bridge header file
+ */
+
+#ifndef _NFP_CPP_BRIDGE_H_
+#define _NFP_CPP_BRIDGE_H_
+
+#define NFP_CPP_MEMIO_BOUNDARY	(1 << 20)
+#define NFP_BRIDGE_OP_READ	20
+#define NFP_BRIDGE_OP_WRITE	30
+#define NFP_BRIDGE_OP_IOCTL	40
+
+#define NFP_IOCTL 'n'
+#define NFP_IOCTL_CPP_IDENTIFICATION _IOW(NFP_IOCTL, 0x8f, uint32_t)
+
+void nfp_register_cpp_service(struct nfp_cpp *cpp);
+int32_t nfp_cpp_bridge_service_func(void *args);
+
+#endif /* _NFP_CPP_BRIDGE_H_ */
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 5bfc23ba04..d79c70c5b7 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -41,6 +41,7 @@
 #include "nfp_rxtx.h"
 #include "nfp_net_logs.h"
 #include "nfp_net_ctrl.h"
+#include "nfp_cpp_bridge.h"
 
 #include <sys/types.h>
 #include <sys/socket.h>
@@ -81,8 +82,6 @@ static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf);
 static int nfp_set_mac_addr(struct rte_eth_dev *dev,
 			     struct rte_ether_addr *mac_addr);
-static int32_t nfp_cpp_bridge_service_func(void *args);
-static void nfp_register_cpp_service(struct nfp_cpp *cpp);
 static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_cpp *cpp,
 			struct nfp_eth_table *nfp_eth_table,
@@ -1926,353 +1925,6 @@ nfp_net_init(struct rte_eth_dev *eth_dev)
 	return err;
 }
 
-#define NFP_CPP_MEMIO_BOUNDARY		(1 << 20)
-
-/*
- * Serving a write request to NFP from host programs. The request
- * sends the write size and the CPP target. The bridge makes use
- * of CPP interface handler configured by the PMD setup.
- */
-static int
-nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp)
-{
-	struct nfp_cpp_area *area;
-	off_t offset, nfp_offset;
-	uint32_t cpp_id, pos, len;
-	uint32_t tmpbuf[16];
-	size_t count, curlen, totlen = 0;
-	int err = 0;
-
-	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
-		sizeof(off_t), sizeof(size_t));
-
-	/* Reading the count param */
-	err = recv(sockfd, &count, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	curlen = count;
-
-	/* Reading the offset param */
-	err = recv(sockfd, &offset, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	/* Obtain target's CPP ID and offset in target */
-	cpp_id = (offset >> 40) << 8;
-	nfp_offset = offset & ((1ull << 40) - 1);
-
-	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
-		offset);
-	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
-		cpp_id, nfp_offset);
-
-	/* Adjust length if not aligned */
-	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
-	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
-		curlen = NFP_CPP_MEMIO_BOUNDARY -
-			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
-	}
-
-	while (count > 0) {
-		/* configure a CPP PCIe2CPP BAR for mapping the CPP target */
-		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
-						    nfp_offset, curlen);
-		if (!area) {
-			RTE_LOG(ERR, PMD, "%s: area alloc fail\n", __func__);
-			return -EIO;
-		}
-
-		/* mapping the target */
-		err = nfp_cpp_area_acquire(area);
-		if (err < 0) {
-			RTE_LOG(ERR, PMD, "area acquire failed\n");
-			nfp_cpp_area_free(area);
-			return -EIO;
-		}
-
-		for (pos = 0; pos < curlen; pos += len) {
-			len = curlen - pos;
-			if (len > sizeof(tmpbuf))
-				len = sizeof(tmpbuf);
-
-			PMD_CPP_LOG(DEBUG, "%s: Receive %u of %zu\n", __func__,
-					   len, count);
-			err = recv(sockfd, tmpbuf, len, MSG_WAITALL);
-			if (err != (int)len) {
-				RTE_LOG(ERR, PMD,
-					"%s: error when receiving, %d of %zu\n",
-					__func__, err, count);
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-			err = nfp_cpp_area_write(area, pos, tmpbuf, len);
-			if (err < 0) {
-				RTE_LOG(ERR, PMD, "nfp_cpp_area_write error\n");
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-		}
-
-		nfp_offset += pos;
-		totlen += pos;
-		nfp_cpp_area_release(area);
-		nfp_cpp_area_free(area);
-
-		count -= pos;
-		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
-			 NFP_CPP_MEMIO_BOUNDARY : count;
-	}
-
-	return 0;
-}
-
-/*
- * Serving a read request to NFP from host programs. The request
- * sends the read size and the CPP target. The bridge makes use
- * of CPP interface handler configured by the PMD setup. The read
- * data is sent to the requester using the same socket.
- */
-static int
-nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp)
-{
-	struct nfp_cpp_area *area;
-	off_t offset, nfp_offset;
-	uint32_t cpp_id, pos, len;
-	uint32_t tmpbuf[16];
-	size_t count, curlen, totlen = 0;
-	int err = 0;
-
-	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
-		sizeof(off_t), sizeof(size_t));
-
-	/* Reading the count param */
-	err = recv(sockfd, &count, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	curlen = count;
-
-	/* Reading the offset param */
-	err = recv(sockfd, &offset, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	/* Obtain target's CPP ID and offset in target */
-	cpp_id = (offset >> 40) << 8;
-	nfp_offset = offset & ((1ull << 40) - 1);
-
-	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
-			   offset);
-	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
-			   cpp_id, nfp_offset);
-
-	/* Adjust length if not aligned */
-	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
-	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
-		curlen = NFP_CPP_MEMIO_BOUNDARY -
-			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
-	}
-
-	while (count > 0) {
-		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
-						    nfp_offset, curlen);
-		if (!area) {
-			RTE_LOG(ERR, PMD, "%s: area alloc failed\n", __func__);
-			return -EIO;
-		}
-
-		err = nfp_cpp_area_acquire(area);
-		if (err < 0) {
-			RTE_LOG(ERR, PMD, "area acquire failed\n");
-			nfp_cpp_area_free(area);
-			return -EIO;
-		}
-
-		for (pos = 0; pos < curlen; pos += len) {
-			len = curlen - pos;
-			if (len > sizeof(tmpbuf))
-				len = sizeof(tmpbuf);
-
-			err = nfp_cpp_area_read(area, pos, tmpbuf, len);
-			if (err < 0) {
-				RTE_LOG(ERR, PMD, "nfp_cpp_area_read error\n");
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-			PMD_CPP_LOG(DEBUG, "%s: sending %u of %zu\n", __func__,
-					   len, count);
-
-			err = send(sockfd, tmpbuf, len, 0);
-			if (err != (int)len) {
-				RTE_LOG(ERR, PMD,
-					"%s: error when sending: %d of %zu\n",
-					__func__, err, count);
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-		}
-
-		nfp_offset += pos;
-		totlen += pos;
-		nfp_cpp_area_release(area);
-		nfp_cpp_area_free(area);
-
-		count -= pos;
-		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
-			NFP_CPP_MEMIO_BOUNDARY : count;
-	}
-	return 0;
-}
-
-#define NFP_IOCTL 'n'
-#define NFP_IOCTL_CPP_IDENTIFICATION _IOW(NFP_IOCTL, 0x8f, uint32_t)
-/*
- * Serving a ioctl command from host NFP tools. This usually goes to
- * a kernel driver char driver but it is not available when the PF is
- * bound to the PMD. Currently just one ioctl command is served and it
- * does not require any CPP access at all.
- */
-static int
-nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp)
-{
-	uint32_t cmd, ident_size, tmp;
-	int err;
-
-	/* Reading now the IOCTL command */
-	err = recv(sockfd, &cmd, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
-		return -EIO;
-	}
-
-	/* Only supporting NFP_IOCTL_CPP_IDENTIFICATION */
-	if (cmd != NFP_IOCTL_CPP_IDENTIFICATION) {
-		RTE_LOG(ERR, PMD, "%s: unknown cmd %d\n", __func__, cmd);
-		return -EINVAL;
-	}
-
-	err = recv(sockfd, &ident_size, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
-		return -EIO;
-	}
-
-	tmp = nfp_cpp_model(cpp);
-
-	PMD_CPP_LOG(DEBUG, "%s: sending NFP model %08x\n", __func__, tmp);
-
-	err = send(sockfd, &tmp, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
-		return -EIO;
-	}
-
-	tmp = cpp->interface;
-
-	PMD_CPP_LOG(DEBUG, "%s: sending NFP interface %08x\n", __func__, tmp);
-
-	err = send(sockfd, &tmp, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
-		return -EIO;
-	}
-
-	return 0;
-}
-
-#define NFP_BRIDGE_OP_READ	20
-#define NFP_BRIDGE_OP_WRITE	30
-#define NFP_BRIDGE_OP_IOCTL	40
-
-/*
- * This is the code to be executed by a service core. The CPP bridge interface
- * is based on a unix socket and requests usually received by a kernel char
- * driver, read, write and ioctl, are handled by the CPP bridge. NFP host tools
- * can be executed with a wrapper library and LD_LIBRARY being completely
- * unaware of the CPP bridge performing the NFP kernel char driver for CPP
- * accesses.
- */
-static int32_t
-nfp_cpp_bridge_service_func(void *args)
-{
-	struct sockaddr address;
-	struct nfp_cpp *cpp = args;
-	int sockfd, datafd, op, ret;
-
-	unlink("/tmp/nfp_cpp");
-	sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
-	if (sockfd < 0) {
-		RTE_LOG(ERR, PMD, "%s: socket creation error. Service failed\n",
-			__func__);
-		return -EIO;
-	}
-
-	memset(&address, 0, sizeof(struct sockaddr));
-
-	address.sa_family = AF_UNIX;
-	strcpy(address.sa_data, "/tmp/nfp_cpp");
-
-	ret = bind(sockfd, (const struct sockaddr *)&address,
-		   sizeof(struct sockaddr));
-	if (ret < 0) {
-		RTE_LOG(ERR, PMD, "%s: bind error (%d). Service failed\n",
-				  __func__, errno);
-		close(sockfd);
-		return ret;
-	}
-
-	ret = listen(sockfd, 20);
-	if (ret < 0) {
-		RTE_LOG(ERR, PMD, "%s: listen error(%d). Service failed\n",
-				  __func__, errno);
-		close(sockfd);
-		return ret;
-	}
-
-	for (;;) {
-		datafd = accept(sockfd, NULL, NULL);
-		if (datafd < 0) {
-			RTE_LOG(ERR, PMD, "%s: accept call error (%d)\n",
-					  __func__, errno);
-			RTE_LOG(ERR, PMD, "%s: service failed\n", __func__);
-			close(sockfd);
-			return -EIO;
-		}
-
-		while (1) {
-			ret = recv(datafd, &op, 4, 0);
-			if (ret <= 0) {
-				PMD_CPP_LOG(DEBUG, "%s: socket close\n",
-						   __func__);
-				break;
-			}
-
-			PMD_CPP_LOG(DEBUG, "%s: getting op %u\n", __func__, op);
-
-			if (op == NFP_BRIDGE_OP_READ)
-				nfp_cpp_bridge_serve_read(datafd, cpp);
-
-			if (op == NFP_BRIDGE_OP_WRITE)
-				nfp_cpp_bridge_serve_write(datafd, cpp);
-
-			if (op == NFP_BRIDGE_OP_IOCTL)
-				nfp_cpp_bridge_serve_ioctl(datafd, cpp);
-
-			if (op == 0)
-				break;
-		}
-		close(datafd);
-	}
-	close(sockfd);
-
-	return 0;
-}
-
 #define DEFAULT_FW_PATH       "/lib/firmware/netronome"
 
 static int
@@ -2491,23 +2143,6 @@ static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
 	return ret;
 }
 
-static void nfp_register_cpp_service(struct nfp_cpp *cpp)
-{
-	uint32_t *cpp_service_id = NULL;
-	struct rte_service_spec service;
-
-	memset(&service, 0, sizeof(struct rte_service_spec));
-	snprintf(service.name, sizeof(service.name), "nfp_cpp_service");
-	service.callback = nfp_cpp_bridge_service_func;
-	service.callback_userdata = (void *)cpp;
-
-	if (rte_service_component_register(&service,
-					   cpp_service_id))
-		RTE_LOG(WARNING, PMD, "NFP CPP bridge service register() failed");
-	else
-		RTE_LOG(DEBUG, PMD, "NFP CPP bridge service registered");
-}
-
 static int nfp_pf_init(struct rte_pci_device *pci_dev)
 {
 	struct nfp_pf_dev *pf_dev = NULL;
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 4/7] net/nfp: prototype common functions in header file
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
                     ` (2 preceding siblings ...)
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
@ 2021-07-16  8:35   ` Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
                     ` (4 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

The majority of "ethdev" type functions are used for both PF devices and
VF devices. Prototype these functions in the nfp_net_pmd header file in
preparation of splitting PF and VF specific functions.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/nfp_net.c     | 87 +++++++++++++----------------------
 drivers/net/nfp/nfp_net_pmd.h | 49 ++++++++++++++++++++
 2 files changed, 81 insertions(+), 55 deletions(-)

diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index d79c70c5b7..da35bba4ef 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -53,35 +53,12 @@
 
 /* Prototypes */
 static int nfp_net_close(struct rte_eth_dev *dev);
-static int nfp_net_configure(struct rte_eth_dev *dev);
-static void nfp_net_dev_interrupt_handler(void *param);
-static void nfp_net_dev_interrupt_delayed_handler(void *param);
-static int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
-static int nfp_net_infos_get(struct rte_eth_dev *dev,
-			     struct rte_eth_dev_info *dev_info);
 static int nfp_net_init(struct rte_eth_dev *eth_dev);
 static int nfp_pf_init(struct rte_pci_device *pci_dev);
 static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
 static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
-static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
-static int nfp_net_promisc_enable(struct rte_eth_dev *dev);
-static int nfp_net_promisc_disable(struct rte_eth_dev *dev);
-static int nfp_net_start(struct rte_eth_dev *dev);
-static int nfp_net_stats_get(struct rte_eth_dev *dev,
-			      struct rte_eth_stats *stats);
-static int nfp_net_stats_reset(struct rte_eth_dev *dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
-static int nfp_net_rss_config_default(struct rte_eth_dev *dev);
-static int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
-				   struct rte_eth_rss_conf *rss_conf);
-static int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
-		    struct rte_eth_rss_reta_entry64 *reta_conf,
-		    uint16_t reta_size);
-static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
-			struct rte_eth_rss_conf *rss_conf);
-static int nfp_set_mac_addr(struct rte_eth_dev *dev,
-			     struct rte_ether_addr *mac_addr);
 static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_cpp *cpp,
 			struct nfp_eth_table *nfp_eth_table,
@@ -136,7 +113,7 @@ __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
  * Write the update word to the BAR and ping the reconfig queue. Then poll
  * until the firmware has acknowledged the update by zeroing the update word.
  */
-static int
+int
 nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update)
 {
 	uint32_t err;
@@ -172,7 +149,7 @@ nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update)
  * before any other function in the Ethernet API. This function can
  * also be re-invoked when a device is in the stopped state.
  */
-static int
+int
 nfp_net_configure(struct rte_eth_dev *dev)
 {
 	struct rte_eth_conf *dev_conf;
@@ -215,7 +192,7 @@ nfp_net_configure(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static void
+void
 nfp_net_enable_queues(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -239,7 +216,7 @@ nfp_net_enable_queues(struct rte_eth_dev *dev)
 	nn_cfg_writeq(hw, NFP_NET_CFG_RXRS_ENABLE, enabled_queues);
 }
 
-static void
+void
 nfp_net_disable_queues(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -264,14 +241,14 @@ nfp_net_disable_queues(struct rte_eth_dev *dev)
 	hw->ctrl = new_ctrl;
 }
 
-static void
+void
 nfp_net_params_setup(struct nfp_net_hw *hw)
 {
 	nn_cfg_writel(hw, NFP_NET_CFG_MTU, hw->mtu);
 	nn_cfg_writel(hw, NFP_NET_CFG_FLBUFSZ, hw->flbufsz);
 }
 
-static void
+void
 nfp_net_cfg_queue_setup(struct nfp_net_hw *hw)
 {
 	hw->qcp_cfg = hw->tx_bar + NFP_QCP_QUEUE_ADDR_SZ;
@@ -279,7 +256,7 @@ nfp_net_cfg_queue_setup(struct nfp_net_hw *hw)
 
 #define ETH_ADDR_LEN	6
 
-static void
+void
 nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
 {
 	int i;
@@ -318,7 +295,7 @@ nfp_net_vf_read_mac(struct nfp_net_hw *hw)
 	memcpy(&hw->mac_addr[4], &tmp, 2);
 }
 
-static void
+void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
 	uint32_t mac0 = *(uint32_t *)mac;
@@ -366,7 +343,7 @@ nfp_set_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr)
 	return 0;
 }
 
-static int
+int
 nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
 			   struct rte_intr_handle *intr_handle)
 {
@@ -410,7 +387,7 @@ nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static uint32_t
+uint32_t
 nfp_check_offloads(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -746,7 +723,7 @@ nfp_net_close(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_promisc_enable(struct rte_eth_dev *dev)
 {
 	uint32_t new_ctrl, update = 0;
@@ -783,7 +760,7 @@ nfp_net_promisc_enable(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_promisc_disable(struct rte_eth_dev *dev)
 {
 	uint32_t new_ctrl, update = 0;
@@ -819,7 +796,7 @@ nfp_net_promisc_disable(struct rte_eth_dev *dev)
  * Wait to complete is needed as it can take up to 9 seconds to get the Link
  * status.
  */
-static int
+int
 nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
 {
 	struct nfp_net_hw *hw;
@@ -869,7 +846,7 @@ nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
 	return ret;
 }
 
-static int
+int
 nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 {
 	int i;
@@ -964,7 +941,7 @@ nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 	return -EINVAL;
 }
 
-static int
+int
 nfp_net_stats_reset(struct rte_eth_dev *dev)
 {
 	int i;
@@ -1029,7 +1006,7 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 {
 	struct nfp_net_hw *hw;
@@ -1123,7 +1100,7 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 	return 0;
 }
 
-static const uint32_t *
+const uint32_t *
 nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 {
 	static const uint32_t ptypes[] = {
@@ -1140,7 +1117,7 @@ nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 	return NULL;
 }
 
-static int
+int
 nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
 	struct rte_pci_device *pci_dev;
@@ -1160,7 +1137,7 @@ nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 	return 0;
 }
 
-static int
+int
 nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
 	struct rte_pci_device *pci_dev;
@@ -1179,7 +1156,7 @@ nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
 	return 0;
 }
 
-static void
+void
 nfp_net_dev_link_status_print(struct rte_eth_dev *dev)
 {
 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
@@ -1208,7 +1185,7 @@ nfp_net_dev_link_status_print(struct rte_eth_dev *dev)
  * If MSI-X auto-masking is enabled clear the mask bit, otherwise
  * clear the ICR for the entry.
  */
-static void
+void
 nfp_net_irq_unmask(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -1229,7 +1206,7 @@ nfp_net_irq_unmask(struct rte_eth_dev *dev)
 	}
 }
 
-static void
+void
 nfp_net_dev_interrupt_handler(void *param)
 {
 	int64_t timeout;
@@ -1272,7 +1249,7 @@ nfp_net_dev_interrupt_handler(void *param)
  *
  * @return  void
  */
-static void
+void
 nfp_net_dev_interrupt_delayed_handler(void *param)
 {
 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
@@ -1286,7 +1263,7 @@ nfp_net_dev_interrupt_delayed_handler(void *param)
 	nfp_net_irq_unmask(dev);
 }
 
-static int
+int
 nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 {
 	struct nfp_net_hw *hw;
@@ -1321,7 +1298,7 @@ nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	return 0;
 }
 
-static int
+int
 nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 {
 	uint32_t new_ctrl, update;
@@ -1353,7 +1330,7 @@ nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 	return ret;
 }
 
-static int
+int
 nfp_net_rss_reta_write(struct rte_eth_dev *dev,
 		    struct rte_eth_rss_reta_entry64 *reta_conf,
 		    uint16_t reta_size)
@@ -1404,7 +1381,7 @@ nfp_net_rss_reta_write(struct rte_eth_dev *dev,
 }
 
 /* Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device */
-static int
+int
 nfp_net_reta_update(struct rte_eth_dev *dev,
 		    struct rte_eth_rss_reta_entry64 *reta_conf,
 		    uint16_t reta_size)
@@ -1430,7 +1407,7 @@ nfp_net_reta_update(struct rte_eth_dev *dev,
 }
 
  /* Query Redirection Table(RETA) of Receive Side Scaling of Ethernet device. */
-static int
+int
 nfp_net_reta_query(struct rte_eth_dev *dev,
 		   struct rte_eth_rss_reta_entry64 *reta_conf,
 		   uint16_t reta_size)
@@ -1477,7 +1454,7 @@ nfp_net_reta_query(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf)
 {
@@ -1527,7 +1504,7 @@ nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf)
 {
@@ -1563,7 +1540,7 @@ nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
 			  struct rte_eth_rss_conf *rss_conf)
 {
@@ -1614,7 +1591,7 @@ nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_config_default(struct rte_eth_dev *dev)
 {
 	struct rte_eth_conf *dev_conf;
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index 9265496bf0..dc05e888df 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -349,6 +349,55 @@ nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
 		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
 }
 
+/* Prototypes for common NFP functions */
+int nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update);
+int nfp_net_configure(struct rte_eth_dev *dev);
+void nfp_net_enable_queues(struct rte_eth_dev *dev);
+void nfp_net_disable_queues(struct rte_eth_dev *dev);
+void nfp_net_params_setup(struct nfp_net_hw *hw);
+void nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src);
+void nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac);
+int nfp_set_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr);
+int nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
+			       struct rte_intr_handle *intr_handle);
+uint32_t nfp_check_offloads(struct rte_eth_dev *dev);
+int nfp_net_promisc_enable(struct rte_eth_dev *dev);
+int nfp_net_promisc_disable(struct rte_eth_dev *dev);
+int nfp_net_link_update(struct rte_eth_dev *dev,
+			__rte_unused int wait_to_complete);
+int nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats);
+int nfp_net_stats_reset(struct rte_eth_dev *dev);
+int nfp_net_infos_get(struct rte_eth_dev *dev,
+		      struct rte_eth_dev_info *dev_info);
+const uint32_t *nfp_net_supported_ptypes_get(struct rte_eth_dev *dev);
+int nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
+int nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
+void nfp_net_params_setup(struct nfp_net_hw *hw);
+void nfp_net_cfg_queue_setup(struct nfp_net_hw *hw);
+void nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src);
+void nfp_net_dev_link_status_print(struct rte_eth_dev *dev);
+void nfp_net_irq_unmask(struct rte_eth_dev *dev);
+void nfp_net_dev_interrupt_handler(void *param);
+void nfp_net_dev_interrupt_delayed_handler(void *param);
+int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
+int nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask);
+int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
+			   struct rte_eth_rss_reta_entry64 *reta_conf,
+			   uint16_t reta_size);
+int nfp_net_reta_update(struct rte_eth_dev *dev,
+			struct rte_eth_rss_reta_entry64 *reta_conf,
+			uint16_t reta_size);
+int nfp_net_reta_query(struct rte_eth_dev *dev,
+		       struct rte_eth_rss_reta_entry64 *reta_conf,
+		       uint16_t reta_size);
+int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
+			   struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
+			    struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
+			      struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_config_default(struct rte_eth_dev *dev);
+
 #define NFP_NET_DEV_PRIVATE_TO_HW(adapter)\
 	(&((struct nfp_net_adapter *)adapter)->hw)
 
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 5/7] net/nfp: move VF functions into new file
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
                     ` (3 preceding siblings ...)
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
@ 2021-07-16  8:35   ` Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 6/7] net/nfp: move PF " Heinrich Kuhn
                     ` (3 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Move any ethdev functionality specific to VF devices into a new file
called nfp_ethdev_vf.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build     |   1 +
 drivers/net/nfp/nfp_ethdev_vf.c | 504 ++++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_net.c       |  42 +--
 3 files changed, 506 insertions(+), 41 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_ethdev_vf.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index b46ac2d40f..34f4054b3c 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -21,4 +21,5 @@ sources = files(
         'nfp_net.c',
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
+        'nfp_ethdev_vf.c',
 )
diff --git a/drivers/net/nfp/nfp_ethdev_vf.c b/drivers/net/nfp/nfp_ethdev_vf.c
new file mode 100644
index 0000000000..223142c0ed
--- /dev/null
+++ b/drivers/net/nfp/nfp_ethdev_vf.c
@@ -0,0 +1,504 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_ethdev_vf.c
+ *
+ * Netronome vNIC  VF DPDK Poll-Mode Driver: Main entry point
+ */
+
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_rtsym.h"
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+
+static void nfp_netvf_read_mac(struct nfp_net_hw *hw);
+static int nfp_netvf_start(struct rte_eth_dev *dev);
+static int nfp_netvf_stop(struct rte_eth_dev *dev);
+static int nfp_netvf_set_link_up(struct rte_eth_dev *dev);
+static int nfp_netvf_set_link_down(struct rte_eth_dev *dev);
+static int nfp_netvf_close(struct rte_eth_dev *dev);
+static int nfp_netvf_init(struct rte_eth_dev *eth_dev);
+static int nfp_vf_pci_uninit(struct rte_eth_dev *eth_dev);
+static int eth_nfp_vf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+	struct rte_pci_device *pci_dev);
+static int eth_nfp_vf_pci_remove(struct rte_pci_device *pci_dev);
+
+static void
+nfp_netvf_read_mac(struct nfp_net_hw *hw)
+{
+	uint32_t tmp;
+
+	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR));
+	memcpy(&hw->mac_addr[0], &tmp, 4);
+
+	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR + 4));
+	memcpy(&hw->mac_addr[4], &tmp, 2);
+}
+
+static int
+nfp_netvf_start(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+	uint32_t new_ctrl, update = 0;
+	struct nfp_net_hw *hw;
+	struct rte_eth_conf *dev_conf;
+	struct rte_eth_rxmode *rxmode;
+	uint32_t intr_vector;
+	int ret;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_LOG(DEBUG, "Start");
+
+	/* Disabling queues just in case... */
+	nfp_net_disable_queues(dev);
+
+	/* Enabling the required queues in the device */
+	nfp_net_enable_queues(dev);
+
+	/* check and configure queue intr-vector mapping */
+	if (dev->data->dev_conf.intr_conf.rxq != 0) {
+		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+			/*
+			 * Better not to share LSC with RX interrupts.
+			 * Unregistering LSC interrupt handler
+			 */
+			rte_intr_callback_unregister(&pci_dev->intr_handle,
+				nfp_net_dev_interrupt_handler, (void *)dev);
+
+			if (dev->data->nb_rx_queues > 1) {
+				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+					     "supports 1 queue with UIO");
+				return -EIO;
+			}
+		}
+		intr_vector = dev->data->nb_rx_queues;
+		if (rte_intr_efd_enable(intr_handle, intr_vector))
+			return -1;
+
+		nfp_configure_rx_interrupt(dev, intr_handle);
+		update = NFP_NET_CFG_UPDATE_MSIX;
+	}
+
+	rte_intr_enable(intr_handle);
+
+	new_ctrl = nfp_check_offloads(dev);
+
+	/* Writing configuration parameters in the device */
+	nfp_net_params_setup(hw);
+
+	dev_conf = &dev->data->dev_conf;
+	rxmode = &dev_conf->rxmode;
+
+	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
+		nfp_net_rss_config_default(dev);
+		update |= NFP_NET_CFG_UPDATE_RSS;
+		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
+	}
+
+	/* Enable device */
+	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
+
+	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+
+	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
+		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
+
+	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
+	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+		return -EIO;
+
+	/*
+	 * Allocating rte mbufs for configured rx queues.
+	 * This requires queues being enabled before
+	 */
+	if (nfp_net_rx_freelist_setup(dev) < 0) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	hw->ctrl = new_ctrl;
+
+	return 0;
+
+error:
+	/*
+	 * An error returned by this function should mean the app
+	 * exiting and then the system releasing all the memory
+	 * allocated even memory coming from hugepages.
+	 *
+	 * The device could be enabled at this point with some queues
+	 * ready for getting packets. This is true if the call to
+	 * nfp_net_rx_freelist_setup() succeeds for some queues but
+	 * fails for subsequent queues.
+	 *
+	 * This should make the app exiting but better if we tell the
+	 * device first.
+	 */
+	nfp_net_disable_queues(dev);
+
+	return ret;
+}
+
+static int
+nfp_netvf_stop(struct rte_eth_dev *dev)
+{
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	PMD_INIT_LOG(DEBUG, "Stop");
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	return 0;
+}
+
+static int
+nfp_netvf_set_link_up(struct rte_eth_dev *dev __rte_unused)
+{
+	return -ENOTSUP;
+}
+
+/* Set the link down. */
+static int
+nfp_netvf_set_link_down(struct rte_eth_dev *dev __rte_unused)
+{
+	return -ENOTSUP;
+}
+
+/* Reset and stop device. The device can not be restarted. */
+static int
+nfp_netvf_close(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	PMD_INIT_LOG(DEBUG, "Close");
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+	/*
+	 * We assume that the DPDK application is stopping all the
+	 * threads/queues before calling the device close function.
+	 */
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q =  (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q =  (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	rte_intr_disable(&pci_dev->intr_handle);
+
+	/* unregister callback func from eal lib */
+	rte_intr_callback_unregister(&pci_dev->intr_handle,
+				     nfp_net_dev_interrupt_handler,
+				     (void *)dev);
+
+	/*
+	 * The ixgbe PMD driver disables the pcie master on the
+	 * device. The i40e does not...
+	 */
+
+	return 0;
+}
+
+/* Initialise and register VF driver with DPDK Application */
+static const struct eth_dev_ops nfp_netvf_eth_dev_ops = {
+	.dev_configure		= nfp_net_configure,
+	.dev_start		= nfp_netvf_start,
+	.dev_stop		= nfp_netvf_stop,
+	.dev_set_link_up	= nfp_netvf_set_link_up,
+	.dev_set_link_down	= nfp_netvf_set_link_down,
+	.dev_close		= nfp_netvf_close,
+	.promiscuous_enable	= nfp_net_promisc_enable,
+	.promiscuous_disable	= nfp_net_promisc_disable,
+	.link_update		= nfp_net_link_update,
+	.stats_get		= nfp_net_stats_get,
+	.stats_reset		= nfp_net_stats_reset,
+	.dev_infos_get		= nfp_net_infos_get,
+	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
+	.mtu_set		= nfp_net_dev_mtu_set,
+	.mac_addr_set           = nfp_set_mac_addr,
+	.vlan_offload_set	= nfp_net_vlan_offload_set,
+	.reta_update		= nfp_net_reta_update,
+	.reta_query		= nfp_net_reta_query,
+	.rss_hash_update	= nfp_net_rss_hash_update,
+	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
+	.rx_queue_setup		= nfp_net_rx_queue_setup,
+	.rx_queue_release	= nfp_net_rx_queue_release,
+	.tx_queue_setup		= nfp_net_tx_queue_setup,
+	.tx_queue_release	= nfp_net_tx_queue_release,
+	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
+	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
+};
+
+static int
+nfp_netvf_init(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_net_hw *hw;
+	struct rte_ether_addr *tmp_ether_addr;
+
+	uint64_t tx_bar_off = 0, rx_bar_off = 0;
+	uint32_t start_q;
+	int stride = 4;
+	int port = 0;
+	int err;
+
+	PMD_INIT_FUNC_TRACE();
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* NFP can not handle DMA addresses requiring more than 40 bits */
+	if (rte_mem_check_dma_mask(40)) {
+		RTE_LOG(ERR, PMD, "device %s can not be used:",
+				   pci_dev->device.name);
+		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
+		return -ENODEV;
+	};
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+
+	eth_dev->dev_ops = &nfp_netvf_eth_dev_ops;
+	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	rte_eth_copy_pci_info(eth_dev, pci_dev);
+
+	hw->device_id = pci_dev->id.device_id;
+	hw->vendor_id = pci_dev->id.vendor_id;
+	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
+	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
+
+	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
+		     pci_dev->id.vendor_id, pci_dev->id.device_id,
+		     pci_dev->addr.domain, pci_dev->addr.bus,
+		     pci_dev->addr.devid, pci_dev->addr.function);
+
+	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
+	if (hw->ctrl_bar == NULL) {
+		PMD_DRV_LOG(ERR,
+			"hw->ctrl_bar is NULL. BAR0 not configured");
+		return -ENODEV;
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+
+	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
+	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
+
+	/* Work out where in the BAR the queues start. */
+	switch (pci_dev->id.device_id) {
+	case PCI_DEVICE_ID_NFP6000_VF_NIC:
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
+		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
+		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
+		err = -ENODEV;
+		goto dev_err_ctrl_map;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
+	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
+
+	hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+		     tx_bar_off;
+	hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+		     rx_bar_off;
+
+	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
+		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
+
+	nfp_net_cfg_queue_setup(hw);
+
+	/* Get some of the read-only fields from the config BAR */
+	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
+	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
+	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
+	hw->mtu = RTE_ETHER_MTU;
+
+	/* VLAN insertion is incompatible with LSOv2 */
+	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
+		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
+
+	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
+		hw->rx_offset = NFP_NET_RX_OFFSET;
+	else
+		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
+
+	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
+			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
+			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
+
+	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
+		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
+
+	hw->ctrl = 0;
+
+	hw->stride_rx = stride;
+	hw->stride_tx = stride;
+
+	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
+		     hw->max_rx_queues, hw->max_tx_queues);
+
+	/* Initializing spinlock for reconfigs */
+	rte_spinlock_init(&hw->reconfig_lock);
+
+	/* Allocating memory for mac addr */
+	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
+					       RTE_ETHER_ADDR_LEN, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
+		err = -ENOMEM;
+		goto dev_err_queues_map;
+	}
+
+	nfp_netvf_read_mac(hw);
+
+	tmp_ether_addr = (struct rte_ether_addr *)&hw->mac_addr;
+	if (!rte_is_valid_assigned_ether_addr(tmp_ether_addr)) {
+		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
+				   port);
+		/* Using random mac addresses for VFs */
+		rte_eth_random_addr(&hw->mac_addr[0]);
+		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+	}
+
+	/* Copying mac address to DPDK eth_dev struct */
+	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
+			&eth_dev->data->mac_addrs[0]);
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
+
+	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
+
+	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
+		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
+		     eth_dev->data->port_id, pci_dev->id.vendor_id,
+		     pci_dev->id.device_id,
+		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
+		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		/* Registering LSC interrupt handler */
+		rte_intr_callback_register(&pci_dev->intr_handle,
+					   nfp_net_dev_interrupt_handler,
+					   (void *)eth_dev);
+		/* Telling the firmware about the LSC interrupt entry */
+		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
+		/* Recording current stats counters values */
+		nfp_net_stats_reset(eth_dev);
+	}
+
+	return 0;
+
+dev_err_queues_map:
+		nfp_cpp_area_free(hw->hwqueues_area);
+dev_err_ctrl_map:
+		nfp_cpp_area_free(hw->ctrl_area);
+
+	return err;
+}
+
+static const struct rte_pci_id pci_id_nfp_vf_net_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP6000_VF_NIC)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static int nfp_vf_pci_uninit(struct rte_eth_dev *eth_dev)
+{
+	/* VF cleanup, just free private port data */
+	return nfp_netvf_close(eth_dev);
+}
+
+static int eth_nfp_vf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+	struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_probe(pci_dev,
+		sizeof(struct nfp_net_adapter), nfp_netvf_init);
+}
+
+static int eth_nfp_vf_pci_remove(struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_vf_pci_uninit);
+}
+
+static struct rte_pci_driver rte_nfp_net_vf_pmd = {
+	.id_table = pci_id_nfp_vf_net_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	.probe = eth_nfp_vf_pci_probe,
+	.remove = eth_nfp_vf_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_nfp_vf, rte_nfp_net_vf_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_vf, pci_id_nfp_vf_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_vf, "* igb_uio | uio_pci_generic | vfio");
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index da35bba4ef..a1722886ba 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -56,6 +56,7 @@ static int nfp_net_close(struct rte_eth_dev *dev);
 static int nfp_net_init(struct rte_eth_dev *eth_dev);
 static int nfp_pf_init(struct rte_pci_device *pci_dev);
 static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
+static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
 static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
@@ -283,18 +284,6 @@ nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
 	return 0;
 }
 
-static void
-nfp_net_vf_read_mac(struct nfp_net_hw *hw)
-{
-	uint32_t tmp;
-
-	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR));
-	memcpy(&hw->mac_addr[0], &tmp, 4);
-
-	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR + 4));
-	memcpy(&hw->mac_addr[4], &tmp, 2);
-}
-
 void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
@@ -1852,8 +1841,6 @@ nfp_net_init(struct rte_eth_dev *eth_dev)
 	if (hw->is_phyport) {
 		nfp_net_pf_read_mac(pf_dev, port);
 		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	} else {
-		nfp_net_vf_read_mac(hw);
 	}
 
 	if (!rte_is_valid_assigned_ether_addr(
@@ -2369,16 +2356,6 @@ static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
 	},
 };
 
-static const struct rte_pci_id pci_id_nfp_vf_net_map[] = {
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP6000_VF_NIC)
-	},
-	{
-		.vendor_id = 0,
-	},
-};
-
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
 {
 	struct rte_pci_device *pci_dev;
@@ -2402,13 +2379,6 @@ static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
 	return nfp_net_close(eth_dev);
 }
 
-static int eth_nfp_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
-	struct rte_pci_device *pci_dev)
-{
-	return rte_eth_dev_pci_generic_probe(pci_dev,
-		sizeof(struct nfp_net_adapter), nfp_net_init);
-}
-
 static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
 {
 	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
@@ -2421,19 +2391,9 @@ static struct rte_pci_driver rte_nfp_net_pf_pmd = {
 	.remove = eth_nfp_pci_remove,
 };
 
-static struct rte_pci_driver rte_nfp_net_vf_pmd = {
-	.id_table = pci_id_nfp_vf_net_map,
-	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
-	.probe = eth_nfp_pci_probe,
-	.remove = eth_nfp_pci_remove,
-};
-
 RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
-RTE_PMD_REGISTER_PCI(net_nfp_vf, rte_nfp_net_vf_pmd);
 RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
-RTE_PMD_REGISTER_PCI_TABLE(net_nfp_vf, pci_id_nfp_vf_net_map);
 RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
-RTE_PMD_REGISTER_KMOD_DEP(net_nfp_vf, "* igb_uio | uio_pci_generic | vfio");
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_init, init, NOTICE);
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_driver, driver, NOTICE);
 /*
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 6/7] net/nfp: move PF functions into new file
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
                     ` (4 preceding siblings ...)
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
@ 2021-07-16  8:35   ` Heinrich Kuhn
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
                     ` (2 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Similar to the last commit, this changeset moves all the PF specific
functions to a new file called nfp_ethdev.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build  |    1 +
 drivers/net/nfp/nfp_ethdev.c | 1099 ++++++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_net.c    | 1088 +--------------------------------
 3 files changed, 1103 insertions(+), 1085 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_ethdev.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 34f4054b3c..ab64d0cac3 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -22,4 +22,5 @@ sources = files(
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
         'nfp_ethdev_vf.c',
+        'nfp_ethdev.c',
 )
diff --git a/drivers/net/nfp/nfp_ethdev.c b/drivers/net/nfp/nfp_ethdev.c
new file mode 100644
index 0000000000..ab08906704
--- /dev/null
+++ b/drivers/net/nfp/nfp_ethdev.c
@@ -0,0 +1,1099 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_ethdev.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: Main entry point
+ */
+
+#include <rte_common.h>
+#include <ethdev_driver.h>
+#include <ethdev_pci.h>
+#include <rte_dev.h>
+#include <rte_ether.h>
+#include <rte_malloc.h>
+#include <rte_memzone.h>
+#include <rte_mempool.h>
+#include <rte_service_component.h>
+
+#include "nfpcore/nfp_cpp.h"
+#include "nfpcore/nfp_nffw.h"
+#include "nfpcore/nfp_hwinfo.h"
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_rtsym.h"
+#include "nfpcore/nfp_nsp.h"
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+#include "nfp_cpp_bridge.h"
+
+
+static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
+static int nfp_net_start(struct rte_eth_dev *dev);
+static int nfp_net_stop(struct rte_eth_dev *dev);
+static int nfp_net_set_link_up(struct rte_eth_dev *dev);
+static int nfp_net_set_link_down(struct rte_eth_dev *dev);
+static int nfp_net_close(struct rte_eth_dev *dev);
+static int nfp_net_init(struct rte_eth_dev *eth_dev);
+static int nfp_fw_upload(struct rte_pci_device *dev,
+			 struct nfp_nsp *nsp, char *card);
+static int nfp_fw_setup(struct rte_pci_device *dev,
+			struct nfp_cpp *cpp,
+			struct nfp_eth_table *nfp_eth_table,
+			struct nfp_hwinfo *hwinfo);
+static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
+static int nfp_pf_init(struct rte_pci_device *pci_dev);
+static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
+static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+			    struct rte_pci_device *dev);
+static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
+static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev);
+
+static int
+nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
+{
+	struct nfp_eth_table *nfp_eth_table;
+	struct nfp_net_hw *hw = NULL;
+
+	/* Grab a pointer to the correct physical port */
+	hw = pf_dev->ports[port];
+
+	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
+
+	nfp_eth_copy_mac((uint8_t *)&hw->mac_addr,
+			 (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
+
+	free(nfp_eth_table);
+	return 0;
+}
+
+static int
+nfp_net_start(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+	uint32_t new_ctrl, update = 0;
+	struct nfp_net_hw *hw;
+	struct nfp_pf_dev *pf_dev;
+	struct rte_eth_conf *dev_conf;
+	struct rte_eth_rxmode *rxmode;
+	uint32_t intr_vector;
+	int ret;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+
+	PMD_INIT_LOG(DEBUG, "Start");
+
+	/* Disabling queues just in case... */
+	nfp_net_disable_queues(dev);
+
+	/* Enabling the required queues in the device */
+	nfp_net_enable_queues(dev);
+
+	/* check and configure queue intr-vector mapping */
+	if (dev->data->dev_conf.intr_conf.rxq != 0) {
+		if (pf_dev->multiport) {
+			PMD_INIT_LOG(ERR, "PMD rx interrupt is not supported "
+					  "with NFP multiport PF");
+				return -EINVAL;
+		}
+		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+			/*
+			 * Better not to share LSC with RX interrupts.
+			 * Unregistering LSC interrupt handler
+			 */
+			rte_intr_callback_unregister(&pci_dev->intr_handle,
+				nfp_net_dev_interrupt_handler, (void *)dev);
+
+			if (dev->data->nb_rx_queues > 1) {
+				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+					     "supports 1 queue with UIO");
+				return -EIO;
+			}
+		}
+		intr_vector = dev->data->nb_rx_queues;
+		if (rte_intr_efd_enable(intr_handle, intr_vector))
+			return -1;
+
+		nfp_configure_rx_interrupt(dev, intr_handle);
+		update = NFP_NET_CFG_UPDATE_MSIX;
+	}
+
+	rte_intr_enable(intr_handle);
+
+	new_ctrl = nfp_check_offloads(dev);
+
+	/* Writing configuration parameters in the device */
+	nfp_net_params_setup(hw);
+
+	dev_conf = &dev->data->dev_conf;
+	rxmode = &dev_conf->rxmode;
+
+	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
+		nfp_net_rss_config_default(dev);
+		update |= NFP_NET_CFG_UPDATE_RSS;
+		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
+	}
+
+	/* Enable device */
+	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
+
+	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+
+	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
+		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
+
+	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
+	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+		return -EIO;
+
+	/*
+	 * Allocating rte mbufs for configured rx queues.
+	 * This requires queues being enabled before
+	 */
+	if (nfp_net_rx_freelist_setup(dev) < 0) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port up */
+		nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
+	else
+		nfp_eth_set_configured(dev->process_private,
+				       hw->nfp_idx, 1);
+
+	hw->ctrl = new_ctrl;
+
+	return 0;
+
+error:
+	/*
+	 * An error returned by this function should mean the app
+	 * exiting and then the system releasing all the memory
+	 * allocated even memory coming from hugepages.
+	 *
+	 * The device could be enabled at this point with some queues
+	 * ready for getting packets. This is true if the call to
+	 * nfp_net_rx_freelist_setup() succeeds for some queues but
+	 * fails for subsequent queues.
+	 *
+	 * This should make the app exiting but better if we tell the
+	 * device first.
+	 */
+	nfp_net_disable_queues(dev);
+
+	return ret;
+}
+
+/* Stop device: disable rx and tx functions to allow for reconfiguring. */
+static int
+nfp_net_stop(struct rte_eth_dev *dev)
+{
+	int i;
+	struct nfp_net_hw *hw;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+
+	PMD_INIT_LOG(DEBUG, "Stop");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
+	else
+		nfp_eth_set_configured(dev->process_private,
+				       hw->nfp_idx, 0);
+
+	return 0;
+}
+
+/* Set the link up. */
+static int
+nfp_net_set_link_up(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+
+	PMD_DRV_LOG(DEBUG, "Set link up");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
+	else
+		return nfp_eth_set_configured(dev->process_private,
+					      hw->nfp_idx, 1);
+}
+
+/* Set the link down. */
+static int
+nfp_net_set_link_down(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+
+	PMD_DRV_LOG(DEBUG, "Set link down");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
+	else
+		return nfp_eth_set_configured(dev->process_private,
+					      hw->nfp_idx, 0);
+}
+
+/* Reset and stop device. The device can not be restarted. */
+static int
+nfp_net_close(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+	struct rte_pci_device *pci_dev;
+	struct nfp_pf_dev *pf_dev;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	PMD_INIT_LOG(DEBUG, "Close");
+
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+	/*
+	 * We assume that the DPDK application is stopping all the
+	 * threads/queues before calling the device close function.
+	 */
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	/* Only free PF resources after all physical ports have been closed */
+	/* Mark this port as unused and free device priv resources*/
+	nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
+	pf_dev->ports[hw->idx] = NULL;
+	rte_eth_dev_release_port(dev);
+
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		/* Check to see if ports are still in use */
+		if (pf_dev->ports[i])
+			return 0;
+	}
+
+	/* Now it is safe to free all PF resources */
+	PMD_INIT_LOG(INFO, "Freeing PF resources");
+	nfp_cpp_area_free(pf_dev->ctrl_area);
+	nfp_cpp_area_free(pf_dev->hwqueues_area);
+	free(pf_dev->hwinfo);
+	free(pf_dev->sym_tbl);
+	nfp_cpp_free(pf_dev->cpp);
+	rte_free(pf_dev);
+
+	rte_intr_disable(&pci_dev->intr_handle);
+
+	/* unregister callback func from eal lib */
+	rte_intr_callback_unregister(&pci_dev->intr_handle,
+				     nfp_net_dev_interrupt_handler,
+				     (void *)dev);
+
+	/*
+	 * The ixgbe PMD driver disables the pcie master on the
+	 * device. The i40e does not...
+	 */
+
+	return 0;
+}
+
+/* Initialise and register driver with DPDK Application */
+static const struct eth_dev_ops nfp_net_eth_dev_ops = {
+	.dev_configure		= nfp_net_configure,
+	.dev_start		= nfp_net_start,
+	.dev_stop		= nfp_net_stop,
+	.dev_set_link_up	= nfp_net_set_link_up,
+	.dev_set_link_down	= nfp_net_set_link_down,
+	.dev_close		= nfp_net_close,
+	.promiscuous_enable	= nfp_net_promisc_enable,
+	.promiscuous_disable	= nfp_net_promisc_disable,
+	.link_update		= nfp_net_link_update,
+	.stats_get		= nfp_net_stats_get,
+	.stats_reset		= nfp_net_stats_reset,
+	.dev_infos_get		= nfp_net_infos_get,
+	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
+	.mtu_set		= nfp_net_dev_mtu_set,
+	.mac_addr_set           = nfp_set_mac_addr,
+	.vlan_offload_set	= nfp_net_vlan_offload_set,
+	.reta_update		= nfp_net_reta_update,
+	.reta_query		= nfp_net_reta_query,
+	.rss_hash_update	= nfp_net_rss_hash_update,
+	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
+	.rx_queue_setup		= nfp_net_rx_queue_setup,
+	.rx_queue_release	= nfp_net_rx_queue_release,
+	.tx_queue_setup		= nfp_net_tx_queue_setup,
+	.tx_queue_release	= nfp_net_tx_queue_release,
+	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
+	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
+};
+
+static int
+nfp_net_init(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_pf_dev *pf_dev;
+	struct nfp_net_hw *hw;
+	struct rte_ether_addr *tmp_ether_addr;
+
+	uint64_t tx_bar_off = 0, rx_bar_off = 0;
+	uint32_t start_q;
+	int stride = 4;
+	int port = 0;
+	int err;
+
+	PMD_INIT_FUNC_TRACE();
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* Use backpointer here to the PF of this eth_dev */
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(eth_dev->data->dev_private);
+
+	/* NFP can not handle DMA addresses requiring more than 40 bits */
+	if (rte_mem_check_dma_mask(40)) {
+		RTE_LOG(ERR, PMD, "device %s can not be used:",
+				   pci_dev->device.name);
+		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
+		return -ENODEV;
+	};
+
+	port = ((struct nfp_net_hw *)eth_dev->data->dev_private)->idx;
+	if (port < 0 || port > 7) {
+		PMD_DRV_LOG(ERR, "Port value is wrong");
+		return -ENODEV;
+	}
+
+	/* Use PF array of physical ports to get pointer to
+	 * this specific port
+	 */
+	hw = pf_dev->ports[port];
+
+	PMD_INIT_LOG(DEBUG, "Working with physical port number: %d, "
+			    "NFP internal port number: %d",
+			    port, hw->nfp_idx);
+
+	eth_dev->dev_ops = &nfp_net_eth_dev_ops;
+	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	rte_eth_copy_pci_info(eth_dev, pci_dev);
+
+	hw->device_id = pci_dev->id.device_id;
+	hw->vendor_id = pci_dev->id.vendor_id;
+	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
+	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
+
+	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
+		     pci_dev->id.vendor_id, pci_dev->id.device_id,
+		     pci_dev->addr.domain, pci_dev->addr.bus,
+		     pci_dev->addr.devid, pci_dev->addr.function);
+
+	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
+	if (hw->ctrl_bar == NULL) {
+		PMD_DRV_LOG(ERR,
+			"hw->ctrl_bar is NULL. BAR0 not configured");
+		return -ENODEV;
+	}
+
+	if (port == 0) {
+		hw->ctrl_bar = pf_dev->ctrl_bar;
+	} else {
+		if (!pf_dev->ctrl_bar)
+			return -ENODEV;
+		/* Use port offset in pf ctrl_bar for this
+		 * ports control bar
+		 */
+		hw->ctrl_bar = pf_dev->ctrl_bar +
+			       (port * NFP_PF_CSR_SLICE_SIZE);
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+
+	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
+	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
+
+	/* Work out where in the BAR the queues start. */
+	switch (pci_dev->id.device_id) {
+	case PCI_DEVICE_ID_NFP4000_PF_NIC:
+	case PCI_DEVICE_ID_NFP6000_PF_NIC:
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
+		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
+		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
+		err = -ENODEV;
+		goto dev_err_ctrl_map;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
+	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
+
+	hw->tx_bar = pf_dev->hw_queues + tx_bar_off;
+	hw->rx_bar = pf_dev->hw_queues + rx_bar_off;
+	eth_dev->data->dev_private = hw;
+
+	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
+		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
+
+	nfp_net_cfg_queue_setup(hw);
+
+	/* Get some of the read-only fields from the config BAR */
+	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
+	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
+	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
+	hw->mtu = RTE_ETHER_MTU;
+
+	/* VLAN insertion is incompatible with LSOv2 */
+	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
+		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
+
+	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
+		hw->rx_offset = NFP_NET_RX_OFFSET;
+	else
+		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
+
+	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
+			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
+			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
+
+	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
+		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
+
+	hw->ctrl = 0;
+
+	hw->stride_rx = stride;
+	hw->stride_tx = stride;
+
+	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
+		     hw->max_rx_queues, hw->max_tx_queues);
+
+	/* Initializing spinlock for reconfigs */
+	rte_spinlock_init(&hw->reconfig_lock);
+
+	/* Allocating memory for mac addr */
+	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
+					       RTE_ETHER_ADDR_LEN, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
+		err = -ENOMEM;
+		goto dev_err_queues_map;
+	}
+
+	nfp_net_pf_read_mac(pf_dev, port);
+	nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+
+	tmp_ether_addr = (struct rte_ether_addr *)&hw->mac_addr;
+	if (!rte_is_valid_assigned_ether_addr(tmp_ether_addr)) {
+		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
+				   port);
+		/* Using random mac addresses for VFs */
+		rte_eth_random_addr(&hw->mac_addr[0]);
+		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+	}
+
+	/* Copying mac address to DPDK eth_dev struct */
+	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
+			&eth_dev->data->mac_addrs[0]);
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
+
+	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
+
+	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
+		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
+		     eth_dev->data->port_id, pci_dev->id.vendor_id,
+		     pci_dev->id.device_id,
+		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
+		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		/* Registering LSC interrupt handler */
+		rte_intr_callback_register(&pci_dev->intr_handle,
+					   nfp_net_dev_interrupt_handler,
+					   (void *)eth_dev);
+		/* Telling the firmware about the LSC interrupt entry */
+		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
+		/* Recording current stats counters values */
+		nfp_net_stats_reset(eth_dev);
+	}
+
+	return 0;
+
+dev_err_queues_map:
+		nfp_cpp_area_free(hw->hwqueues_area);
+dev_err_ctrl_map:
+		nfp_cpp_area_free(hw->ctrl_area);
+
+	return err;
+}
+
+#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
+
+static int
+nfp_fw_upload(struct rte_pci_device *dev, struct nfp_nsp *nsp, char *card)
+{
+	struct nfp_cpp *cpp = nsp->cpp;
+	int fw_f;
+	char *fw_buf;
+	char fw_name[125];
+	char serial[40];
+	struct stat file_stat;
+	off_t fsize, bytes;
+
+	/* Looking for firmware file in order of priority */
+
+	/* First try to find a firmware image specific for this device */
+	snprintf(serial, sizeof(serial),
+			"serial-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x",
+		cpp->serial[0], cpp->serial[1], cpp->serial[2], cpp->serial[3],
+		cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
+		cpp->interface & 0xff);
+
+	snprintf(fw_name, sizeof(fw_name), "%s/%s.nffw", DEFAULT_FW_PATH,
+			serial);
+
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f >= 0)
+		goto read_fw;
+
+	/* Then try the PCI name */
+	snprintf(fw_name, sizeof(fw_name), "%s/pci-%s.nffw", DEFAULT_FW_PATH,
+			dev->device.name);
+
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f >= 0)
+		goto read_fw;
+
+	/* Finally try the card type and media */
+	snprintf(fw_name, sizeof(fw_name), "%s/%s", DEFAULT_FW_PATH, card);
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	fw_f = open(fw_name, O_RDONLY);
+	if (fw_f < 0) {
+		PMD_DRV_LOG(INFO, "Firmware file %s not found.", fw_name);
+		return -ENOENT;
+	}
+
+read_fw:
+	if (fstat(fw_f, &file_stat) < 0) {
+		PMD_DRV_LOG(INFO, "Firmware file %s size is unknown", fw_name);
+		close(fw_f);
+		return -ENOENT;
+	}
+
+	fsize = file_stat.st_size;
+	PMD_DRV_LOG(INFO, "Firmware file found at %s with size: %" PRIu64 "",
+			    fw_name, (uint64_t)fsize);
+
+	fw_buf = malloc((size_t)fsize);
+	if (!fw_buf) {
+		PMD_DRV_LOG(INFO, "malloc failed for fw buffer");
+		close(fw_f);
+		return -ENOMEM;
+	}
+	memset(fw_buf, 0, fsize);
+
+	bytes = read(fw_f, fw_buf, fsize);
+	if (bytes != fsize) {
+		PMD_DRV_LOG(INFO, "Reading fw to buffer failed."
+				   "Just %" PRIu64 " of %" PRIu64 " bytes read",
+				   (uint64_t)bytes, (uint64_t)fsize);
+		free(fw_buf);
+		close(fw_f);
+		return -EIO;
+	}
+
+	PMD_DRV_LOG(INFO, "Uploading the firmware ...");
+	nfp_nsp_load_fw(nsp, fw_buf, bytes);
+	PMD_DRV_LOG(INFO, "Done");
+
+	free(fw_buf);
+	close(fw_f);
+
+	return 0;
+}
+
+static int
+nfp_fw_setup(struct rte_pci_device *dev, struct nfp_cpp *cpp,
+	     struct nfp_eth_table *nfp_eth_table, struct nfp_hwinfo *hwinfo)
+{
+	struct nfp_nsp *nsp;
+	const char *nfp_fw_model;
+	char card_desc[100];
+	int err = 0;
+
+	nfp_fw_model = nfp_hwinfo_lookup(hwinfo, "assembly.partno");
+
+	if (nfp_fw_model) {
+		PMD_DRV_LOG(INFO, "firmware model found: %s", nfp_fw_model);
+	} else {
+		PMD_DRV_LOG(ERR, "firmware model NOT found");
+		return -EIO;
+	}
+
+	if (nfp_eth_table->count == 0 || nfp_eth_table->count > 8) {
+		PMD_DRV_LOG(ERR, "NFP ethernet table reports wrong ports: %u",
+		       nfp_eth_table->count);
+		return -EIO;
+	}
+
+	PMD_DRV_LOG(INFO, "NFP ethernet port table reports %u ports",
+			   nfp_eth_table->count);
+
+	PMD_DRV_LOG(INFO, "Port speed: %u", nfp_eth_table->ports[0].speed);
+
+	snprintf(card_desc, sizeof(card_desc), "nic_%s_%dx%d.nffw",
+			nfp_fw_model, nfp_eth_table->count,
+			nfp_eth_table->ports[0].speed / 1000);
+
+	nsp = nfp_nsp_open(cpp);
+	if (!nsp) {
+		PMD_DRV_LOG(ERR, "NFP error when obtaining NSP handle");
+		return -EIO;
+	}
+
+	nfp_nsp_device_soft_reset(nsp);
+	err = nfp_fw_upload(dev, nsp, card_desc);
+
+	nfp_nsp_close(nsp);
+	return err;
+}
+
+static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
+{
+	struct nfp_net_hw *hw;
+	struct rte_eth_dev *eth_dev;
+	struct nfp_eth_table *nfp_eth_table = NULL;
+	int ret = 0;
+	int i;
+
+	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
+	if (!nfp_eth_table) {
+		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
+		ret = -EIO;
+		goto error;
+	}
+
+	/* Loop through all physical ports on PF */
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		const unsigned int numa_node = rte_socket_id();
+		char port_name[RTE_ETH_NAME_MAX_LEN];
+
+		snprintf(port_name, sizeof(port_name), "%s_port%d",
+			 pf_dev->pci_dev->device.name, i);
+
+		/* Allocate a eth_dev for this phyport */
+		eth_dev = rte_eth_dev_allocate(port_name);
+		if (!eth_dev) {
+			ret = -ENODEV;
+			goto port_cleanup;
+		}
+
+		/* Allocate memory for this phyport */
+		eth_dev->data->dev_private =
+			rte_zmalloc_socket(port_name, sizeof(struct nfp_net_hw),
+					   RTE_CACHE_LINE_SIZE, numa_node);
+		if (!eth_dev->data->dev_private) {
+			ret = -ENOMEM;
+			rte_eth_dev_release_port(eth_dev);
+			goto port_cleanup;
+		}
+
+		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+
+		/* Add this device to the PF's array of physical ports */
+		pf_dev->ports[i] = hw;
+
+		hw->pf_dev = pf_dev;
+		hw->cpp = pf_dev->cpp;
+		hw->eth_dev = eth_dev;
+		hw->idx = i;
+		hw->nfp_idx = nfp_eth_table->ports[i].index;
+		hw->is_phyport = true;
+
+		eth_dev->device = &pf_dev->pci_dev->device;
+
+		/* ctrl/tx/rx BAR mappings and remaining init happens in
+		 * nfp_net_init
+		 */
+		ret = nfp_net_init(eth_dev);
+
+		if (ret) {
+			ret = -ENODEV;
+			goto port_cleanup;
+		}
+
+		rte_eth_dev_probing_finish(eth_dev);
+
+	} /* End loop, all ports on this PF */
+	ret = 0;
+	goto eth_table_cleanup;
+
+port_cleanup:
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		if (pf_dev->ports[i] && pf_dev->ports[i]->eth_dev) {
+			struct rte_eth_dev *tmp_dev;
+			tmp_dev = pf_dev->ports[i]->eth_dev;
+			rte_eth_dev_release_port(tmp_dev);
+			pf_dev->ports[i] = NULL;
+		}
+	}
+eth_table_cleanup:
+	free(nfp_eth_table);
+error:
+	return ret;
+}
+
+static int nfp_pf_init(struct rte_pci_device *pci_dev)
+{
+	struct nfp_pf_dev *pf_dev = NULL;
+	struct nfp_cpp *cpp;
+	struct nfp_hwinfo *hwinfo;
+	struct nfp_rtsym_table *sym_tbl;
+	struct nfp_eth_table *nfp_eth_table = NULL;
+	char name[RTE_ETH_NAME_MAX_LEN];
+	int total_ports;
+	int ret = -ENODEV;
+	int err;
+
+	if (!pci_dev)
+		return ret;
+
+	/*
+	 * When device bound to UIO, the device could be used, by mistake,
+	 * by two DPDK apps, and the UIO driver does not avoid it. This
+	 * could lead to a serious problem when configuring the NFP CPP
+	 * interface. Here we avoid this telling to the CPP init code to
+	 * use a lock file if UIO is being used.
+	 */
+	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
+		cpp = nfp_cpp_from_device_name(pci_dev, 0);
+	else
+		cpp = nfp_cpp_from_device_name(pci_dev, 1);
+
+	if (!cpp) {
+		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
+		ret = -EIO;
+		goto error;
+	}
+
+	hwinfo = nfp_hwinfo_read(cpp);
+	if (!hwinfo) {
+		PMD_INIT_LOG(ERR, "Error reading hwinfo table");
+		ret = -EIO;
+		goto error;
+	}
+
+	nfp_eth_table = nfp_eth_read_ports(cpp);
+	if (!nfp_eth_table) {
+		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
+		ret = -EIO;
+		goto hwinfo_cleanup;
+	}
+
+	if (nfp_fw_setup(pci_dev, cpp, nfp_eth_table, hwinfo)) {
+		PMD_INIT_LOG(ERR, "Error when uploading firmware");
+		ret = -EIO;
+		goto eth_table_cleanup;
+	}
+
+	/* Now the symbol table should be there */
+	sym_tbl = nfp_rtsym_table_read(cpp);
+	if (!sym_tbl) {
+		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
+				" symbol table");
+		ret = -EIO;
+		goto eth_table_cleanup;
+	}
+
+	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
+	if (total_ports != (int)nfp_eth_table->count) {
+		PMD_DRV_LOG(ERR, "Inconsistent number of ports");
+		ret = -EIO;
+		goto sym_tbl_cleanup;
+	}
+
+	PMD_INIT_LOG(INFO, "Total physical ports: %d", total_ports);
+
+	if (total_ports <= 0 || total_ports > 8) {
+		PMD_INIT_LOG(ERR, "nfd_cfg_pf0_num_ports symbol with wrong value");
+		ret = -ENODEV;
+		goto sym_tbl_cleanup;
+	}
+	/* Allocate memory for the PF "device" */
+	snprintf(name, sizeof(name), "nfp_pf%d", 0);
+	pf_dev = rte_zmalloc(name, sizeof(*pf_dev), 0);
+	if (!pf_dev) {
+		ret = -ENOMEM;
+		goto sym_tbl_cleanup;
+	}
+
+	/* Populate the newly created PF device */
+	pf_dev->cpp = cpp;
+	pf_dev->hwinfo = hwinfo;
+	pf_dev->sym_tbl = sym_tbl;
+	pf_dev->total_phyports = total_ports;
+
+	if (total_ports > 1)
+		pf_dev->multiport = true;
+
+	pf_dev->pci_dev = pci_dev;
+
+	/* Map the symbol table */
+	pf_dev->ctrl_bar = nfp_rtsym_map(pf_dev->sym_tbl, "_pf0_net_bar0",
+				     pf_dev->total_phyports * 32768,
+				     &pf_dev->ctrl_area);
+	if (!pf_dev->ctrl_bar) {
+		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for _pf0_net_ctrl_bar");
+		ret = -EIO;
+		goto pf_cleanup;
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", pf_dev->ctrl_bar);
+
+	/* configure access to tx/rx vNIC BARs */
+	pf_dev->hw_queues = nfp_cpp_map_area(pf_dev->cpp, 0, 0,
+					      NFP_PCIE_QUEUE(0),
+					      NFP_QCP_QUEUE_AREA_SZ,
+					      &pf_dev->hwqueues_area);
+	if (!pf_dev->hw_queues) {
+		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for net.qc");
+		ret = -EIO;
+		goto ctrl_area_cleanup;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p", pf_dev->hw_queues);
+
+	/* Initialize and prep physical ports now
+	 * This will loop through all physical ports
+	 */
+	ret = nfp_init_phyports(pf_dev);
+	if (ret) {
+		PMD_INIT_LOG(ERR, "Could not create physical ports");
+		goto hwqueues_cleanup;
+	}
+
+	/* register the CPP bridge service here for primary use */
+	nfp_register_cpp_service(pf_dev->cpp);
+
+	return 0;
+
+hwqueues_cleanup:
+	nfp_cpp_area_free(pf_dev->hwqueues_area);
+ctrl_area_cleanup:
+	nfp_cpp_area_free(pf_dev->ctrl_area);
+pf_cleanup:
+	rte_free(pf_dev);
+sym_tbl_cleanup:
+	free(sym_tbl);
+eth_table_cleanup:
+	free(nfp_eth_table);
+hwinfo_cleanup:
+	free(hwinfo);
+error:
+	return ret;
+}
+
+/*
+ * When attaching to the NFP4000/6000 PF on a secondary process there
+ * is no need to initilise the PF again. Only minimal work is required
+ * here
+ */
+static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev)
+{
+	struct nfp_cpp *cpp;
+	struct nfp_rtsym_table *sym_tbl;
+	int total_ports;
+	int i;
+	int err;
+
+	if (!pci_dev)
+		return -ENODEV;
+
+	/*
+	 * When device bound to UIO, the device could be used, by mistake,
+	 * by two DPDK apps, and the UIO driver does not avoid it. This
+	 * could lead to a serious problem when configuring the NFP CPP
+	 * interface. Here we avoid this telling to the CPP init code to
+	 * use a lock file if UIO is being used.
+	 */
+	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
+		cpp = nfp_cpp_from_device_name(pci_dev, 0);
+	else
+		cpp = nfp_cpp_from_device_name(pci_dev, 1);
+
+	if (!cpp) {
+		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
+		return -EIO;
+	}
+
+	/*
+	 * We don't have access to the PF created in the primary process
+	 * here so we have to read the number of ports from firmware
+	 */
+	sym_tbl = nfp_rtsym_table_read(cpp);
+	if (!sym_tbl) {
+		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
+				" symbol table");
+		return -EIO;
+	}
+
+	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
+
+	for (i = 0; i < total_ports; i++) {
+		struct rte_eth_dev *eth_dev;
+		char port_name[RTE_ETH_NAME_MAX_LEN];
+
+		snprintf(port_name, sizeof(port_name), "%s_port%d",
+			 pci_dev->device.name, i);
+
+		PMD_DRV_LOG(DEBUG, "Secondary attaching to port %s",
+		    port_name);
+		eth_dev = rte_eth_dev_attach_secondary(port_name);
+		if (!eth_dev) {
+			RTE_LOG(ERR, EAL,
+			"secondary process attach failed, "
+			"ethdev doesn't exist");
+			return -ENODEV;
+		}
+		eth_dev->process_private = cpp;
+		eth_dev->dev_ops = &nfp_net_eth_dev_ops;
+		eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+		eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+		eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+		rte_eth_dev_probing_finish(eth_dev);
+	}
+
+	/* Register the CPP bridge service for the secondary too */
+	nfp_register_cpp_service(cpp);
+
+	return 0;
+}
+
+static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+			    struct rte_pci_device *dev)
+{
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		return nfp_pf_init(dev);
+	else
+		return nfp_pf_secondary_init(dev);
+}
+
+static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP4000_PF_NIC)
+	},
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP6000_PF_NIC)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	uint16_t port_id;
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* Free up all physical ports under PF */
+	RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
+		rte_eth_dev_close(port_id);
+	/*
+	 * Ports can be closed and freed but hotplugging is not
+	 * currently supported
+	 */
+	return -ENOTSUP;
+}
+
+static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
+}
+
+static struct rte_pci_driver rte_nfp_net_pf_pmd = {
+	.id_table = pci_id_nfp_pf_net_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	.probe = nfp_pf_pci_probe,
+	.remove = eth_nfp_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index a1722886ba..a6097eaab0 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -37,10 +37,10 @@
 #include "nfpcore/nfp_rtsym.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 #include "nfp_cpp_bridge.h"
 
 #include <sys/types.h>
@@ -51,20 +51,6 @@
 #include <sys/ioctl.h>
 #include <errno.h>
 
-/* Prototypes */
-static int nfp_net_close(struct rte_eth_dev *dev);
-static int nfp_net_init(struct rte_eth_dev *eth_dev);
-static int nfp_pf_init(struct rte_pci_device *pci_dev);
-static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
-static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
-static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
-static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
-static int nfp_net_stop(struct rte_eth_dev *dev);
-static int nfp_fw_setup(struct rte_pci_device *dev,
-			struct nfp_cpp *cpp,
-			struct nfp_eth_table *nfp_eth_table,
-			struct nfp_hwinfo *hwinfo);
-
 static int
 __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
 {
@@ -266,24 +252,6 @@ nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
 		dst[i] = src[i];
 }
 
-static int
-nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
-{
-	struct nfp_eth_table *nfp_eth_table;
-	struct nfp_net_hw *hw = NULL;
-
-	/* Grab a pointer to the correct physical port */
-	hw = pf_dev->ports[port];
-
-	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
-
-	nfp_eth_copy_mac((uint8_t *)&hw->mac_addr,
-			 (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
-
-	free(nfp_eth_table);
-	return 0;
-}
-
 void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
@@ -436,282 +404,6 @@ nfp_check_offloads(struct rte_eth_dev *dev)
 	return ctrl;
 }
 
-static int
-nfp_net_start(struct rte_eth_dev *dev)
-{
-	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
-	uint32_t new_ctrl, update = 0;
-	struct nfp_net_hw *hw;
-	struct nfp_pf_dev *pf_dev;
-	struct rte_eth_conf *dev_conf;
-	struct rte_eth_rxmode *rxmode;
-	uint32_t intr_vector;
-	int ret;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
-
-	PMD_INIT_LOG(DEBUG, "Start");
-
-	/* Disabling queues just in case... */
-	nfp_net_disable_queues(dev);
-
-	/* Enabling the required queues in the device */
-	nfp_net_enable_queues(dev);
-
-	/* check and configure queue intr-vector mapping */
-	if (dev->data->dev_conf.intr_conf.rxq != 0) {
-		if (pf_dev->multiport) {
-			PMD_INIT_LOG(ERR, "PMD rx interrupt is not supported "
-					  "with NFP multiport PF");
-				return -EINVAL;
-		}
-		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
-			/*
-			 * Better not to share LSC with RX interrupts.
-			 * Unregistering LSC interrupt handler
-			 */
-			rte_intr_callback_unregister(&pci_dev->intr_handle,
-				nfp_net_dev_interrupt_handler, (void *)dev);
-
-			if (dev->data->nb_rx_queues > 1) {
-				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
-					     "supports 1 queue with UIO");
-				return -EIO;
-			}
-		}
-		intr_vector = dev->data->nb_rx_queues;
-		if (rte_intr_efd_enable(intr_handle, intr_vector))
-			return -1;
-
-		nfp_configure_rx_interrupt(dev, intr_handle);
-		update = NFP_NET_CFG_UPDATE_MSIX;
-	}
-
-	rte_intr_enable(intr_handle);
-
-	new_ctrl = nfp_check_offloads(dev);
-
-	/* Writing configuration parameters in the device */
-	nfp_net_params_setup(hw);
-
-	dev_conf = &dev->data->dev_conf;
-	rxmode = &dev_conf->rxmode;
-
-	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
-		nfp_net_rss_config_default(dev);
-		update |= NFP_NET_CFG_UPDATE_RSS;
-		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
-	}
-
-	/* Enable device */
-	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
-
-	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
-
-	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
-		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
-
-	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
-	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
-		return -EIO;
-
-	/*
-	 * Allocating rte mbufs for configured rx queues.
-	 * This requires queues being enabled before
-	 */
-	if (nfp_net_rx_freelist_setup(dev) < 0) {
-		ret = -ENOMEM;
-		goto error;
-	}
-
-	if (hw->is_phyport) {
-		if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-			/* Configure the physical port up */
-			nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
-		else
-			nfp_eth_set_configured(dev->process_private,
-					       hw->nfp_idx, 1);
-	}
-
-	hw->ctrl = new_ctrl;
-
-	return 0;
-
-error:
-	/*
-	 * An error returned by this function should mean the app
-	 * exiting and then the system releasing all the memory
-	 * allocated even memory coming from hugepages.
-	 *
-	 * The device could be enabled at this point with some queues
-	 * ready for getting packets. This is true if the call to
-	 * nfp_net_rx_freelist_setup() succeeds for some queues but
-	 * fails for subsequent queues.
-	 *
-	 * This should make the app exiting but better if we tell the
-	 * device first.
-	 */
-	nfp_net_disable_queues(dev);
-
-	return ret;
-}
-
-/* Stop device: disable rx and tx functions to allow for reconfiguring. */
-static int
-nfp_net_stop(struct rte_eth_dev *dev)
-{
-	int i;
-	struct nfp_net_hw *hw;
-
-	PMD_INIT_LOG(DEBUG, "Stop");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	nfp_net_disable_queues(dev);
-
-	/* Clear queues */
-	for (i = 0; i < dev->data->nb_tx_queues; i++) {
-		nfp_net_reset_tx_queue(
-			(struct nfp_net_txq *)dev->data->tx_queues[i]);
-	}
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		nfp_net_reset_rx_queue(
-			(struct nfp_net_rxq *)dev->data->rx_queues[i]);
-	}
-
-	if (hw->is_phyport) {
-		if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-			/* Configure the physical port down */
-			nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
-		else
-			nfp_eth_set_configured(dev->process_private,
-					       hw->nfp_idx, 0);
-	}
-
-	return 0;
-}
-
-/* Set the link up. */
-static int
-nfp_net_set_link_up(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-
-	PMD_DRV_LOG(DEBUG, "Set link up");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	if (!hw->is_phyport)
-		return -ENOTSUP;
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		/* Configure the physical port down */
-		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
-	else
-		return nfp_eth_set_configured(dev->process_private,
-					      hw->nfp_idx, 1);
-}
-
-/* Set the link down. */
-static int
-nfp_net_set_link_down(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-
-	PMD_DRV_LOG(DEBUG, "Set link down");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	if (!hw->is_phyport)
-		return -ENOTSUP;
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		/* Configure the physical port down */
-		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
-	else
-		return nfp_eth_set_configured(dev->process_private,
-					      hw->nfp_idx, 0);
-}
-
-/* Reset and stop device. The device can not be restarted. */
-static int
-nfp_net_close(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-	struct rte_pci_device *pci_dev;
-	int i;
-
-	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
-		return 0;
-
-	PMD_INIT_LOG(DEBUG, "Close");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-
-	/*
-	 * We assume that the DPDK application is stopping all the
-	 * threads/queues before calling the device close function.
-	 */
-
-	nfp_net_disable_queues(dev);
-
-	/* Clear queues */
-	for (i = 0; i < dev->data->nb_tx_queues; i++) {
-		nfp_net_reset_tx_queue(
-			(struct nfp_net_txq *)dev->data->tx_queues[i]);
-	}
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		nfp_net_reset_rx_queue(
-			(struct nfp_net_rxq *)dev->data->rx_queues[i]);
-	}
-
-	/* Only free PF resources after all physical ports have been closed */
-	if (pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC ||
-	    pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC) {
-		struct nfp_pf_dev *pf_dev;
-		pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
-
-		/* Mark this port as unused and free device priv resources*/
-		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
-		pf_dev->ports[hw->idx] = NULL;
-		rte_eth_dev_release_port(dev);
-
-		for (i = 0; i < pf_dev->total_phyports; i++) {
-			/* Check to see if ports are still in use */
-			if (pf_dev->ports[i])
-				return 0;
-		}
-
-		/* Now it is safe to free all PF resources */
-		PMD_INIT_LOG(INFO, "Freeing PF resources");
-		nfp_cpp_area_free(pf_dev->ctrl_area);
-		nfp_cpp_area_free(pf_dev->hwqueues_area);
-		free(pf_dev->hwinfo);
-		free(pf_dev->sym_tbl);
-		nfp_cpp_free(pf_dev->cpp);
-		rte_free(pf_dev);
-	}
-
-	rte_intr_disable(&pci_dev->intr_handle);
-
-	/* unregister callback func from eal lib */
-	rte_intr_callback_unregister(&pci_dev->intr_handle,
-				     nfp_net_dev_interrupt_handler,
-				     (void *)dev);
-
-	/*
-	 * The ixgbe PMD driver disables the pcie master on the
-	 * device. The i40e does not...
-	 */
-
-	return 0;
-}
-
 int
 nfp_net_promisc_enable(struct rte_eth_dev *dev)
 {
@@ -1620,780 +1312,6 @@ nfp_net_rss_config_default(struct rte_eth_dev *dev)
 	return ret;
 }
 
-
-/* Initialise and register driver with DPDK Application */
-static const struct eth_dev_ops nfp_net_eth_dev_ops = {
-	.dev_configure		= nfp_net_configure,
-	.dev_start		= nfp_net_start,
-	.dev_stop		= nfp_net_stop,
-	.dev_set_link_up	= nfp_net_set_link_up,
-	.dev_set_link_down	= nfp_net_set_link_down,
-	.dev_close		= nfp_net_close,
-	.promiscuous_enable	= nfp_net_promisc_enable,
-	.promiscuous_disable	= nfp_net_promisc_disable,
-	.link_update		= nfp_net_link_update,
-	.stats_get		= nfp_net_stats_get,
-	.stats_reset		= nfp_net_stats_reset,
-	.dev_infos_get		= nfp_net_infos_get,
-	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
-	.mtu_set		= nfp_net_dev_mtu_set,
-	.mac_addr_set           = nfp_set_mac_addr,
-	.vlan_offload_set	= nfp_net_vlan_offload_set,
-	.reta_update		= nfp_net_reta_update,
-	.reta_query		= nfp_net_reta_query,
-	.rss_hash_update	= nfp_net_rss_hash_update,
-	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
-	.rx_queue_setup		= nfp_net_rx_queue_setup,
-	.rx_queue_release	= nfp_net_rx_queue_release,
-	.tx_queue_setup		= nfp_net_tx_queue_setup,
-	.tx_queue_release	= nfp_net_tx_queue_release,
-	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
-	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
-};
-
-
-static int
-nfp_net_init(struct rte_eth_dev *eth_dev)
-{
-	struct rte_pci_device *pci_dev;
-	struct nfp_pf_dev *pf_dev;
-	struct nfp_net_hw *hw;
-
-	uint64_t tx_bar_off = 0, rx_bar_off = 0;
-	uint32_t start_q;
-	int stride = 4;
-	int port = 0;
-	int err;
-
-	PMD_INIT_FUNC_TRACE();
-
-	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
-
-	/* Use backpointer here to the PF of this eth_dev */
-	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(eth_dev->data->dev_private);
-
-	/* NFP can not handle DMA addresses requiring more than 40 bits */
-	if (rte_mem_check_dma_mask(40)) {
-		RTE_LOG(ERR, PMD, "device %s can not be used:",
-				   pci_dev->device.name);
-		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
-		return -ENODEV;
-	};
-
-	if ((pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC) ||
-	    (pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC)) {
-		port = ((struct nfp_net_hw *)eth_dev->data->dev_private)->idx;
-		if (port < 0 || port > 7) {
-			PMD_DRV_LOG(ERR, "Port value is wrong");
-			return -ENODEV;
-		}
-
-		/* Use PF array of physical ports to get pointer to
-		 * this specific port
-		 */
-		hw = pf_dev->ports[port];
-
-		PMD_INIT_LOG(DEBUG, "Working with physical port number: %d, "
-				    "NFP internal port number: %d",
-				    port, hw->nfp_idx);
-
-	} else {
-		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
-	}
-
-	eth_dev->dev_ops = &nfp_net_eth_dev_ops;
-	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
-	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
-	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
-
-	/* For secondary processes, the primary has done all the work */
-	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
-		return 0;
-
-	rte_eth_copy_pci_info(eth_dev, pci_dev);
-
-	hw->device_id = pci_dev->id.device_id;
-	hw->vendor_id = pci_dev->id.vendor_id;
-	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
-	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
-
-	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
-		     pci_dev->id.vendor_id, pci_dev->id.device_id,
-		     pci_dev->addr.domain, pci_dev->addr.bus,
-		     pci_dev->addr.devid, pci_dev->addr.function);
-
-	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
-	if (hw->ctrl_bar == NULL) {
-		PMD_DRV_LOG(ERR,
-			"hw->ctrl_bar is NULL. BAR0 not configured");
-		return -ENODEV;
-	}
-
-	if (hw->is_phyport) {
-		if (port == 0) {
-			hw->ctrl_bar = pf_dev->ctrl_bar;
-		} else {
-			if (!pf_dev->ctrl_bar)
-				return -ENODEV;
-			/* Use port offset in pf ctrl_bar for this
-			 * ports control bar
-			 */
-			hw->ctrl_bar = pf_dev->ctrl_bar +
-				       (port * NFP_PF_CSR_SLICE_SIZE);
-		}
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
-
-	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
-	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
-
-	/* Work out where in the BAR the queues start. */
-	switch (pci_dev->id.device_id) {
-	case PCI_DEVICE_ID_NFP4000_PF_NIC:
-	case PCI_DEVICE_ID_NFP6000_PF_NIC:
-	case PCI_DEVICE_ID_NFP6000_VF_NIC:
-		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
-		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
-		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
-		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
-		break;
-	default:
-		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
-		err = -ENODEV;
-		goto dev_err_ctrl_map;
-	}
-
-	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
-	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
-
-	if (hw->is_phyport) {
-		hw->tx_bar = pf_dev->hw_queues + tx_bar_off;
-		hw->rx_bar = pf_dev->hw_queues + rx_bar_off;
-		eth_dev->data->dev_private = hw;
-	} else {
-		hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
-			     tx_bar_off;
-		hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
-			     rx_bar_off;
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
-		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
-
-	nfp_net_cfg_queue_setup(hw);
-
-	/* Get some of the read-only fields from the config BAR */
-	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
-	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
-	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
-	hw->mtu = RTE_ETHER_MTU;
-
-	/* VLAN insertion is incompatible with LSOv2 */
-	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
-		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
-
-	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
-		hw->rx_offset = NFP_NET_RX_OFFSET;
-	else
-		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
-
-	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
-			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
-			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
-
-	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
-		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
-
-	hw->ctrl = 0;
-
-	hw->stride_rx = stride;
-	hw->stride_tx = stride;
-
-	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
-		     hw->max_rx_queues, hw->max_tx_queues);
-
-	/* Initializing spinlock for reconfigs */
-	rte_spinlock_init(&hw->reconfig_lock);
-
-	/* Allocating memory for mac addr */
-	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
-					       RTE_ETHER_ADDR_LEN, 0);
-	if (eth_dev->data->mac_addrs == NULL) {
-		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
-		err = -ENOMEM;
-		goto dev_err_queues_map;
-	}
-
-	if (hw->is_phyport) {
-		nfp_net_pf_read_mac(pf_dev, port);
-		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	}
-
-	if (!rte_is_valid_assigned_ether_addr(
-		    (struct rte_ether_addr *)&hw->mac_addr)) {
-		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
-				   port);
-		/* Using random mac addresses for VFs */
-		rte_eth_random_addr(&hw->mac_addr[0]);
-		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	}
-
-	/* Copying mac address to DPDK eth_dev struct */
-	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
-			&eth_dev->data->mac_addrs[0]);
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
-		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
-
-	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
-
-	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
-		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
-		     eth_dev->data->port_id, pci_dev->id.vendor_id,
-		     pci_dev->id.device_id,
-		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
-		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
-		/* Registering LSC interrupt handler */
-		rte_intr_callback_register(&pci_dev->intr_handle,
-					   nfp_net_dev_interrupt_handler,
-					   (void *)eth_dev);
-		/* Telling the firmware about the LSC interrupt entry */
-		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
-		/* Recording current stats counters values */
-		nfp_net_stats_reset(eth_dev);
-	}
-
-	return 0;
-
-dev_err_queues_map:
-		nfp_cpp_area_free(hw->hwqueues_area);
-dev_err_ctrl_map:
-		nfp_cpp_area_free(hw->ctrl_area);
-
-	return err;
-}
-
-#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
-
-static int
-nfp_fw_upload(struct rte_pci_device *dev, struct nfp_nsp *nsp, char *card)
-{
-	struct nfp_cpp *cpp = nsp->cpp;
-	int fw_f;
-	char *fw_buf;
-	char fw_name[125];
-	char serial[40];
-	struct stat file_stat;
-	off_t fsize, bytes;
-
-	/* Looking for firmware file in order of priority */
-
-	/* First try to find a firmware image specific for this device */
-	snprintf(serial, sizeof(serial),
-			"serial-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x",
-		cpp->serial[0], cpp->serial[1], cpp->serial[2], cpp->serial[3],
-		cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
-		cpp->interface & 0xff);
-
-	snprintf(fw_name, sizeof(fw_name), "%s/%s.nffw", DEFAULT_FW_PATH,
-			serial);
-
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	fw_f = open(fw_name, O_RDONLY);
-	if (fw_f >= 0)
-		goto read_fw;
-
-	/* Then try the PCI name */
-	snprintf(fw_name, sizeof(fw_name), "%s/pci-%s.nffw", DEFAULT_FW_PATH,
-			dev->device.name);
-
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	fw_f = open(fw_name, O_RDONLY);
-	if (fw_f >= 0)
-		goto read_fw;
-
-	/* Finally try the card type and media */
-	snprintf(fw_name, sizeof(fw_name), "%s/%s", DEFAULT_FW_PATH, card);
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	fw_f = open(fw_name, O_RDONLY);
-	if (fw_f < 0) {
-		PMD_DRV_LOG(INFO, "Firmware file %s not found.", fw_name);
-		return -ENOENT;
-	}
-
-read_fw:
-	if (fstat(fw_f, &file_stat) < 0) {
-		PMD_DRV_LOG(INFO, "Firmware file %s size is unknown", fw_name);
-		close(fw_f);
-		return -ENOENT;
-	}
-
-	fsize = file_stat.st_size;
-	PMD_DRV_LOG(INFO, "Firmware file found at %s with size: %" PRIu64 "",
-			    fw_name, (uint64_t)fsize);
-
-	fw_buf = malloc((size_t)fsize);
-	if (!fw_buf) {
-		PMD_DRV_LOG(INFO, "malloc failed for fw buffer");
-		close(fw_f);
-		return -ENOMEM;
-	}
-	memset(fw_buf, 0, fsize);
-
-	bytes = read(fw_f, fw_buf, fsize);
-	if (bytes != fsize) {
-		PMD_DRV_LOG(INFO, "Reading fw to buffer failed."
-				   "Just %" PRIu64 " of %" PRIu64 " bytes read",
-				   (uint64_t)bytes, (uint64_t)fsize);
-		free(fw_buf);
-		close(fw_f);
-		return -EIO;
-	}
-
-	PMD_DRV_LOG(INFO, "Uploading the firmware ...");
-	nfp_nsp_load_fw(nsp, fw_buf, bytes);
-	PMD_DRV_LOG(INFO, "Done");
-
-	free(fw_buf);
-	close(fw_f);
-
-	return 0;
-}
-
-static int
-nfp_fw_setup(struct rte_pci_device *dev, struct nfp_cpp *cpp,
-	     struct nfp_eth_table *nfp_eth_table, struct nfp_hwinfo *hwinfo)
-{
-	struct nfp_nsp *nsp;
-	const char *nfp_fw_model;
-	char card_desc[100];
-	int err = 0;
-
-	nfp_fw_model = nfp_hwinfo_lookup(hwinfo, "assembly.partno");
-
-	if (nfp_fw_model) {
-		PMD_DRV_LOG(INFO, "firmware model found: %s", nfp_fw_model);
-	} else {
-		PMD_DRV_LOG(ERR, "firmware model NOT found");
-		return -EIO;
-	}
-
-	if (nfp_eth_table->count == 0 || nfp_eth_table->count > 8) {
-		PMD_DRV_LOG(ERR, "NFP ethernet table reports wrong ports: %u",
-		       nfp_eth_table->count);
-		return -EIO;
-	}
-
-	PMD_DRV_LOG(INFO, "NFP ethernet port table reports %u ports",
-			   nfp_eth_table->count);
-
-	PMD_DRV_LOG(INFO, "Port speed: %u", nfp_eth_table->ports[0].speed);
-
-	snprintf(card_desc, sizeof(card_desc), "nic_%s_%dx%d.nffw",
-			nfp_fw_model, nfp_eth_table->count,
-			nfp_eth_table->ports[0].speed / 1000);
-
-	nsp = nfp_nsp_open(cpp);
-	if (!nsp) {
-		PMD_DRV_LOG(ERR, "NFP error when obtaining NSP handle");
-		return -EIO;
-	}
-
-	nfp_nsp_device_soft_reset(nsp);
-	err = nfp_fw_upload(dev, nsp, card_desc);
-
-	nfp_nsp_close(nsp);
-	return err;
-}
-
-static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
-{
-	struct nfp_net_hw *hw;
-	struct rte_eth_dev *eth_dev;
-	struct nfp_eth_table *nfp_eth_table = NULL;
-	int ret = 0;
-	int i;
-
-	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
-	if (!nfp_eth_table) {
-		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
-		ret = -EIO;
-		goto error;
-	}
-
-	/* Loop through all physical ports on PF */
-	for (i = 0; i < pf_dev->total_phyports; i++) {
-		const unsigned int numa_node = rte_socket_id();
-		char port_name[RTE_ETH_NAME_MAX_LEN];
-
-		snprintf(port_name, sizeof(port_name), "%s_port%d",
-			 pf_dev->pci_dev->device.name, i);
-
-		/* Allocate a eth_dev for this phyport */
-		eth_dev = rte_eth_dev_allocate(port_name);
-		if (!eth_dev) {
-			ret = -ENODEV;
-			goto port_cleanup;
-		}
-
-		/* Allocate memory for this phyport */
-		eth_dev->data->dev_private =
-			rte_zmalloc_socket(port_name, sizeof(struct nfp_net_hw),
-					   RTE_CACHE_LINE_SIZE, numa_node);
-		if (!eth_dev->data->dev_private) {
-			ret = -ENOMEM;
-			rte_eth_dev_release_port(eth_dev);
-			goto port_cleanup;
-		}
-
-		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
-
-		/* Add this device to the PF's array of physical ports */
-		pf_dev->ports[i] = hw;
-
-		hw->pf_dev = pf_dev;
-		hw->cpp = pf_dev->cpp;
-		hw->eth_dev = eth_dev;
-		hw->idx = i;
-		hw->nfp_idx = nfp_eth_table->ports[i].index;
-		hw->is_phyport = true;
-
-		eth_dev->device = &pf_dev->pci_dev->device;
-
-		/* ctrl/tx/rx BAR mappings and remaining init happens in
-		 * nfp_net_init
-		 */
-		ret = nfp_net_init(eth_dev);
-
-		if (ret) {
-			ret = -ENODEV;
-			goto port_cleanup;
-		}
-
-		rte_eth_dev_probing_finish(eth_dev);
-
-	} /* End loop, all ports on this PF */
-	ret = 0;
-	goto eth_table_cleanup;
-
-port_cleanup:
-	for (i = 0; i < pf_dev->total_phyports; i++) {
-		if (pf_dev->ports[i] && pf_dev->ports[i]->eth_dev) {
-			struct rte_eth_dev *tmp_dev;
-			tmp_dev = pf_dev->ports[i]->eth_dev;
-			rte_eth_dev_release_port(tmp_dev);
-			pf_dev->ports[i] = NULL;
-		}
-	}
-eth_table_cleanup:
-	free(nfp_eth_table);
-error:
-	return ret;
-}
-
-static int nfp_pf_init(struct rte_pci_device *pci_dev)
-{
-	struct nfp_pf_dev *pf_dev = NULL;
-	struct nfp_cpp *cpp;
-	struct nfp_hwinfo *hwinfo;
-	struct nfp_rtsym_table *sym_tbl;
-	struct nfp_eth_table *nfp_eth_table = NULL;
-	char name[RTE_ETH_NAME_MAX_LEN];
-	int total_ports;
-	int ret = -ENODEV;
-	int err;
-
-	if (!pci_dev)
-		return ret;
-
-	/*
-	 * When device bound to UIO, the device could be used, by mistake,
-	 * by two DPDK apps, and the UIO driver does not avoid it. This
-	 * could lead to a serious problem when configuring the NFP CPP
-	 * interface. Here we avoid this telling to the CPP init code to
-	 * use a lock file if UIO is being used.
-	 */
-	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
-		cpp = nfp_cpp_from_device_name(pci_dev, 0);
-	else
-		cpp = nfp_cpp_from_device_name(pci_dev, 1);
-
-	if (!cpp) {
-		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
-		ret = -EIO;
-		goto error;
-	}
-
-	hwinfo = nfp_hwinfo_read(cpp);
-	if (!hwinfo) {
-		PMD_INIT_LOG(ERR, "Error reading hwinfo table");
-		ret = -EIO;
-		goto error;
-	}
-
-	nfp_eth_table = nfp_eth_read_ports(cpp);
-	if (!nfp_eth_table) {
-		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
-		ret = -EIO;
-		goto hwinfo_cleanup;
-	}
-
-	if (nfp_fw_setup(pci_dev, cpp, nfp_eth_table, hwinfo)) {
-		PMD_INIT_LOG(ERR, "Error when uploading firmware");
-		ret = -EIO;
-		goto eth_table_cleanup;
-	}
-
-	/* Now the symbol table should be there */
-	sym_tbl = nfp_rtsym_table_read(cpp);
-	if (!sym_tbl) {
-		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
-				" symbol table");
-		ret = -EIO;
-		goto eth_table_cleanup;
-	}
-
-	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
-	if (total_ports != (int)nfp_eth_table->count) {
-		PMD_DRV_LOG(ERR, "Inconsistent number of ports");
-		ret = -EIO;
-		goto sym_tbl_cleanup;
-	}
-
-	PMD_INIT_LOG(INFO, "Total physical ports: %d", total_ports);
-
-	if (total_ports <= 0 || total_ports > 8) {
-		PMD_INIT_LOG(ERR, "nfd_cfg_pf0_num_ports symbol with wrong value");
-		ret = -ENODEV;
-		goto sym_tbl_cleanup;
-	}
-	/* Allocate memory for the PF "device" */
-	snprintf(name, sizeof(name), "nfp_pf%d", 0);
-	pf_dev = rte_zmalloc(name, sizeof(*pf_dev), 0);
-	if (!pf_dev) {
-		ret = -ENOMEM;
-		goto sym_tbl_cleanup;
-	}
-
-	/* Populate the newly created PF device */
-	pf_dev->cpp = cpp;
-	pf_dev->hwinfo = hwinfo;
-	pf_dev->sym_tbl = sym_tbl;
-	pf_dev->total_phyports = total_ports;
-
-	if (total_ports > 1)
-		pf_dev->multiport = true;
-
-	pf_dev->pci_dev = pci_dev;
-
-	/* Map the symbol table */
-	pf_dev->ctrl_bar = nfp_rtsym_map(pf_dev->sym_tbl, "_pf0_net_bar0",
-				     pf_dev->total_phyports * 32768,
-				     &pf_dev->ctrl_area);
-	if (!pf_dev->ctrl_bar) {
-		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for _pf0_net_ctrl_bar");
-		ret = -EIO;
-		goto pf_cleanup;
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", pf_dev->ctrl_bar);
-
-	/* configure access to tx/rx vNIC BARs */
-	pf_dev->hw_queues = nfp_cpp_map_area(pf_dev->cpp, 0, 0,
-					      NFP_PCIE_QUEUE(0),
-					      NFP_QCP_QUEUE_AREA_SZ,
-					      &pf_dev->hwqueues_area);
-	if (!pf_dev->hw_queues) {
-		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for net.qc");
-		ret = -EIO;
-		goto ctrl_area_cleanup;
-	}
-
-	PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p", pf_dev->hw_queues);
-
-	/* Initialize and prep physical ports now
-	 * This will loop through all physical ports
-	 */
-	ret = nfp_init_phyports(pf_dev);
-	if (ret) {
-		PMD_INIT_LOG(ERR, "Could not create physical ports");
-		goto hwqueues_cleanup;
-	}
-
-	/* register the CPP bridge service here for primary use */
-	nfp_register_cpp_service(pf_dev->cpp);
-
-	return 0;
-
-hwqueues_cleanup:
-	nfp_cpp_area_free(pf_dev->hwqueues_area);
-ctrl_area_cleanup:
-	nfp_cpp_area_free(pf_dev->ctrl_area);
-pf_cleanup:
-	rte_free(pf_dev);
-sym_tbl_cleanup:
-	free(sym_tbl);
-eth_table_cleanup:
-	free(nfp_eth_table);
-hwinfo_cleanup:
-	free(hwinfo);
-error:
-	return ret;
-}
-
-/*
- * When attaching to the NFP4000/6000 PF on a secondary process there
- * is no need to initialize the PF again. Only minimal work is required
- * here
- */
-static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev)
-{
-	struct nfp_cpp *cpp;
-	struct nfp_rtsym_table *sym_tbl;
-	int total_ports;
-	int i;
-	int err;
-
-	if (!pci_dev)
-		return -ENODEV;
-
-	/*
-	 * When device bound to UIO, the device could be used, by mistake,
-	 * by two DPDK apps, and the UIO driver does not avoid it. This
-	 * could lead to a serious problem when configuring the NFP CPP
-	 * interface. Here we avoid this telling to the CPP init code to
-	 * use a lock file if UIO is being used.
-	 */
-	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
-		cpp = nfp_cpp_from_device_name(pci_dev, 0);
-	else
-		cpp = nfp_cpp_from_device_name(pci_dev, 1);
-
-	if (!cpp) {
-		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
-		return -EIO;
-	}
-
-	/*
-	 * We don't have access to the PF created in the primary process
-	 * here so we have to read the number of ports from firmware
-	 */
-	sym_tbl = nfp_rtsym_table_read(cpp);
-	if (!sym_tbl) {
-		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
-				" symbol table");
-		return -EIO;
-	}
-
-	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
-
-	for (i = 0; i < total_ports; i++) {
-		struct rte_eth_dev *eth_dev;
-		char port_name[RTE_ETH_NAME_MAX_LEN];
-
-		snprintf(port_name, sizeof(port_name), "%s_port%d",
-			 pci_dev->device.name, i);
-
-		PMD_DRV_LOG(DEBUG, "Secondary attaching to port %s",
-		    port_name);
-		eth_dev = rte_eth_dev_attach_secondary(port_name);
-		if (!eth_dev) {
-			RTE_LOG(ERR, EAL,
-			"secondary process attach failed, "
-			"ethdev doesn't exist");
-			return -ENODEV;
-		}
-		eth_dev->process_private = cpp;
-		eth_dev->dev_ops = &nfp_net_eth_dev_ops;
-		eth_dev->rx_queue_count = nfp_net_rx_queue_count;
-		eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
-		eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
-		rte_eth_dev_probing_finish(eth_dev);
-	}
-
-	/* Register the CPP bridge service for the secondary too */
-	nfp_register_cpp_service(cpp);
-
-	return 0;
-}
-
-static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
-			    struct rte_pci_device *dev)
-{
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		return nfp_pf_init(dev);
-	else
-		return nfp_pf_secondary_init(dev);
-}
-
-static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP4000_PF_NIC)
-	},
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP6000_PF_NIC)
-	},
-	{
-		.vendor_id = 0,
-	},
-};
-
-static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
-{
-	struct rte_pci_device *pci_dev;
-	uint16_t port_id;
-
-	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
-
-	if (pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC ||
-	    pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC) {
-		/* Free up all physical ports under PF */
-		RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
-			rte_eth_dev_close(port_id);
-		/*
-		 * Ports can be closed and freed but hotplugging is not
-		 * currently supported
-		 */
-		return -ENOTSUP;
-	}
-
-	/* VF cleanup, just free private port data */
-	return nfp_net_close(eth_dev);
-}
-
-static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
-{
-	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
-}
-
-static struct rte_pci_driver rte_nfp_net_pf_pmd = {
-	.id_table = pci_id_nfp_pf_net_map,
-	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
-	.probe = nfp_pf_pci_probe,
-	.remove = eth_nfp_pci_remove,
-};
-
-RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
-RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
-RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_init, init, NOTICE);
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_driver, driver, NOTICE);
 /*
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v2 7/7] net/nfp: batch file rename for consistency
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
                     ` (5 preceding siblings ...)
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 6/7] net/nfp: move PF " Heinrich Kuhn
@ 2021-07-16  8:35   ` Heinrich Kuhn
  2021-07-23  9:18   ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Thomas Monjalon
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-16  8:35 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Rename the nfp_net.c file to nfp_common as it now contains functions
common to VF and PF functionality. Rename the header file too to be
consistent. Also remove the "net" naming from the _ctrl and _logs files
for consistency across the PMD.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build                     | 2 +-
 drivers/net/nfp/{nfp_net.c => nfp_common.c}     | 4 ++--
 drivers/net/nfp/{nfp_net_pmd.h => nfp_common.h} | 6 +++---
 drivers/net/nfp/nfp_cpp_bridge.c                | 2 +-
 drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h}  | 6 +++---
 drivers/net/nfp/nfp_ethdev.c                    | 6 +++---
 drivers/net/nfp/nfp_ethdev_vf.c                 | 6 +++---
 drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h}  | 6 +++---
 drivers/net/nfp/nfp_rxtx.c                      | 6 +++---
 9 files changed, 22 insertions(+), 22 deletions(-)
 rename drivers/net/nfp/{nfp_net.c => nfp_common.c} (99%)
 rename drivers/net/nfp/{nfp_net_pmd.h => nfp_common.h} (99%)
 rename drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h} (99%)
 rename drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h} (94%)

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index ab64d0cac3..810f02ae5b 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -18,7 +18,7 @@ sources = files(
         'nfpcore/nfp_mutex.c',
         'nfpcore/nfp_nsp_eth.c',
         'nfpcore/nfp_hwinfo.c',
-        'nfp_net.c',
+        'nfp_common.c',
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
         'nfp_ethdev_vf.c',
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_common.c
similarity index 99%
rename from drivers/net/nfp/nfp_net.c
rename to drivers/net/nfp/nfp_common.c
index a6097eaab0..87e8f5f333 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_common.c
@@ -8,9 +8,9 @@
 /*
  * vim:shiftwidth=8:noexpandtab
  *
- * @file dpdk/pmd/nfp_net.c
+ * @file dpdk/pmd/nfp_common.c
  *
- * Netronome vNIC DPDK Poll-Mode Driver: Main entry point
+ * Netronome vNIC DPDK Poll-Mode Driver: Common files
  */
 
 #include <rte_byteorder.h>
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_common.h
similarity index 99%
rename from drivers/net/nfp/nfp_net_pmd.h
rename to drivers/net/nfp/nfp_common.h
index dc05e888df..54ac937bd2 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_common.h
@@ -11,8 +11,8 @@
  * Netronome NFP_NET PMD driver
  */
 
-#ifndef _NFP_NET_PMD_H_
-#define _NFP_NET_PMD_H_
+#ifndef _NFP_COMMON_H_
+#define _NFP_COMMON_H_
 
 #define NFP_NET_PMD_VERSION "0.1"
 #define PCI_VENDOR_ID_NETRONOME         0x19ee
@@ -404,7 +404,7 @@ int nfp_net_rss_config_default(struct rte_eth_dev *dev);
 #define NFP_NET_DEV_PRIVATE_TO_PF(dev_priv)\
 	(((struct nfp_net_hw *)dev_priv)->pf_dev)
 
-#endif /* _NFP_NET_PMD_H_ */
+#endif /* _NFP_COMMON_H_ */
 /*
  * Local variables:
  * c-file-style: "Linux"
diff --git a/drivers/net/nfp/nfp_cpp_bridge.c b/drivers/net/nfp/nfp_cpp_bridge.c
index d916793338..74a0eacb3f 100644
--- a/drivers/net/nfp/nfp_cpp_bridge.c
+++ b/drivers/net/nfp/nfp_cpp_bridge.c
@@ -19,7 +19,7 @@
 #include "nfpcore/nfp_mip.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_logs.h"
+#include "nfp_logs.h"
 #include "nfp_cpp_bridge.h"
 
 #include <sys/ioctl.h>
diff --git a/drivers/net/nfp/nfp_net_ctrl.h b/drivers/net/nfp/nfp_ctrl.h
similarity index 99%
rename from drivers/net/nfp/nfp_net_ctrl.h
rename to drivers/net/nfp/nfp_ctrl.h
index 4f26ccf483..4dd62ef194 100644
--- a/drivers/net/nfp/nfp_net_ctrl.h
+++ b/drivers/net/nfp/nfp_ctrl.h
@@ -8,8 +8,8 @@
  *
  * Netronome network device driver: Control BAR layout
  */
-#ifndef _NFP_NET_CTRL_H_
-#define _NFP_NET_CTRL_H_
+#ifndef _NFP_CTRL_H_
+#define _NFP_CTRL_H_
 
 /*
  * Configuration BAR size.
@@ -317,7 +317,7 @@
 /* PF multiport offset */
 #define NFP_PF_CSR_SLICE_SIZE	(32 * 1024)
 
-#endif /* _NFP_NET_CTRL_H_ */
+#endif /* _NFP_CTRL_H_ */
 /*
  * Local variables:
  * c-file-style: "Linux"
diff --git a/drivers/net/nfp/nfp_ethdev.c b/drivers/net/nfp/nfp_ethdev.c
index ab08906704..29e6bb128a 100644
--- a/drivers/net/nfp/nfp_ethdev.c
+++ b/drivers/net/nfp/nfp_ethdev.c
@@ -30,10 +30,10 @@
 #include "nfpcore/nfp_rtsym.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 #include "nfp_cpp_bridge.h"
 
 
diff --git a/drivers/net/nfp/nfp_ethdev_vf.c b/drivers/net/nfp/nfp_ethdev_vf.c
index 223142c0ed..b697b55865 100644
--- a/drivers/net/nfp/nfp_ethdev_vf.c
+++ b/drivers/net/nfp/nfp_ethdev_vf.c
@@ -16,10 +16,10 @@
 #include "nfpcore/nfp_mip.h"
 #include "nfpcore/nfp_rtsym.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 
 static void nfp_netvf_read_mac(struct nfp_net_hw *hw);
 static int nfp_netvf_start(struct rte_eth_dev *dev);
diff --git a/drivers/net/nfp/nfp_net_logs.h b/drivers/net/nfp/nfp_logs.h
similarity index 94%
rename from drivers/net/nfp/nfp_net_logs.h
rename to drivers/net/nfp/nfp_logs.h
index 27dd87611b..bd5a5e1ec5 100644
--- a/drivers/net/nfp/nfp_net_logs.h
+++ b/drivers/net/nfp/nfp_logs.h
@@ -3,8 +3,8 @@
  * All rights reserved.
  */
 
-#ifndef _NFP_NET_LOGS_H_
-#define _NFP_NET_LOGS_H_
+#ifndef _NFP_LOGS_H_
+#define _NFP_LOGS_H_
 
 #include <rte_log.h>
 
@@ -44,4 +44,4 @@ extern int nfp_logtype_driver;
 	rte_log(RTE_LOG_ ## level, nfp_logtype_driver, \
 		"%s(): " fmt "\n", __func__, ## args)
 
-#endif /* _NFP_NET_LOGS_H_ */
+#endif /* _NFP_LOGS_H_ */
diff --git a/drivers/net/nfp/nfp_rxtx.c b/drivers/net/nfp/nfp_rxtx.c
index 9ee9e5c9a3..1402c5f84a 100644
--- a/drivers/net/nfp/nfp_rxtx.c
+++ b/drivers/net/nfp/nfp_rxtx.c
@@ -16,10 +16,10 @@
 #include <ethdev_driver.h>
 #include <ethdev_pci.h>
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 
 /* Prototypes */
 static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
                     ` (6 preceding siblings ...)
  2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
@ 2021-07-23  9:18   ` Thomas Monjalon
  2021-07-29 13:48     ` Heinrich Kuhn
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
  8 siblings, 1 reply; 31+ messages in thread
From: Thomas Monjalon @ 2021-07-23  9:18 UTC (permalink / raw)
  To: Heinrich Kuhn, Simon Horman; +Cc: dev, david.marchand

16/07/2021 10:35, Heinrich Kuhn:
> This patch set restructures the NFP PMD, aligning it more with the
> common layout adopted by most other PMD's. Although the changes look
> fairly large, functionally nothing is added or removed from the driver
> and the existing code is mostly just reorganized into the familiar
> structure seen in other PMD's.

It seems this refactoring is destroying this firmware improvement:
	https://git.dpdk.org/dpdk/commit/?id=40edb9c0d36b781

Please rebase carefully, keeping the behaviour really intact.

> Apart form adopting the common PMD layout
> this change should also aid in future feature development to the NFP
> PMD. The previous layout where most of the logic resided in a single
> file (nfp_net.c) would have become tedious to support going forward.



^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD
  2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
                     ` (7 preceding siblings ...)
  2021-07-23  9:18   ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Thomas Monjalon
@ 2021-07-29 13:47   ` Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
                       ` (8 more replies)
  8 siblings, 9 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn

This patch set restructures the NFP PMD, aligning it more with the
common layout adopted by most other PMD's. Although the changes look
fairly large, functionally nothing is added or removed from the driver
and the existing code is mostly just reorganized into the familiar
structure seen in other PMD's. Apart form adopting the common PMD layout
this change should also aid in future feature development to the NFP
PMD. The previous layout where most of the logic resided in a single
file (nfp_net.c) would have become tedious to support going forward.

v3:
* Avoid squashing the new firmware loader helper added in: 
  https://git.dpdk.org/dpdk/commit/?id=40edb9c0d36b781
* Add dependency to patch-93299

v2:
* Added missing sign-off's

---
Depends-on: patch-93299 ("net/nfp: remove compile time log")

Heinrich Kuhn (7):
  net/nfp: split rxtx headers into separate file
  net/nfp: move rxtx functions to their own file
  net/nfp: move CPP bridge to a separate file
  net/nfp: prototype common functions in header file
  net/nfp: move VF functions into new file
  net/nfp: move PF functions into new file
  net/nfp: batch file rename for consistency

 drivers/net/nfp/meson.build                   |    6 +-
 drivers/net/nfp/nfp_common.c                  | 1322 ++++++
 drivers/net/nfp/nfp_common.h                  |  413 ++
 drivers/net/nfp/nfp_cpp_bridge.c              |  392 ++
 drivers/net/nfp/nfp_cpp_bridge.h              |   36 +
 .../net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h}    |    6 +-
 drivers/net/nfp/nfp_ethdev.c                  | 1067 +++++
 drivers/net/nfp/nfp_ethdev_vf.c               |  504 +++
 .../net/nfp/{nfp_net_logs.h => nfp_logs.h}    |    6 +-
 drivers/net/nfp/nfp_net.c                     | 3889 -----------------
 drivers/net/nfp/nfp_rxtx.c                    | 1002 +++++
 drivers/net/nfp/{nfp_net_pmd.h => nfp_rxtx.h} |  279 +-
 12 files changed, 4783 insertions(+), 4139 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_common.c
 create mode 100644 drivers/net/nfp/nfp_common.h
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.c
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.h
 rename drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h} (99%)
 create mode 100644 drivers/net/nfp/nfp_ethdev.c
 create mode 100644 drivers/net/nfp/nfp_ethdev_vf.c
 rename drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h} (94%)
 delete mode 100644 drivers/net/nfp/nfp_net.c
 create mode 100644 drivers/net/nfp/nfp_rxtx.c
 rename drivers/net/nfp/{nfp_net_pmd.h => nfp_rxtx.h} (54%)

-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 1/7] net/nfp: split rxtx headers into separate file
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
@ 2021-07-29 13:47     ` Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
                       ` (7 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

This change splits out the rx/tx specific structs and defines from the
main nfp_net_pmd header file and into their own header file.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/nfp_net.c     |   1 +
 drivers/net/nfp/nfp_net_pmd.h | 248 ------------------------------
 drivers/net/nfp/nfp_rxtx.h    | 276 ++++++++++++++++++++++++++++++++++
 3 files changed, 277 insertions(+), 248 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_rxtx.h

diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index a30e78db16..29eb8c84ad 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -40,6 +40,7 @@
 #include "nfpcore/nfp_nsp.h"
 
 #include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
 #include "nfp_net_logs.h"
 #include "nfp_net_ctrl.h"
 
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index 212f9ef162..a3a3ba32d6 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -23,19 +23,6 @@
 /* Forward declaration */
 struct nfp_net_adapter;
 
-/*
- * The maximum number of descriptors is limited by design as
- * DPDK uses uint16_t variables for these values
- */
-#define NFP_NET_MAX_TX_DESC (32 * 1024)
-#define NFP_NET_MIN_TX_DESC 64
-
-#define NFP_NET_MAX_RX_DESC (32 * 1024)
-#define NFP_NET_MIN_RX_DESC 64
-
-/* Descriptor alignment */
-#define NFP_ALIGN_RING_DESC 128
-
 #define NFP_TX_MAX_SEG     UINT8_MAX
 #define NFP_TX_MAX_MTU_SEG 8
 
@@ -150,241 +137,6 @@ static inline void nn_writeq(uint64_t val, volatile void *addr)
 	nn_writel(val, addr);
 }
 
-/* TX descriptor format */
-#define PCIE_DESC_TX_EOP                (1 << 7)
-#define PCIE_DESC_TX_OFFSET_MASK        (0x7f)
-
-/* Flags in the host TX descriptor */
-#define PCIE_DESC_TX_CSUM               (1 << 7)
-#define PCIE_DESC_TX_IP4_CSUM           (1 << 6)
-#define PCIE_DESC_TX_TCP_CSUM           (1 << 5)
-#define PCIE_DESC_TX_UDP_CSUM           (1 << 4)
-#define PCIE_DESC_TX_VLAN               (1 << 3)
-#define PCIE_DESC_TX_LSO                (1 << 2)
-#define PCIE_DESC_TX_ENCAP_NONE         (0)
-#define PCIE_DESC_TX_ENCAP_VXLAN        (1 << 1)
-#define PCIE_DESC_TX_ENCAP_GRE          (1 << 0)
-
-struct nfp_net_tx_desc {
-	union {
-		struct {
-			uint8_t dma_addr_hi; /* High bits of host buf address */
-			__le16 dma_len;     /* Length to DMA for this desc */
-			uint8_t offset_eop; /* Offset in buf where pkt starts +
-					     * highest bit is eop flag.
-					     */
-			__le32 dma_addr_lo; /* Low 32bit of host buf addr */
-
-			__le16 mss;         /* MSS to be used for LSO */
-			uint8_t lso_hdrlen; /* LSO, where the data starts */
-			uint8_t flags;      /* TX Flags, see @PCIE_DESC_TX_* */
-
-			union {
-				struct {
-					/*
-					 * L3 and L4 header offsets required
-					 * for TSOv2
-					 */
-					uint8_t l3_offset;
-					uint8_t l4_offset;
-				};
-				__le16 vlan; /* VLAN tag to add if indicated */
-			};
-			__le16 data_len;    /* Length of frame + meta data */
-		} __rte_packed;
-		__le32 vals[4];
-	};
-};
-
-struct nfp_net_txq {
-	struct nfp_net_hw *hw; /* Backpointer to nfp_net structure */
-
-	/*
-	 * Queue information: @qidx is the queue index from Linux's
-	 * perspective.  @tx_qcidx is the index of the Queue
-	 * Controller Peripheral queue relative to the TX queue BAR.
-	 * @cnt is the size of the queue in number of
-	 * descriptors. @qcp_q is a pointer to the base of the queue
-	 * structure on the NFP
-	 */
-	uint8_t *qcp_q;
-
-	/*
-	 * Read and Write pointers.  @wr_p and @rd_p are host side pointer,
-	 * they are free running and have little relation to the QCP pointers *
-	 * @qcp_rd_p is a local copy queue controller peripheral read pointer
-	 */
-
-	uint32_t wr_p;
-	uint32_t rd_p;
-
-	uint32_t tx_count;
-
-	uint32_t tx_free_thresh;
-
-	/*
-	 * For each descriptor keep a reference to the mbuf and
-	 * DMA address used until completion is signalled.
-	 */
-	struct {
-		struct rte_mbuf *mbuf;
-	} *txbufs;
-
-	/*
-	 * Information about the host side queue location. @txds is
-	 * the virtual address for the queue, @dma is the DMA address
-	 * of the queue and @size is the size in bytes for the queue
-	 * (needed for free)
-	 */
-	struct nfp_net_tx_desc *txds;
-
-	/*
-	 * At this point 48 bytes have been used for all the fields in the
-	 * TX critical path. We have room for 8 bytes and still all placed
-	 * in a cache line. We are not using the threshold values below but
-	 * if we need to, we can add the most used in the remaining bytes.
-	 */
-	uint32_t tx_rs_thresh; /* not used by now. Future? */
-	uint32_t tx_pthresh;   /* not used by now. Future? */
-	uint32_t tx_hthresh;   /* not used by now. Future? */
-	uint32_t tx_wthresh;   /* not used by now. Future? */
-	uint16_t port_id;
-	int qidx;
-	int tx_qcidx;
-	__le64 dma;
-} __rte_aligned(64);
-
-/* RX and freelist descriptor format */
-#define PCIE_DESC_RX_DD                 (1 << 7)
-#define PCIE_DESC_RX_META_LEN_MASK      (0x7f)
-
-/* Flags in the RX descriptor */
-#define PCIE_DESC_RX_RSS                (1 << 15)
-#define PCIE_DESC_RX_I_IP4_CSUM         (1 << 14)
-#define PCIE_DESC_RX_I_IP4_CSUM_OK      (1 << 13)
-#define PCIE_DESC_RX_I_TCP_CSUM         (1 << 12)
-#define PCIE_DESC_RX_I_TCP_CSUM_OK      (1 << 11)
-#define PCIE_DESC_RX_I_UDP_CSUM         (1 << 10)
-#define PCIE_DESC_RX_I_UDP_CSUM_OK      (1 <<  9)
-#define PCIE_DESC_RX_SPARE              (1 <<  8)
-#define PCIE_DESC_RX_EOP                (1 <<  7)
-#define PCIE_DESC_RX_IP4_CSUM           (1 <<  6)
-#define PCIE_DESC_RX_IP4_CSUM_OK        (1 <<  5)
-#define PCIE_DESC_RX_TCP_CSUM           (1 <<  4)
-#define PCIE_DESC_RX_TCP_CSUM_OK        (1 <<  3)
-#define PCIE_DESC_RX_UDP_CSUM           (1 <<  2)
-#define PCIE_DESC_RX_UDP_CSUM_OK        (1 <<  1)
-#define PCIE_DESC_RX_VLAN               (1 <<  0)
-
-#define PCIE_DESC_RX_L4_CSUM_OK         (PCIE_DESC_RX_TCP_CSUM_OK | \
-					 PCIE_DESC_RX_UDP_CSUM_OK)
-struct nfp_net_rx_desc {
-	union {
-		/* Freelist descriptor */
-		struct {
-			uint8_t dma_addr_hi;
-			__le16 spare;
-			uint8_t dd;
-
-			__le32 dma_addr_lo;
-		} __rte_packed fld;
-
-		/* RX descriptor */
-		struct {
-			__le16 data_len;
-			uint8_t reserved;
-			uint8_t meta_len_dd;
-
-			__le16 flags;
-			__le16 vlan;
-		} __rte_packed rxd;
-
-		__le32 vals[2];
-	};
-};
-
-struct nfp_net_rx_buff {
-	struct rte_mbuf *mbuf;
-};
-
-struct nfp_net_rxq {
-	struct nfp_net_hw *hw;	/* Backpointer to nfp_net structure */
-
-	 /*
-	  * @qcp_fl and @qcp_rx are pointers to the base addresses of the
-	  * freelist and RX queue controller peripheral queue structures on the
-	  * NFP
-	  */
-	uint8_t *qcp_fl;
-	uint8_t *qcp_rx;
-
-	/*
-	 * Read and Write pointers.  @wr_p and @rd_p are host side
-	 * pointer, they are free running and have little relation to
-	 * the QCP pointers. @wr_p is where the driver adds new
-	 * freelist descriptors and @rd_p is where the driver start
-	 * reading descriptors for newly arrive packets from.
-	 */
-	uint32_t rd_p;
-
-	/*
-	 * For each buffer placed on the freelist, record the
-	 * associated SKB
-	 */
-	struct nfp_net_rx_buff *rxbufs;
-
-	/*
-	 * Information about the host side queue location.  @rxds is
-	 * the virtual address for the queue
-	 */
-	struct nfp_net_rx_desc *rxds;
-
-	/*
-	 * The mempool is created by the user specifying a mbuf size.
-	 * We save here the reference of the mempool needed in the RX
-	 * path and the mbuf size for checking received packets can be
-	 * safely copied to the mbuf using the NFP_NET_RX_OFFSET
-	 */
-	struct rte_mempool *mem_pool;
-	uint16_t mbuf_size;
-
-	/*
-	 * Next two fields are used for giving more free descriptors
-	 * to the NFP
-	 */
-	uint16_t rx_free_thresh;
-	uint16_t nb_rx_hold;
-
-	 /* the size of the queue in number of descriptors */
-	uint16_t rx_count;
-
-	/*
-	 * Fields above this point fit in a single cache line and are all used
-	 * in the RX critical path. Fields below this point are just used
-	 * during queue configuration or not used at all (yet)
-	 */
-
-	/* referencing dev->data->port_id */
-	uint16_t port_id;
-
-	uint8_t  crc_len; /* Not used by now */
-	uint8_t  drop_en; /* Not used by now */
-
-	/* DMA address of the queue */
-	__le64 dma;
-
-	/*
-	 * Queue information: @qidx is the queue index from Linux's
-	 * perspective.  @fl_qcidx is the index of the Queue
-	 * Controller peripheral queue relative to the RX queue BAR
-	 * used for the freelist and @rx_qcidx is the Queue Controller
-	 * Peripheral index for the RX queue.
-	 */
-	int qidx;
-	int fl_qcidx;
-	int rx_qcidx;
-} __rte_aligned(64);
-
 struct nfp_pf_dev {
 	/* Backpointer to associated pci device */
 	struct rte_pci_device *pci_dev;
diff --git a/drivers/net/nfp/nfp_rxtx.h b/drivers/net/nfp/nfp_rxtx.h
new file mode 100644
index 0000000000..41a3a4b4e7
--- /dev/null
+++ b/drivers/net/nfp/nfp_rxtx.h
@@ -0,0 +1,276 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_rxtx.h
+ *
+ * Netronome NFP Rx/Tx specific header file
+ */
+
+#ifndef _NFP_RXTX_H_
+#define _NFP_RXTX_H_
+
+#include <linux/types.h>
+#include <rte_io.h>
+
+/*
+ * The maximum number of descriptors is limited by design as
+ * DPDK uses uint16_t variables for these values
+ */
+#define NFP_NET_MAX_TX_DESC (32 * 1024)
+#define NFP_NET_MIN_TX_DESC 64
+
+#define NFP_NET_MAX_RX_DESC (32 * 1024)
+#define NFP_NET_MIN_RX_DESC 64
+
+/* Descriptor alignment */
+#define NFP_ALIGN_RING_DESC 128
+
+/* TX descriptor format */
+#define PCIE_DESC_TX_EOP                (1 << 7)
+#define PCIE_DESC_TX_OFFSET_MASK        (0x7f)
+
+/* Flags in the host TX descriptor */
+#define PCIE_DESC_TX_CSUM               (1 << 7)
+#define PCIE_DESC_TX_IP4_CSUM           (1 << 6)
+#define PCIE_DESC_TX_TCP_CSUM           (1 << 5)
+#define PCIE_DESC_TX_UDP_CSUM           (1 << 4)
+#define PCIE_DESC_TX_VLAN               (1 << 3)
+#define PCIE_DESC_TX_LSO                (1 << 2)
+#define PCIE_DESC_TX_ENCAP_NONE         (0)
+#define PCIE_DESC_TX_ENCAP_VXLAN        (1 << 1)
+#define PCIE_DESC_TX_ENCAP_GRE          (1 << 0)
+
+struct nfp_net_tx_desc {
+	union {
+		struct {
+			uint8_t dma_addr_hi; /* High bits of host buf address */
+			__le16 dma_len;     /* Length to DMA for this desc */
+			uint8_t offset_eop; /* Offset in buf where pkt starts +
+					     * highest bit is eop flag.
+					     */
+			__le32 dma_addr_lo; /* Low 32bit of host buf addr */
+
+			__le16 mss;         /* MSS to be used for LSO */
+			uint8_t lso_hdrlen; /* LSO, where the data starts */
+			uint8_t flags;      /* TX Flags, see @PCIE_DESC_TX_* */
+
+			union {
+				struct {
+					/*
+					 * L3 and L4 header offsets required
+					 * for TSOv2
+					 */
+					uint8_t l3_offset;
+					uint8_t l4_offset;
+				};
+				__le16 vlan; /* VLAN tag to add if indicated */
+			};
+			__le16 data_len;    /* Length of frame + meta data */
+		} __rte_packed;
+		__le32 vals[4];
+	};
+};
+
+struct nfp_net_txq {
+	struct nfp_net_hw *hw; /* Backpointer to nfp_net structure */
+
+	/*
+	 * Queue information: @qidx is the queue index from Linux's
+	 * perspective.  @tx_qcidx is the index of the Queue
+	 * Controller Peripheral queue relative to the TX queue BAR.
+	 * @cnt is the size of the queue in number of
+	 * descriptors. @qcp_q is a pointer to the base of the queue
+	 * structure on the NFP
+	 */
+	uint8_t *qcp_q;
+
+	/*
+	 * Read and Write pointers.  @wr_p and @rd_p are host side pointer,
+	 * they are free running and have little relation to the QCP pointers *
+	 * @qcp_rd_p is a local copy queue controller peripheral read pointer
+	 */
+
+	uint32_t wr_p;
+	uint32_t rd_p;
+
+	uint32_t tx_count;
+
+	uint32_t tx_free_thresh;
+
+	/*
+	 * For each descriptor keep a reference to the mbuf and
+	 * DMA address used until completion is signalled.
+	 */
+	struct {
+		struct rte_mbuf *mbuf;
+	} *txbufs;
+
+	/*
+	 * Information about the host side queue location. @txds is
+	 * the virtual address for the queue, @dma is the DMA address
+	 * of the queue and @size is the size in bytes for the queue
+	 * (needed for free)
+	 */
+	struct nfp_net_tx_desc *txds;
+
+	/*
+	 * At this point 48 bytes have been used for all the fields in the
+	 * TX critical path. We have room for 8 bytes and still all placed
+	 * in a cache line. We are not using the threshold values below but
+	 * if we need to, we can add the most used in the remaining bytes.
+	 */
+	uint32_t tx_rs_thresh; /* not used by now. Future? */
+	uint32_t tx_pthresh;   /* not used by now. Future? */
+	uint32_t tx_hthresh;   /* not used by now. Future? */
+	uint32_t tx_wthresh;   /* not used by now. Future? */
+	uint16_t port_id;
+	int qidx;
+	int tx_qcidx;
+	__le64 dma;
+} __rte_aligned(64);
+
+/* RX and freelist descriptor format */
+#define PCIE_DESC_RX_DD                 (1 << 7)
+#define PCIE_DESC_RX_META_LEN_MASK      (0x7f)
+
+/* Flags in the RX descriptor */
+#define PCIE_DESC_RX_RSS                (1 << 15)
+#define PCIE_DESC_RX_I_IP4_CSUM         (1 << 14)
+#define PCIE_DESC_RX_I_IP4_CSUM_OK      (1 << 13)
+#define PCIE_DESC_RX_I_TCP_CSUM         (1 << 12)
+#define PCIE_DESC_RX_I_TCP_CSUM_OK      (1 << 11)
+#define PCIE_DESC_RX_I_UDP_CSUM         (1 << 10)
+#define PCIE_DESC_RX_I_UDP_CSUM_OK      (1 <<  9)
+#define PCIE_DESC_RX_SPARE              (1 <<  8)
+#define PCIE_DESC_RX_EOP                (1 <<  7)
+#define PCIE_DESC_RX_IP4_CSUM           (1 <<  6)
+#define PCIE_DESC_RX_IP4_CSUM_OK        (1 <<  5)
+#define PCIE_DESC_RX_TCP_CSUM           (1 <<  4)
+#define PCIE_DESC_RX_TCP_CSUM_OK        (1 <<  3)
+#define PCIE_DESC_RX_UDP_CSUM           (1 <<  2)
+#define PCIE_DESC_RX_UDP_CSUM_OK        (1 <<  1)
+#define PCIE_DESC_RX_VLAN               (1 <<  0)
+
+#define PCIE_DESC_RX_L4_CSUM_OK         (PCIE_DESC_RX_TCP_CSUM_OK | \
+					 PCIE_DESC_RX_UDP_CSUM_OK)
+
+struct nfp_net_rx_desc {
+	union {
+		/* Freelist descriptor */
+		struct {
+			uint8_t dma_addr_hi;
+			__le16 spare;
+			uint8_t dd;
+
+			__le32 dma_addr_lo;
+		} __rte_packed fld;
+
+		/* RX descriptor */
+		struct {
+			__le16 data_len;
+			uint8_t reserved;
+			uint8_t meta_len_dd;
+
+			__le16 flags;
+			__le16 vlan;
+		} __rte_packed rxd;
+
+		__le32 vals[2];
+	};
+};
+
+struct nfp_net_rx_buff {
+	struct rte_mbuf *mbuf;
+};
+
+struct nfp_net_rxq {
+	struct nfp_net_hw *hw;	/* Backpointer to nfp_net structure */
+
+	 /*
+	  * @qcp_fl and @qcp_rx are pointers to the base addresses of the
+	  * freelist and RX queue controller peripheral queue structures on the
+	  * NFP
+	  */
+	uint8_t *qcp_fl;
+	uint8_t *qcp_rx;
+
+	/*
+	 * Read and Write pointers.  @wr_p and @rd_p are host side
+	 * pointer, they are free running and have little relation to
+	 * the QCP pointers. @wr_p is where the driver adds new
+	 * freelist descriptors and @rd_p is where the driver start
+	 * reading descriptors for newly arrive packets from.
+	 */
+	uint32_t rd_p;
+
+	/*
+	 * For each buffer placed on the freelist, record the
+	 * associated SKB
+	 */
+	struct nfp_net_rx_buff *rxbufs;
+
+	/*
+	 * Information about the host side queue location.  @rxds is
+	 * the virtual address for the queue
+	 */
+	struct nfp_net_rx_desc *rxds;
+
+	/*
+	 * The mempool is created by the user specifying a mbuf size.
+	 * We save here the reference of the mempool needed in the RX
+	 * path and the mbuf size for checking received packets can be
+	 * safely copied to the mbuf using the NFP_NET_RX_OFFSET
+	 */
+	struct rte_mempool *mem_pool;
+	uint16_t mbuf_size;
+
+	/*
+	 * Next two fields are used for giving more free descriptors
+	 * to the NFP
+	 */
+	uint16_t rx_free_thresh;
+	uint16_t nb_rx_hold;
+
+	 /* the size of the queue in number of descriptors */
+	uint16_t rx_count;
+
+	/*
+	 * Fields above this point fit in a single cache line and are all used
+	 * in the RX critical path. Fields below this point are just used
+	 * during queue configuration or not used at all (yet)
+	 */
+
+	/* referencing dev->data->port_id */
+	uint16_t port_id;
+
+	uint8_t  crc_len; /* Not used by now */
+	uint8_t  drop_en; /* Not used by now */
+
+	/* DMA address of the queue */
+	__le64 dma;
+
+	/*
+	 * Queue information: @qidx is the queue index from Linux's
+	 * perspective.  @fl_qcidx is the index of the Queue
+	 * Controller peripheral queue relative to the RX queue BAR
+	 * used for the freelist and @rx_qcidx is the Queue Controller
+	 * Peripheral index for the RX queue.
+	 */
+	int qidx;
+	int fl_qcidx;
+	int rx_qcidx;
+} __rte_aligned(64);
+
+#endif /* _NFP_RXTX_H_ */
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
+
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 2/7] net/nfp: move rxtx functions to their own file
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
@ 2021-07-29 13:47     ` Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
                       ` (6 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Create a new rxtx file and move the Rx/Tx functions to this file. This
commit will also move the needed shared functions to the nfp_net_pmd.h
file as needed.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build   |    1 +
 drivers/net/nfp/nfp_net.c     | 1090 ---------------------------------
 drivers/net/nfp/nfp_net_pmd.h |  184 ++++--
 drivers/net/nfp/nfp_rxtx.c    | 1002 ++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_rxtx.h    |   27 +
 5 files changed, 1173 insertions(+), 1131 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_rxtx.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index b51e2e5f20..1b289e2354 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -19,4 +19,5 @@ sources = files(
         'nfpcore/nfp_nsp_eth.c',
         'nfpcore/nfp_hwinfo.c',
         'nfp_net.c',
+        'nfp_rxtx.c',
 )
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 29eb8c84ad..de0de808f4 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -68,29 +68,11 @@ static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
 static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
 static int nfp_net_promisc_enable(struct rte_eth_dev *dev);
 static int nfp_net_promisc_disable(struct rte_eth_dev *dev);
-static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
-static uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
-				       uint16_t queue_idx);
-static uint16_t nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
-				  uint16_t nb_pkts);
-static void nfp_net_rx_queue_release(void *rxq);
-static int nfp_net_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-				  uint16_t nb_desc, unsigned int socket_id,
-				  const struct rte_eth_rxconf *rx_conf,
-				  struct rte_mempool *mp);
-static int nfp_net_tx_free_bufs(struct nfp_net_txq *txq);
-static void nfp_net_tx_queue_release(void *txq);
-static int nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-				  uint16_t nb_desc, unsigned int socket_id,
-				  const struct rte_eth_txconf *tx_conf);
 static int nfp_net_start(struct rte_eth_dev *dev);
 static int nfp_net_stats_get(struct rte_eth_dev *dev,
 			      struct rte_eth_stats *stats);
 static int nfp_net_stats_reset(struct rte_eth_dev *dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
-static uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
-				  uint16_t nb_pkts);
-
 static int nfp_net_rss_config_default(struct rte_eth_dev *dev);
 static int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 				   struct rte_eth_rss_conf *rss_conf);
@@ -108,184 +90,6 @@ static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_eth_table *nfp_eth_table,
 			struct nfp_hwinfo *hwinfo);
 
-
-/* The offset of the queue controller queues in the PCIe Target */
-#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
-
-/* Maximum value which can be added to a queue with one transaction */
-#define NFP_QCP_MAX_ADD	0x7f
-
-#define RTE_MBUF_DMA_ADDR_DEFAULT(mb) \
-	(uint64_t)((mb)->buf_iova + RTE_PKTMBUF_HEADROOM)
-
-/* nfp_qcp_ptr - Read or Write Pointer of a queue */
-enum nfp_qcp_ptr {
-	NFP_QCP_READ_PTR = 0,
-	NFP_QCP_WRITE_PTR
-};
-
-/*
- * nfp_qcp_ptr_add - Add the value to the selected pointer of a queue
- * @q: Base address for queue structure
- * @ptr: Add to the Read or Write pointer
- * @val: Value to add to the queue pointer
- *
- * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed.
- */
-static inline void
-nfp_qcp_ptr_add(uint8_t *q, enum nfp_qcp_ptr ptr, uint32_t val)
-{
-	uint32_t off;
-
-	if (ptr == NFP_QCP_READ_PTR)
-		off = NFP_QCP_QUEUE_ADD_RPTR;
-	else
-		off = NFP_QCP_QUEUE_ADD_WPTR;
-
-	while (val > NFP_QCP_MAX_ADD) {
-		nn_writel(rte_cpu_to_le_32(NFP_QCP_MAX_ADD), q + off);
-		val -= NFP_QCP_MAX_ADD;
-	}
-
-	nn_writel(rte_cpu_to_le_32(val), q + off);
-}
-
-/*
- * nfp_qcp_read - Read the current Read/Write pointer value for a queue
- * @q:  Base address for queue structure
- * @ptr: Read or Write pointer
- */
-static inline uint32_t
-nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
-{
-	uint32_t off;
-	uint32_t val;
-
-	if (ptr == NFP_QCP_READ_PTR)
-		off = NFP_QCP_QUEUE_STS_LO;
-	else
-		off = NFP_QCP_QUEUE_STS_HI;
-
-	val = rte_cpu_to_le_32(nn_readl(q + off));
-
-	if (ptr == NFP_QCP_READ_PTR)
-		return val & NFP_QCP_QUEUE_STS_LO_READPTR_mask;
-	else
-		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
-}
-
-/*
- * Functions to read/write from/to Config BAR
- * Performs any endian conversion necessary.
- */
-static inline uint8_t
-nn_cfg_readb(struct nfp_net_hw *hw, int off)
-{
-	return nn_readb(hw->ctrl_bar + off);
-}
-
-static inline void
-nn_cfg_writeb(struct nfp_net_hw *hw, int off, uint8_t val)
-{
-	nn_writeb(val, hw->ctrl_bar + off);
-}
-
-static inline uint32_t
-nn_cfg_readl(struct nfp_net_hw *hw, int off)
-{
-	return rte_le_to_cpu_32(nn_readl(hw->ctrl_bar + off));
-}
-
-static inline void
-nn_cfg_writel(struct nfp_net_hw *hw, int off, uint32_t val)
-{
-	nn_writel(rte_cpu_to_le_32(val), hw->ctrl_bar + off);
-}
-
-static inline uint64_t
-nn_cfg_readq(struct nfp_net_hw *hw, int off)
-{
-	return rte_le_to_cpu_64(nn_readq(hw->ctrl_bar + off));
-}
-
-static inline void
-nn_cfg_writeq(struct nfp_net_hw *hw, int off, uint64_t val)
-{
-	nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
-}
-
-static void
-nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
-{
-	unsigned i;
-
-	if (rxq->rxbufs == NULL)
-		return;
-
-	for (i = 0; i < rxq->rx_count; i++) {
-		if (rxq->rxbufs[i].mbuf) {
-			rte_pktmbuf_free_seg(rxq->rxbufs[i].mbuf);
-			rxq->rxbufs[i].mbuf = NULL;
-		}
-	}
-}
-
-static void
-nfp_net_rx_queue_release(void *rx_queue)
-{
-	struct nfp_net_rxq *rxq = rx_queue;
-
-	if (rxq) {
-		nfp_net_rx_queue_release_mbufs(rxq);
-		rte_free(rxq->rxbufs);
-		rte_free(rxq);
-	}
-}
-
-static void
-nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
-{
-	nfp_net_rx_queue_release_mbufs(rxq);
-	rxq->rd_p = 0;
-	rxq->nb_rx_hold = 0;
-}
-
-static void
-nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq)
-{
-	unsigned i;
-
-	if (txq->txbufs == NULL)
-		return;
-
-	for (i = 0; i < txq->tx_count; i++) {
-		if (txq->txbufs[i].mbuf) {
-			rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
-			txq->txbufs[i].mbuf = NULL;
-		}
-	}
-}
-
-static void
-nfp_net_tx_queue_release(void *tx_queue)
-{
-	struct nfp_net_txq *txq = tx_queue;
-
-	if (txq) {
-		nfp_net_tx_queue_release_mbufs(txq);
-		rte_free(txq->txbufs);
-		rte_free(txq);
-	}
-}
-
-static void
-nfp_net_reset_tx_queue(struct nfp_net_txq *txq)
-{
-	nfp_net_tx_queue_release_mbufs(txq);
-	txq->wr_p = 0;
-	txq->rd_p = 0;
-}
-
 static int
 __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
 {
@@ -463,18 +267,6 @@ nfp_net_disable_queues(struct rte_eth_dev *dev)
 	hw->ctrl = new_ctrl;
 }
 
-static int
-nfp_net_rx_freelist_setup(struct rte_eth_dev *dev)
-{
-	int i;
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		if (nfp_net_rx_fill_freelist(dev->data->rx_queues[i]) < 0)
-			return -1;
-	}
-	return 0;
-}
-
 static void
 nfp_net_params_setup(struct nfp_net_hw *hw)
 {
@@ -1351,44 +1143,6 @@ nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 	return NULL;
 }
 
-static uint32_t
-nfp_net_rx_queue_count(struct rte_eth_dev *dev, uint16_t queue_idx)
-{
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_rx_desc *rxds;
-	uint32_t idx;
-	uint32_t count;
-
-	rxq = (struct nfp_net_rxq *)dev->data->rx_queues[queue_idx];
-
-	idx = rxq->rd_p;
-
-	count = 0;
-
-	/*
-	 * Other PMDs are just checking the DD bit in intervals of 4
-	 * descriptors and counting all four if the first has the DD
-	 * bit on. Of course, this is not accurate but can be good for
-	 * performance. But ideally that should be done in descriptors
-	 * chunks belonging to the same cache line
-	 */
-
-	while (count < rxq->rx_count) {
-		rxds = &rxq->rxds[idx];
-		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
-			break;
-
-		count++;
-		idx++;
-
-		/* Wrapping? */
-		if ((idx) == rxq->rx_count)
-			idx = 0;
-	}
-
-	return count;
-}
-
 static int
 nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
@@ -1570,850 +1324,6 @@ nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	return 0;
 }
 
-static int
-nfp_net_rx_queue_setup(struct rte_eth_dev *dev,
-		       uint16_t queue_idx, uint16_t nb_desc,
-		       unsigned int socket_id,
-		       const struct rte_eth_rxconf *rx_conf,
-		       struct rte_mempool *mp)
-{
-	const struct rte_memzone *tz;
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_hw *hw;
-	uint32_t rx_desc_sz;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	PMD_INIT_FUNC_TRACE();
-
-	/* Validating number of descriptors */
-	rx_desc_sz = nb_desc * sizeof(struct nfp_net_rx_desc);
-	if (rx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
-	    nb_desc > NFP_NET_MAX_RX_DESC ||
-	    nb_desc < NFP_NET_MIN_RX_DESC) {
-		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
-		return -EINVAL;
-	}
-
-	/*
-	 * Free memory prior to re-allocation if needed. This is the case after
-	 * calling nfp_net_stop
-	 */
-	if (dev->data->rx_queues[queue_idx]) {
-		nfp_net_rx_queue_release(dev->data->rx_queues[queue_idx]);
-		dev->data->rx_queues[queue_idx] = NULL;
-	}
-
-	/* Allocating rx queue data structure */
-	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct nfp_net_rxq),
-				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (rxq == NULL)
-		return -ENOMEM;
-
-	/* Hw queues mapping based on firmware configuration */
-	rxq->qidx = queue_idx;
-	rxq->fl_qcidx = queue_idx * hw->stride_rx;
-	rxq->rx_qcidx = rxq->fl_qcidx + (hw->stride_rx - 1);
-	rxq->qcp_fl = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->fl_qcidx);
-	rxq->qcp_rx = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->rx_qcidx);
-
-	/*
-	 * Tracking mbuf size for detecting a potential mbuf overflow due to
-	 * RX offset
-	 */
-	rxq->mem_pool = mp;
-	rxq->mbuf_size = rxq->mem_pool->elt_size;
-	rxq->mbuf_size -= (sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM);
-	hw->flbufsz = rxq->mbuf_size;
-
-	rxq->rx_count = nb_desc;
-	rxq->port_id = dev->data->port_id;
-	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
-	rxq->drop_en = rx_conf->rx_drop_en;
-
-	/*
-	 * Allocate RX ring hardware descriptors. A memzone large enough to
-	 * handle the maximum ring size is allocated in order to allow for
-	 * resizing in later calls to the queue setup function.
-	 */
-	tz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
-				   sizeof(struct nfp_net_rx_desc) *
-				   NFP_NET_MAX_RX_DESC, NFP_MEMZONE_ALIGN,
-				   socket_id);
-
-	if (tz == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating rx dma");
-		nfp_net_rx_queue_release(rxq);
-		return -ENOMEM;
-	}
-
-	/* Saving physical and virtual addresses for the RX ring */
-	rxq->dma = (uint64_t)tz->iova;
-	rxq->rxds = (struct nfp_net_rx_desc *)tz->addr;
-
-	/* mbuf pointers array for referencing mbufs linked to RX descriptors */
-	rxq->rxbufs = rte_zmalloc_socket("rxq->rxbufs",
-					 sizeof(*rxq->rxbufs) * nb_desc,
-					 RTE_CACHE_LINE_SIZE, socket_id);
-	if (rxq->rxbufs == NULL) {
-		nfp_net_rx_queue_release(rxq);
-		return -ENOMEM;
-	}
-
-	PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
-		   rxq->rxbufs, rxq->rxds, (unsigned long int)rxq->dma);
-
-	nfp_net_reset_rx_queue(rxq);
-
-	dev->data->rx_queues[queue_idx] = rxq;
-	rxq->hw = hw;
-
-	/*
-	 * Telling the HW about the physical address of the RX ring and number
-	 * of descriptors in log2 format
-	 */
-	nn_cfg_writeq(hw, NFP_NET_CFG_RXR_ADDR(queue_idx), rxq->dma);
-	nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), rte_log2_u32(nb_desc));
-
-	return 0;
-}
-
-static int
-nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq)
-{
-	struct nfp_net_rx_buff *rxe = rxq->rxbufs;
-	uint64_t dma_addr;
-	unsigned i;
-
-	PMD_RX_LOG(DEBUG, "nfp_net_rx_fill_freelist for %u descriptors",
-		   rxq->rx_count);
-
-	for (i = 0; i < rxq->rx_count; i++) {
-		struct nfp_net_rx_desc *rxd;
-		struct rte_mbuf *mbuf = rte_pktmbuf_alloc(rxq->mem_pool);
-
-		if (mbuf == NULL) {
-			PMD_DRV_LOG(ERR, "RX mbuf alloc failed queue_id=%u",
-				(unsigned)rxq->qidx);
-			return -ENOMEM;
-		}
-
-		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(mbuf));
-
-		rxd = &rxq->rxds[i];
-		rxd->fld.dd = 0;
-		rxd->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
-		rxd->fld.dma_addr_lo = dma_addr & 0xffffffff;
-		rxe[i].mbuf = mbuf;
-		PMD_RX_LOG(DEBUG, "[%d]: %" PRIx64, i, dma_addr);
-	}
-
-	/* Make sure all writes are flushed before telling the hardware */
-	rte_wmb();
-
-	/* Not advertising the whole ring as the firmware gets confused if so */
-	PMD_RX_LOG(DEBUG, "Increment FL write pointer in %u",
-		   rxq->rx_count - 1);
-
-	nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, rxq->rx_count - 1);
-
-	return 0;
-}
-
-static int
-nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-		       uint16_t nb_desc, unsigned int socket_id,
-		       const struct rte_eth_txconf *tx_conf)
-{
-	const struct rte_memzone *tz;
-	struct nfp_net_txq *txq;
-	uint16_t tx_free_thresh;
-	struct nfp_net_hw *hw;
-	uint32_t tx_desc_sz;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	PMD_INIT_FUNC_TRACE();
-
-	/* Validating number of descriptors */
-	tx_desc_sz = nb_desc * sizeof(struct nfp_net_tx_desc);
-	if (tx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
-	    nb_desc > NFP_NET_MAX_TX_DESC ||
-	    nb_desc < NFP_NET_MIN_TX_DESC) {
-		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
-		return -EINVAL;
-	}
-
-	tx_free_thresh = (uint16_t)((tx_conf->tx_free_thresh) ?
-				    tx_conf->tx_free_thresh :
-				    DEFAULT_TX_FREE_THRESH);
-
-	if (tx_free_thresh > (nb_desc)) {
-		PMD_DRV_LOG(ERR,
-			"tx_free_thresh must be less than the number of TX "
-			"descriptors. (tx_free_thresh=%u port=%d "
-			"queue=%d)", (unsigned int)tx_free_thresh,
-			dev->data->port_id, (int)queue_idx);
-		return -(EINVAL);
-	}
-
-	/*
-	 * Free memory prior to re-allocation if needed. This is the case after
-	 * calling nfp_net_stop
-	 */
-	if (dev->data->tx_queues[queue_idx]) {
-		PMD_TX_LOG(DEBUG, "Freeing memory prior to re-allocation %d",
-			   queue_idx);
-		nfp_net_tx_queue_release(dev->data->tx_queues[queue_idx]);
-		dev->data->tx_queues[queue_idx] = NULL;
-	}
-
-	/* Allocating tx queue data structure */
-	txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct nfp_net_txq),
-				 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating tx dma");
-		return -ENOMEM;
-	}
-
-	/*
-	 * Allocate TX ring hardware descriptors. A memzone large enough to
-	 * handle the maximum ring size is allocated in order to allow for
-	 * resizing in later calls to the queue setup function.
-	 */
-	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
-				   sizeof(struct nfp_net_tx_desc) *
-				   NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN,
-				   socket_id);
-	if (tz == NULL) {
-		PMD_DRV_LOG(ERR, "Error allocating tx dma");
-		nfp_net_tx_queue_release(txq);
-		return -ENOMEM;
-	}
-
-	txq->tx_count = nb_desc;
-	txq->tx_free_thresh = tx_free_thresh;
-	txq->tx_pthresh = tx_conf->tx_thresh.pthresh;
-	txq->tx_hthresh = tx_conf->tx_thresh.hthresh;
-	txq->tx_wthresh = tx_conf->tx_thresh.wthresh;
-
-	/* queue mapping based on firmware configuration */
-	txq->qidx = queue_idx;
-	txq->tx_qcidx = queue_idx * hw->stride_tx;
-	txq->qcp_q = hw->tx_bar + NFP_QCP_QUEUE_OFF(txq->tx_qcidx);
-
-	txq->port_id = dev->data->port_id;
-
-	/* Saving physical and virtual addresses for the TX ring */
-	txq->dma = (uint64_t)tz->iova;
-	txq->txds = (struct nfp_net_tx_desc *)tz->addr;
-
-	/* mbuf pointers array for referencing mbufs linked to TX descriptors */
-	txq->txbufs = rte_zmalloc_socket("txq->txbufs",
-					 sizeof(*txq->txbufs) * nb_desc,
-					 RTE_CACHE_LINE_SIZE, socket_id);
-	if (txq->txbufs == NULL) {
-		nfp_net_tx_queue_release(txq);
-		return -ENOMEM;
-	}
-	PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
-		   txq->txbufs, txq->txds, (unsigned long int)txq->dma);
-
-	nfp_net_reset_tx_queue(txq);
-
-	dev->data->tx_queues[queue_idx] = txq;
-	txq->hw = hw;
-
-	/*
-	 * Telling the HW about the physical address of the TX ring and number
-	 * of descriptors in log2 format
-	 */
-	nn_cfg_writeq(hw, NFP_NET_CFG_TXR_ADDR(queue_idx), txq->dma);
-	nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), rte_log2_u32(nb_desc));
-
-	return 0;
-}
-
-/* nfp_net_tx_tso - Set TX descriptor for TSO */
-static inline void
-nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
-	       struct rte_mbuf *mb)
-{
-	uint64_t ol_flags;
-	struct nfp_net_hw *hw = txq->hw;
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY))
-		goto clean_txd;
-
-	ol_flags = mb->ol_flags;
-
-	if (!(ol_flags & PKT_TX_TCP_SEG))
-		goto clean_txd;
-
-	txd->l3_offset = mb->l2_len;
-	txd->l4_offset = mb->l2_len + mb->l3_len;
-	txd->lso_hdrlen = mb->l2_len + mb->l3_len + mb->l4_len;
-	txd->mss = rte_cpu_to_le_16(mb->tso_segsz);
-	txd->flags = PCIE_DESC_TX_LSO;
-	return;
-
-clean_txd:
-	txd->flags = 0;
-	txd->l3_offset = 0;
-	txd->l4_offset = 0;
-	txd->lso_hdrlen = 0;
-	txd->mss = 0;
-}
-
-/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
-static inline void
-nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
-		 struct rte_mbuf *mb)
-{
-	uint64_t ol_flags;
-	struct nfp_net_hw *hw = txq->hw;
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_TXCSUM))
-		return;
-
-	ol_flags = mb->ol_flags;
-
-	/* IPv6 does not need checksum */
-	if (ol_flags & PKT_TX_IP_CKSUM)
-		txd->flags |= PCIE_DESC_TX_IP4_CSUM;
-
-	switch (ol_flags & PKT_TX_L4_MASK) {
-	case PKT_TX_UDP_CKSUM:
-		txd->flags |= PCIE_DESC_TX_UDP_CSUM;
-		break;
-	case PKT_TX_TCP_CKSUM:
-		txd->flags |= PCIE_DESC_TX_TCP_CSUM;
-		break;
-	}
-
-	if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
-		txd->flags |= PCIE_DESC_TX_CSUM;
-}
-
-/* nfp_net_rx_cksum - set mbuf checksum flags based on RX descriptor flags */
-static inline void
-nfp_net_rx_cksum(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
-		 struct rte_mbuf *mb)
-{
-	struct nfp_net_hw *hw = rxq->hw;
-
-	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RXCSUM))
-		return;
-
-	/* If IPv4 and IP checksum error, fail */
-	if (unlikely((rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM) &&
-	    !(rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM_OK)))
-		mb->ol_flags |= PKT_RX_IP_CKSUM_BAD;
-	else
-		mb->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
-
-	/* If neither UDP nor TCP return */
-	if (!(rxd->rxd.flags & PCIE_DESC_RX_TCP_CSUM) &&
-	    !(rxd->rxd.flags & PCIE_DESC_RX_UDP_CSUM))
-		return;
-
-	if (likely(rxd->rxd.flags & PCIE_DESC_RX_L4_CSUM_OK))
-		mb->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
-	else
-		mb->ol_flags |= PKT_RX_L4_CKSUM_BAD;
-}
-
-#define NFP_HASH_OFFSET      ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 4)
-#define NFP_HASH_TYPE_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 8)
-
-#define NFP_DESC_META_LEN(d) (d->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
-
-/*
- * nfp_net_set_hash - Set mbuf hash data
- *
- * The RSS hash and hash-type are pre-pended to the packet data.
- * Extract and decode it and set the mbuf fields.
- */
-static inline void
-nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
-		 struct rte_mbuf *mbuf)
-{
-	struct nfp_net_hw *hw = rxq->hw;
-	uint8_t *meta_offset;
-	uint32_t meta_info;
-	uint32_t hash = 0;
-	uint32_t hash_type = 0;
-
-	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
-		return;
-
-	/* this is true for new firmwares */
-	if (likely(((hw->cap & NFP_NET_CFG_CTRL_RSS2) ||
-	    (NFD_CFG_MAJOR_VERSION_of(hw->ver) == 4)) &&
-	     NFP_DESC_META_LEN(rxd))) {
-		/*
-		 * new metadata api:
-		 * <----  32 bit  ----->
-		 * m    field type word
-		 * e     data field #2
-		 * t     data field #1
-		 * a     data field #0
-		 * ====================
-		 *    packet data
-		 *
-		 * Field type word contains up to 8 4bit field types
-		 * A 4bit field type refers to a data field word
-		 * A data field word can have several 4bit field types
-		 */
-		meta_offset = rte_pktmbuf_mtod(mbuf, uint8_t *);
-		meta_offset -= NFP_DESC_META_LEN(rxd);
-		meta_info = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
-		meta_offset += 4;
-		/* NFP PMD just supports metadata for hashing */
-		switch (meta_info & NFP_NET_META_FIELD_MASK) {
-		case NFP_NET_META_HASH:
-			/* next field type is about the hash type */
-			meta_info >>= NFP_NET_META_FIELD_SIZE;
-			/* hash value is in the data field */
-			hash = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
-			hash_type = meta_info & NFP_NET_META_FIELD_MASK;
-			break;
-		default:
-			/* Unsupported metadata can be a performance issue */
-			return;
-		}
-	} else {
-		if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
-			return;
-
-		hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
-		hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
-	}
-
-	mbuf->hash.rss = hash;
-	mbuf->ol_flags |= PKT_RX_RSS_HASH;
-
-	switch (hash_type) {
-	case NFP_NET_RSS_IPV4:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV4;
-		break;
-	case NFP_NET_RSS_IPV6:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6;
-		break;
-	case NFP_NET_RSS_IPV6_EX:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV4_TCP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV6_TCP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV4_UDP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	case NFP_NET_RSS_IPV6_UDP:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
-		break;
-	default:
-		mbuf->packet_type |= RTE_PTYPE_INNER_L4_MASK;
-	}
-}
-
-static inline void
-nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq)
-{
-	rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
-}
-
-#define NFP_DESC_META_LEN(d) (d->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
-
-/*
- * RX path design:
- *
- * There are some decisions to take:
- * 1) How to check DD RX descriptors bit
- * 2) How and when to allocate new mbufs
- *
- * Current implementation checks just one single DD bit each loop. As each
- * descriptor is 8 bytes, it is likely a good idea to check descriptors in
- * a single cache line instead. Tests with this change have not shown any
- * performance improvement but it requires further investigation. For example,
- * depending on which descriptor is next, the number of descriptors could be
- * less than 8 for just checking those in the same cache line. This implies
- * extra work which could be counterproductive by itself. Indeed, last firmware
- * changes are just doing this: writing several descriptors with the DD bit
- * for saving PCIe bandwidth and DMA operations from the NFP.
- *
- * Mbuf allocation is done when a new packet is received. Then the descriptor
- * is automatically linked with the new mbuf and the old one is given to the
- * user. The main drawback with this design is mbuf allocation is heavier than
- * using bulk allocations allowed by DPDK with rte_mempool_get_bulk. From the
- * cache point of view it does not seem allocating the mbuf early on as we are
- * doing now have any benefit at all. Again, tests with this change have not
- * shown any improvement. Also, rte_mempool_get_bulk returns all or nothing
- * so looking at the implications of this type of allocation should be studied
- * deeply
- */
-
-static uint16_t
-nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
-{
-	struct nfp_net_rxq *rxq;
-	struct nfp_net_rx_desc *rxds;
-	struct nfp_net_rx_buff *rxb;
-	struct nfp_net_hw *hw;
-	struct rte_mbuf *mb;
-	struct rte_mbuf *new_mb;
-	uint16_t nb_hold;
-	uint64_t dma_addr;
-	int avail;
-
-	rxq = rx_queue;
-	if (unlikely(rxq == NULL)) {
-		/*
-		 * DPDK just checks the queue is lower than max queues
-		 * enabled. But the queue needs to be configured
-		 */
-		RTE_LOG_DP(ERR, PMD, "RX Bad queue\n");
-		return -EINVAL;
-	}
-
-	hw = rxq->hw;
-	avail = 0;
-	nb_hold = 0;
-
-	while (avail < nb_pkts) {
-		rxb = &rxq->rxbufs[rxq->rd_p];
-		if (unlikely(rxb == NULL)) {
-			RTE_LOG_DP(ERR, PMD, "rxb does not exist!\n");
-			break;
-		}
-
-		rxds = &rxq->rxds[rxq->rd_p];
-		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
-			break;
-
-		/*
-		 * Memory barrier to ensure that we won't do other
-		 * reads before the DD bit.
-		 */
-		rte_rmb();
-
-		/*
-		 * We got a packet. Let's alloc a new mbuf for refilling the
-		 * free descriptor ring as soon as possible
-		 */
-		new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
-		if (unlikely(new_mb == NULL)) {
-			RTE_LOG_DP(DEBUG, PMD,
-			"RX mbuf alloc failed port_id=%u queue_id=%u\n",
-				rxq->port_id, (unsigned int)rxq->qidx);
-			nfp_net_mbuf_alloc_failed(rxq);
-			break;
-		}
-
-		nb_hold++;
-
-		/*
-		 * Grab the mbuf and refill the descriptor with the
-		 * previously allocated mbuf
-		 */
-		mb = rxb->mbuf;
-		rxb->mbuf = new_mb;
-
-		PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u",
-			   rxds->rxd.data_len, rxq->mbuf_size);
-
-		/* Size of this segment */
-		mb->data_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
-		/* Size of the whole packet. We just support 1 segment */
-		mb->pkt_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
-
-		if (unlikely((mb->data_len + hw->rx_offset) >
-			     rxq->mbuf_size)) {
-			/*
-			 * This should not happen and the user has the
-			 * responsibility of avoiding it. But we have
-			 * to give some info about the error
-			 */
-			RTE_LOG_DP(ERR, PMD,
-				"mbuf overflow likely due to the RX offset.\n"
-				"\t\tYour mbuf size should have extra space for"
-				" RX offset=%u bytes.\n"
-				"\t\tCurrently you just have %u bytes available"
-				" but the received packet is %u bytes long",
-				hw->rx_offset,
-				rxq->mbuf_size - hw->rx_offset,
-				mb->data_len);
-			return -EINVAL;
-		}
-
-		/* Filling the received mbuf with packet info */
-		if (hw->rx_offset)
-			mb->data_off = RTE_PKTMBUF_HEADROOM + hw->rx_offset;
-		else
-			mb->data_off = RTE_PKTMBUF_HEADROOM +
-				       NFP_DESC_META_LEN(rxds);
-
-		/* No scatter mode supported */
-		mb->nb_segs = 1;
-		mb->next = NULL;
-
-		mb->port = rxq->port_id;
-
-		/* Checking the RSS flag */
-		nfp_net_set_hash(rxq, rxds, mb);
-
-		/* Checking the checksum flag */
-		nfp_net_rx_cksum(rxq, rxds, mb);
-
-		if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
-		    (hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
-			mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
-			mb->ol_flags |= PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
-		}
-
-		/* Adding the mbuf to the mbuf array passed by the app */
-		rx_pkts[avail++] = mb;
-
-		/* Now resetting and updating the descriptor */
-		rxds->vals[0] = 0;
-		rxds->vals[1] = 0;
-		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(new_mb));
-		rxds->fld.dd = 0;
-		rxds->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
-		rxds->fld.dma_addr_lo = dma_addr & 0xffffffff;
-
-		rxq->rd_p++;
-		if (unlikely(rxq->rd_p == rxq->rx_count)) /* wrapping?*/
-			rxq->rd_p = 0;
-	}
-
-	if (nb_hold == 0)
-		return nb_hold;
-
-	PMD_RX_LOG(DEBUG, "RX  port_id=%u queue_id=%u, %d packets received",
-		   rxq->port_id, (unsigned int)rxq->qidx, nb_hold);
-
-	nb_hold += rxq->nb_rx_hold;
-
-	/*
-	 * FL descriptors needs to be written before incrementing the
-	 * FL queue WR pointer
-	 */
-	rte_wmb();
-	if (nb_hold > rxq->rx_free_thresh) {
-		PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u",
-			   rxq->port_id, (unsigned int)rxq->qidx,
-			   (unsigned)nb_hold, (unsigned)avail);
-		nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, nb_hold);
-		nb_hold = 0;
-	}
-	rxq->nb_rx_hold = nb_hold;
-
-	return avail;
-}
-
-/*
- * nfp_net_tx_free_bufs - Check for descriptors with a complete
- * status
- * @txq: TX queue to work with
- * Returns number of descriptors freed
- */
-int
-nfp_net_tx_free_bufs(struct nfp_net_txq *txq)
-{
-	uint32_t qcp_rd_p;
-	int todo;
-
-	PMD_TX_LOG(DEBUG, "queue %u. Check for descriptor with a complete"
-		   " status", txq->qidx);
-
-	/* Work out how many packets have been sent */
-	qcp_rd_p = nfp_qcp_read(txq->qcp_q, NFP_QCP_READ_PTR);
-
-	if (qcp_rd_p == txq->rd_p) {
-		PMD_TX_LOG(DEBUG, "queue %u: It seems harrier is not sending "
-			   "packets (%u, %u)", txq->qidx,
-			   qcp_rd_p, txq->rd_p);
-		return 0;
-	}
-
-	if (qcp_rd_p > txq->rd_p)
-		todo = qcp_rd_p - txq->rd_p;
-	else
-		todo = qcp_rd_p + txq->tx_count - txq->rd_p;
-
-	PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->rd_p: %u, qcp->rd_p: %u",
-		   qcp_rd_p, txq->rd_p, txq->rd_p);
-
-	if (todo == 0)
-		return todo;
-
-	txq->rd_p += todo;
-	if (unlikely(txq->rd_p >= txq->tx_count))
-		txq->rd_p -= txq->tx_count;
-
-	return todo;
-}
-
-/* Leaving always free descriptors for avoiding wrapping confusion */
-static inline
-uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq)
-{
-	if (txq->wr_p >= txq->rd_p)
-		return txq->tx_count - (txq->wr_p - txq->rd_p) - 8;
-	else
-		return txq->rd_p - txq->wr_p - 8;
-}
-
-/*
- * nfp_net_txq_full - Check if the TX queue free descriptors
- * is below tx_free_threshold
- *
- * @txq: TX queue to check
- *
- * This function uses the host copy* of read/write pointers
- */
-static inline
-uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
-{
-	return (nfp_free_tx_desc(txq) < txq->tx_free_thresh);
-}
-
-static uint16_t
-nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
-{
-	struct nfp_net_txq *txq;
-	struct nfp_net_hw *hw;
-	struct nfp_net_tx_desc *txds, txd;
-	struct rte_mbuf *pkt;
-	uint64_t dma_addr;
-	int pkt_size, dma_size;
-	uint16_t free_descs, issued_descs;
-	struct rte_mbuf **lmbuf;
-	int i;
-
-	txq = tx_queue;
-	hw = txq->hw;
-	txds = &txq->txds[txq->wr_p];
-
-	PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets",
-		   txq->qidx, txq->wr_p, nb_pkts);
-
-	if ((nfp_free_tx_desc(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
-		nfp_net_tx_free_bufs(txq);
-
-	free_descs = (uint16_t)nfp_free_tx_desc(txq);
-	if (unlikely(free_descs == 0))
-		return 0;
-
-	pkt = *tx_pkts;
-
-	i = 0;
-	issued_descs = 0;
-	PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets",
-		   txq->qidx, nb_pkts);
-	/* Sending packets */
-	while ((i < nb_pkts) && free_descs) {
-		/* Grabbing the mbuf linked to the current descriptor */
-		lmbuf = &txq->txbufs[txq->wr_p].mbuf;
-		/* Warming the cache for releasing the mbuf later on */
-		RTE_MBUF_PREFETCH_TO_FREE(*lmbuf);
-
-		pkt = *(tx_pkts + i);
-
-		if (unlikely((pkt->nb_segs > 1) &&
-			     !(hw->cap & NFP_NET_CFG_CTRL_GATHER))) {
-			PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set");
-			rte_panic("Multisegment packet unsupported\n");
-		}
-
-		/* Checking if we have enough descriptors */
-		if (unlikely(pkt->nb_segs > free_descs))
-			goto xmit_end;
-
-		/*
-		 * Checksum and VLAN flags just in the first descriptor for a
-		 * multisegment packet, but TSO info needs to be in all of them.
-		 */
-		txd.data_len = pkt->pkt_len;
-		nfp_net_tx_tso(txq, &txd, pkt);
-		nfp_net_tx_cksum(txq, &txd, pkt);
-
-		if ((pkt->ol_flags & PKT_TX_VLAN_PKT) &&
-		    (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)) {
-			txd.flags |= PCIE_DESC_TX_VLAN;
-			txd.vlan = pkt->vlan_tci;
-		}
-
-		/*
-		 * mbuf data_len is the data in one segment and pkt_len data
-		 * in the whole packet. When the packet is just one segment,
-		 * then data_len = pkt_len
-		 */
-		pkt_size = pkt->pkt_len;
-
-		while (pkt) {
-			/* Copying TSO, VLAN and cksum info */
-			*txds = txd;
-
-			/* Releasing mbuf used by this descriptor previously*/
-			if (*lmbuf)
-				rte_pktmbuf_free_seg(*lmbuf);
-
-			/*
-			 * Linking mbuf with descriptor for being released
-			 * next time descriptor is used
-			 */
-			*lmbuf = pkt;
-
-			dma_size = pkt->data_len;
-			dma_addr = rte_mbuf_data_iova(pkt);
-			PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
-				   "%" PRIx64 "", dma_addr);
-
-			/* Filling descriptors fields */
-			txds->dma_len = dma_size;
-			txds->data_len = txd.data_len;
-			txds->dma_addr_hi = (dma_addr >> 32) & 0xff;
-			txds->dma_addr_lo = (dma_addr & 0xffffffff);
-			ASSERT(free_descs > 0);
-			free_descs--;
-
-			txq->wr_p++;
-			if (unlikely(txq->wr_p == txq->tx_count)) /* wrapping?*/
-				txq->wr_p = 0;
-
-			pkt_size -= dma_size;
-
-			/*
-			 * Making the EOP, packets with just one segment
-			 * the priority
-			 */
-			if (likely(!pkt_size))
-				txds->offset_eop = PCIE_DESC_TX_EOP;
-			else
-				txds->offset_eop = 0;
-
-			pkt = pkt->next;
-			/* Referencing next free TX descriptor */
-			txds = &txq->txds[txq->wr_p];
-			lmbuf = &txq->txbufs[txq->wr_p].mbuf;
-			issued_descs++;
-		}
-		i++;
-	}
-
-xmit_end:
-	/* Increment write pointers. Force memory write before we let HW know */
-	rte_wmb();
-	nfp_qcp_ptr_add(txq->qcp_q, NFP_QCP_WRITE_PTR, issued_descs);
-
-	return i;
-}
-
 static int
 nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 {
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index a3a3ba32d6..9265496bf0 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -41,6 +41,12 @@ struct nfp_net_adapter;
 #define NFP_QCP_QUEUE_STS_HI                    0x000c
 #define NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask    (0x3ffff)
 
+/* The offset of the queue controller queues in the PCIe Target */
+#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
+
+/* Maximum value which can be added to a queue with one transaction */
+#define NFP_QCP_MAX_ADD	0x7f
+
 /* Interrupt definitions */
 #define NFP_NET_IRQ_LSC_IDX             0
 
@@ -95,47 +101,11 @@ struct nfp_net_adapter;
 #include <linux/types.h>
 #include <rte_io.h>
 
-static inline uint8_t nn_readb(volatile const void *addr)
-{
-	return rte_read8(addr);
-}
-
-static inline void nn_writeb(uint8_t val, volatile void *addr)
-{
-	rte_write8(val, addr);
-}
-
-static inline uint32_t nn_readl(volatile const void *addr)
-{
-	return rte_read32(addr);
-}
-
-static inline void nn_writel(uint32_t val, volatile void *addr)
-{
-	rte_write32(val, addr);
-}
-
-static inline void nn_writew(uint16_t val, volatile void *addr)
-{
-	rte_write16(val, addr);
-}
-
-static inline uint64_t nn_readq(volatile void *addr)
-{
-	const volatile uint32_t *p = addr;
-	uint32_t low, high;
-
-	high = nn_readl((volatile const void *)(p + 1));
-	low = nn_readl((volatile const void *)p);
-
-	return low + ((uint64_t)high << 32);
-}
-
-static inline void nn_writeq(uint64_t val, volatile void *addr)
-{
-	nn_writel(val >> 32, (volatile char *)addr + 4);
-	nn_writel(val, addr);
-}
+/* nfp_qcp_ptr - Read or Write Pointer of a queue */
+enum nfp_qcp_ptr {
+	NFP_QCP_READ_PTR = 0,
+	NFP_QCP_WRITE_PTR
+};
 
 struct nfp_pf_dev {
 	/* Backpointer to associated pci device */
@@ -247,6 +217,138 @@ struct nfp_net_adapter {
 	struct nfp_net_hw hw;
 };
 
+static inline uint8_t nn_readb(volatile const void *addr)
+{
+	return rte_read8(addr);
+}
+
+static inline void nn_writeb(uint8_t val, volatile void *addr)
+{
+	rte_write8(val, addr);
+}
+
+static inline uint32_t nn_readl(volatile const void *addr)
+{
+	return rte_read32(addr);
+}
+
+static inline void nn_writel(uint32_t val, volatile void *addr)
+{
+	rte_write32(val, addr);
+}
+
+static inline void nn_writew(uint16_t val, volatile void *addr)
+{
+	rte_write16(val, addr);
+}
+
+static inline uint64_t nn_readq(volatile void *addr)
+{
+	const volatile uint32_t *p = addr;
+	uint32_t low, high;
+
+	high = nn_readl((volatile const void *)(p + 1));
+	low = nn_readl((volatile const void *)p);
+
+	return low + ((uint64_t)high << 32);
+}
+
+static inline void nn_writeq(uint64_t val, volatile void *addr)
+{
+	nn_writel(val >> 32, (volatile char *)addr + 4);
+	nn_writel(val, addr);
+}
+
+/*
+ * Functions to read/write from/to Config BAR
+ * Performs any endian conversion necessary.
+ */
+static inline uint8_t
+nn_cfg_readb(struct nfp_net_hw *hw, int off)
+{
+	return nn_readb(hw->ctrl_bar + off);
+}
+
+static inline void
+nn_cfg_writeb(struct nfp_net_hw *hw, int off, uint8_t val)
+{
+	nn_writeb(val, hw->ctrl_bar + off);
+}
+
+static inline uint32_t
+nn_cfg_readl(struct nfp_net_hw *hw, int off)
+{
+	return rte_le_to_cpu_32(nn_readl(hw->ctrl_bar + off));
+}
+
+static inline void
+nn_cfg_writel(struct nfp_net_hw *hw, int off, uint32_t val)
+{
+	nn_writel(rte_cpu_to_le_32(val), hw->ctrl_bar + off);
+}
+
+static inline uint64_t
+nn_cfg_readq(struct nfp_net_hw *hw, int off)
+{
+	return rte_le_to_cpu_64(nn_readq(hw->ctrl_bar + off));
+}
+
+static inline void
+nn_cfg_writeq(struct nfp_net_hw *hw, int off, uint64_t val)
+{
+	nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
+}
+
+/*
+ * nfp_qcp_ptr_add - Add the value to the selected pointer of a queue
+ * @q: Base address for queue structure
+ * @ptr: Add to the Read or Write pointer
+ * @val: Value to add to the queue pointer
+ *
+ * If @val is greater than @NFP_QCP_MAX_ADD multiple writes are performed.
+ */
+static inline void
+nfp_qcp_ptr_add(uint8_t *q, enum nfp_qcp_ptr ptr, uint32_t val)
+{
+	uint32_t off;
+
+	if (ptr == NFP_QCP_READ_PTR)
+		off = NFP_QCP_QUEUE_ADD_RPTR;
+	else
+		off = NFP_QCP_QUEUE_ADD_WPTR;
+
+	while (val > NFP_QCP_MAX_ADD) {
+		nn_writel(rte_cpu_to_le_32(NFP_QCP_MAX_ADD), q + off);
+		val -= NFP_QCP_MAX_ADD;
+}
+
+nn_writel(rte_cpu_to_le_32(val), q + off);
+}
+
+/*
+ * nfp_qcp_read - Read the current Read/Write pointer value for a queue
+ * @q:  Base address for queue structure
+ * @ptr: Read or Write pointer
+ */
+static inline uint32_t
+nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
+{
+	uint32_t off;
+	uint32_t val;
+
+	if (ptr == NFP_QCP_READ_PTR)
+		off = NFP_QCP_QUEUE_STS_LO;
+	else
+		off = NFP_QCP_QUEUE_STS_HI;
+
+	val = rte_cpu_to_le_32(nn_readl(q + off));
+
+	if (ptr == NFP_QCP_READ_PTR)
+		return val & NFP_QCP_QUEUE_STS_LO_READPTR_mask;
+	else
+		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
+}
+
 #define NFP_NET_DEV_PRIVATE_TO_HW(adapter)\
 	(&((struct nfp_net_adapter *)adapter)->hw)
 
diff --git a/drivers/net/nfp/nfp_rxtx.c b/drivers/net/nfp/nfp_rxtx.c
new file mode 100644
index 0000000000..9ee9e5c9a3
--- /dev/null
+++ b/drivers/net/nfp/nfp_rxtx.c
@@ -0,0 +1,1002 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_rxtx.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: Rx/Tx functions
+ */
+
+#include <ethdev_driver.h>
+#include <ethdev_pci.h>
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+
+/* Prototypes */
+static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
+static inline void nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq);
+static inline void nfp_net_set_hash(struct nfp_net_rxq *rxq,
+				    struct nfp_net_rx_desc *rxd,
+				    struct rte_mbuf *mbuf);
+static inline void nfp_net_rx_cksum(struct nfp_net_rxq *rxq,
+				    struct nfp_net_rx_desc *rxd,
+				    struct rte_mbuf *mb);
+static void nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq);
+static int nfp_net_tx_free_bufs(struct nfp_net_txq *txq);
+static void nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq);
+static inline uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq);
+static inline uint32_t nfp_net_txq_full(struct nfp_net_txq *txq);
+static inline void nfp_net_tx_tso(struct nfp_net_txq *txq,
+				  struct nfp_net_tx_desc *txd,
+				  struct rte_mbuf *mb);
+static inline void nfp_net_tx_cksum(struct nfp_net_txq *txq,
+				    struct nfp_net_tx_desc *txd,
+				    struct rte_mbuf *mb);
+
+static int
+nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq)
+{
+	struct nfp_net_rx_buff *rxe = rxq->rxbufs;
+	uint64_t dma_addr;
+	unsigned int i;
+
+	PMD_RX_LOG(DEBUG, "Fill Rx Freelist for %u descriptors",
+		   rxq->rx_count);
+
+	for (i = 0; i < rxq->rx_count; i++) {
+		struct nfp_net_rx_desc *rxd;
+		struct rte_mbuf *mbuf = rte_pktmbuf_alloc(rxq->mem_pool);
+
+		if (mbuf == NULL) {
+			PMD_DRV_LOG(ERR, "RX mbuf alloc failed queue_id=%u",
+				(unsigned int)rxq->qidx);
+			return -ENOMEM;
+		}
+
+		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(mbuf));
+
+		rxd = &rxq->rxds[i];
+		rxd->fld.dd = 0;
+		rxd->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
+		rxd->fld.dma_addr_lo = dma_addr & 0xffffffff;
+		rxe[i].mbuf = mbuf;
+		PMD_RX_LOG(DEBUG, "[%d]: %" PRIx64, i, dma_addr);
+	}
+
+	/* Make sure all writes are flushed before telling the hardware */
+	rte_wmb();
+
+	/* Not advertising the whole ring as the firmware gets confused if so */
+	PMD_RX_LOG(DEBUG, "Increment FL write pointer in %u",
+		   rxq->rx_count - 1);
+
+	nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, rxq->rx_count - 1);
+
+	return 0;
+}
+
+int
+nfp_net_rx_freelist_setup(struct rte_eth_dev *dev)
+{
+	int i;
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		if (nfp_net_rx_fill_freelist(dev->data->rx_queues[i]) < 0)
+			return -1;
+	}
+	return 0;
+}
+
+uint32_t
+nfp_net_rx_queue_count(struct rte_eth_dev *dev, uint16_t queue_idx)
+{
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_rx_desc *rxds;
+	uint32_t idx;
+	uint32_t count;
+
+	rxq = (struct nfp_net_rxq *)dev->data->rx_queues[queue_idx];
+
+	idx = rxq->rd_p;
+
+	count = 0;
+
+	/*
+	 * Other PMDs are just checking the DD bit in intervals of 4
+	 * descriptors and counting all four if the first has the DD
+	 * bit on. Of course, this is not accurate but can be good for
+	 * performance. But ideally that should be done in descriptors
+	 * chunks belonging to the same cache line
+	 */
+
+	while (count < rxq->rx_count) {
+		rxds = &rxq->rxds[idx];
+		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
+			break;
+
+		count++;
+		idx++;
+
+		/* Wrapping? */
+		if ((idx) == rxq->rx_count)
+			idx = 0;
+	}
+
+	return count;
+}
+
+static inline void
+nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq)
+{
+	rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
+}
+
+/*
+ * nfp_net_set_hash - Set mbuf hash data
+ *
+ * The RSS hash and hash-type are pre-pended to the packet data.
+ * Extract and decode it and set the mbuf fields.
+ */
+static inline void
+nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
+		 struct rte_mbuf *mbuf)
+{
+	struct nfp_net_hw *hw = rxq->hw;
+	uint8_t *meta_offset;
+	uint32_t meta_info;
+	uint32_t hash = 0;
+	uint32_t hash_type = 0;
+
+	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
+		return;
+
+	/* this is true for new firmwares */
+	if (likely(((hw->cap & NFP_NET_CFG_CTRL_RSS2) ||
+	    (NFD_CFG_MAJOR_VERSION_of(hw->ver) == 4)) &&
+	     NFP_DESC_META_LEN(rxd))) {
+		/*
+		 * new metadata api:
+		 * <----  32 bit  ----->
+		 * m    field type word
+		 * e     data field #2
+		 * t     data field #1
+		 * a     data field #0
+		 * ====================
+		 *    packet data
+		 *
+		 * Field type word contains up to 8 4bit field types
+		 * A 4bit field type refers to a data field word
+		 * A data field word can have several 4bit field types
+		 */
+		meta_offset = rte_pktmbuf_mtod(mbuf, uint8_t *);
+		meta_offset -= NFP_DESC_META_LEN(rxd);
+		meta_info = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+		meta_offset += 4;
+		/* NFP PMD just supports metadata for hashing */
+		switch (meta_info & NFP_NET_META_FIELD_MASK) {
+		case NFP_NET_META_HASH:
+			/* next field type is about the hash type */
+			meta_info >>= NFP_NET_META_FIELD_SIZE;
+			/* hash value is in the data field */
+			hash = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+			hash_type = meta_info & NFP_NET_META_FIELD_MASK;
+			break;
+		default:
+			/* Unsupported metadata can be a performance issue */
+			return;
+		}
+	} else {
+		if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
+			return;
+
+		hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
+		hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
+	}
+
+	mbuf->hash.rss = hash;
+	mbuf->ol_flags |= PKT_RX_RSS_HASH;
+
+	switch (hash_type) {
+	case NFP_NET_RSS_IPV4:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV4;
+		break;
+	case NFP_NET_RSS_IPV6:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6;
+		break;
+	case NFP_NET_RSS_IPV6_EX:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV4_TCP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV6_TCP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV4_UDP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	case NFP_NET_RSS_IPV6_UDP:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L3_IPV6_EXT;
+		break;
+	default:
+		mbuf->packet_type |= RTE_PTYPE_INNER_L4_MASK;
+	}
+}
+
+/* nfp_net_rx_cksum - set mbuf checksum flags based on RX descriptor flags */
+static inline void
+nfp_net_rx_cksum(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
+		 struct rte_mbuf *mb)
+{
+	struct nfp_net_hw *hw = rxq->hw;
+
+	if (!(hw->ctrl & NFP_NET_CFG_CTRL_RXCSUM))
+		return;
+
+	/* If IPv4 and IP checksum error, fail */
+	if (unlikely((rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM) &&
+	    !(rxd->rxd.flags & PCIE_DESC_RX_IP4_CSUM_OK)))
+		mb->ol_flags |= PKT_RX_IP_CKSUM_BAD;
+	else
+		mb->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+
+	/* If neither UDP nor TCP return */
+	if (!(rxd->rxd.flags & PCIE_DESC_RX_TCP_CSUM) &&
+	    !(rxd->rxd.flags & PCIE_DESC_RX_UDP_CSUM))
+		return;
+
+	if (likely(rxd->rxd.flags & PCIE_DESC_RX_L4_CSUM_OK))
+		mb->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
+	else
+		mb->ol_flags |= PKT_RX_L4_CKSUM_BAD;
+}
+
+/*
+ * RX path design:
+ *
+ * There are some decisions to take:
+ * 1) How to check DD RX descriptors bit
+ * 2) How and when to allocate new mbufs
+ *
+ * Current implementation checks just one single DD bit each loop. As each
+ * descriptor is 8 bytes, it is likely a good idea to check descriptors in
+ * a single cache line instead. Tests with this change have not shown any
+ * performance improvement but it requires further investigation. For example,
+ * depending on which descriptor is next, the number of descriptors could be
+ * less than 8 for just checking those in the same cache line. This implies
+ * extra work which could be counterproductive by itself. Indeed, last firmware
+ * changes are just doing this: writing several descriptors with the DD bit
+ * for saving PCIe bandwidth and DMA operations from the NFP.
+ *
+ * Mbuf allocation is done when a new packet is received. Then the descriptor
+ * is automatically linked with the new mbuf and the old one is given to the
+ * user. The main drawback with this design is mbuf allocation is heavier than
+ * using bulk allocations allowed by DPDK with rte_mempool_get_bulk. From the
+ * cache point of view it does not seem allocating the mbuf early on as we are
+ * doing now have any benefit at all. Again, tests with this change have not
+ * shown any improvement. Also, rte_mempool_get_bulk returns all or nothing
+ * so looking at the implications of this type of allocation should be studied
+ * deeply
+ */
+
+uint16_t
+nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_rx_desc *rxds;
+	struct nfp_net_rx_buff *rxb;
+	struct nfp_net_hw *hw;
+	struct rte_mbuf *mb;
+	struct rte_mbuf *new_mb;
+	uint16_t nb_hold;
+	uint64_t dma_addr;
+	int avail;
+
+	rxq = rx_queue;
+	if (unlikely(rxq == NULL)) {
+		/*
+		 * DPDK just checks the queue is lower than max queues
+		 * enabled. But the queue needs to be configured
+		 */
+		RTE_LOG_DP(ERR, PMD, "RX Bad queue\n");
+		return -EINVAL;
+	}
+
+	hw = rxq->hw;
+	avail = 0;
+	nb_hold = 0;
+
+	while (avail < nb_pkts) {
+		rxb = &rxq->rxbufs[rxq->rd_p];
+		if (unlikely(rxb == NULL)) {
+			RTE_LOG_DP(ERR, PMD, "rxb does not exist!\n");
+			break;
+		}
+
+		rxds = &rxq->rxds[rxq->rd_p];
+		if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
+			break;
+
+		/*
+		 * Memory barrier to ensure that we won't do other
+		 * reads before the DD bit.
+		 */
+		rte_rmb();
+
+		/*
+		 * We got a packet. Let's alloc a new mbuf for refilling the
+		 * free descriptor ring as soon as possible
+		 */
+		new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
+		if (unlikely(new_mb == NULL)) {
+			RTE_LOG_DP(DEBUG, PMD,
+			"RX mbuf alloc failed port_id=%u queue_id=%u\n",
+				rxq->port_id, (unsigned int)rxq->qidx);
+			nfp_net_mbuf_alloc_failed(rxq);
+			break;
+		}
+
+		nb_hold++;
+
+		/*
+		 * Grab the mbuf and refill the descriptor with the
+		 * previously allocated mbuf
+		 */
+		mb = rxb->mbuf;
+		rxb->mbuf = new_mb;
+
+		PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u",
+			   rxds->rxd.data_len, rxq->mbuf_size);
+
+		/* Size of this segment */
+		mb->data_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
+		/* Size of the whole packet. We just support 1 segment */
+		mb->pkt_len = rxds->rxd.data_len - NFP_DESC_META_LEN(rxds);
+
+		if (unlikely((mb->data_len + hw->rx_offset) >
+			     rxq->mbuf_size)) {
+			/*
+			 * This should not happen and the user has the
+			 * responsibility of avoiding it. But we have
+			 * to give some info about the error
+			 */
+			RTE_LOG_DP(ERR, PMD,
+				"mbuf overflow likely due to the RX offset.\n"
+				"\t\tYour mbuf size should have extra space for"
+				" RX offset=%u bytes.\n"
+				"\t\tCurrently you just have %u bytes available"
+				" but the received packet is %u bytes long",
+				hw->rx_offset,
+				rxq->mbuf_size - hw->rx_offset,
+				mb->data_len);
+			return -EINVAL;
+		}
+
+		/* Filling the received mbuf with packet info */
+		if (hw->rx_offset)
+			mb->data_off = RTE_PKTMBUF_HEADROOM + hw->rx_offset;
+		else
+			mb->data_off = RTE_PKTMBUF_HEADROOM +
+				       NFP_DESC_META_LEN(rxds);
+
+		/* No scatter mode supported */
+		mb->nb_segs = 1;
+		mb->next = NULL;
+
+		mb->port = rxq->port_id;
+
+		/* Checking the RSS flag */
+		nfp_net_set_hash(rxq, rxds, mb);
+
+		/* Checking the checksum flag */
+		nfp_net_rx_cksum(rxq, rxds, mb);
+
+		if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
+		    (hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
+			mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
+			mb->ol_flags |= PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
+		}
+
+		/* Adding the mbuf to the mbuf array passed by the app */
+		rx_pkts[avail++] = mb;
+
+		/* Now resetting and updating the descriptor */
+		rxds->vals[0] = 0;
+		rxds->vals[1] = 0;
+		dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(new_mb));
+		rxds->fld.dd = 0;
+		rxds->fld.dma_addr_hi = (dma_addr >> 32) & 0xff;
+		rxds->fld.dma_addr_lo = dma_addr & 0xffffffff;
+
+		rxq->rd_p++;
+		if (unlikely(rxq->rd_p == rxq->rx_count)) /* wrapping?*/
+			rxq->rd_p = 0;
+	}
+
+	if (nb_hold == 0)
+		return nb_hold;
+
+	PMD_RX_LOG(DEBUG, "RX  port_id=%u queue_id=%u, %d packets received",
+		   rxq->port_id, (unsigned int)rxq->qidx, nb_hold);
+
+	nb_hold += rxq->nb_rx_hold;
+
+	/*
+	 * FL descriptors needs to be written before incrementing the
+	 * FL queue WR pointer
+	 */
+	rte_wmb();
+	if (nb_hold > rxq->rx_free_thresh) {
+		PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u",
+			   rxq->port_id, (unsigned int)rxq->qidx,
+			   (unsigned int)nb_hold, (unsigned int)avail);
+		nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, nb_hold);
+		nb_hold = 0;
+	}
+	rxq->nb_rx_hold = nb_hold;
+
+	return avail;
+}
+
+static void
+nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq)
+{
+	unsigned int i;
+
+	if (rxq->rxbufs == NULL)
+		return;
+
+	for (i = 0; i < rxq->rx_count; i++) {
+		if (rxq->rxbufs[i].mbuf) {
+			rte_pktmbuf_free_seg(rxq->rxbufs[i].mbuf);
+			rxq->rxbufs[i].mbuf = NULL;
+		}
+	}
+}
+
+void
+nfp_net_rx_queue_release(void *rx_queue)
+{
+	struct nfp_net_rxq *rxq = rx_queue;
+
+	if (rxq) {
+		nfp_net_rx_queue_release_mbufs(rxq);
+		rte_free(rxq->rxbufs);
+		rte_free(rxq);
+	}
+}
+
+void
+nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
+{
+	nfp_net_rx_queue_release_mbufs(rxq);
+	rxq->rd_p = 0;
+	rxq->nb_rx_hold = 0;
+}
+
+int
+nfp_net_rx_queue_setup(struct rte_eth_dev *dev,
+		       uint16_t queue_idx, uint16_t nb_desc,
+		       unsigned int socket_id,
+		       const struct rte_eth_rxconf *rx_conf,
+		       struct rte_mempool *mp)
+{
+	const struct rte_memzone *tz;
+	struct nfp_net_rxq *rxq;
+	struct nfp_net_hw *hw;
+	uint32_t rx_desc_sz;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* Validating number of descriptors */
+	rx_desc_sz = nb_desc * sizeof(struct nfp_net_rx_desc);
+	if (rx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
+	    nb_desc > NFP_NET_MAX_RX_DESC ||
+	    nb_desc < NFP_NET_MIN_RX_DESC) {
+		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
+		return -EINVAL;
+	}
+
+	/*
+	 * Free memory prior to re-allocation if needed. This is the case after
+	 * calling nfp_net_stop
+	 */
+	if (dev->data->rx_queues[queue_idx]) {
+		nfp_net_rx_queue_release(dev->data->rx_queues[queue_idx]);
+		dev->data->rx_queues[queue_idx] = NULL;
+	}
+
+	/* Allocating rx queue data structure */
+	rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct nfp_net_rxq),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxq == NULL)
+		return -ENOMEM;
+
+	/* Hw queues mapping based on firmware configuration */
+	rxq->qidx = queue_idx;
+	rxq->fl_qcidx = queue_idx * hw->stride_rx;
+	rxq->rx_qcidx = rxq->fl_qcidx + (hw->stride_rx - 1);
+	rxq->qcp_fl = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->fl_qcidx);
+	rxq->qcp_rx = hw->rx_bar + NFP_QCP_QUEUE_OFF(rxq->rx_qcidx);
+
+	/*
+	 * Tracking mbuf size for detecting a potential mbuf overflow due to
+	 * RX offset
+	 */
+	rxq->mem_pool = mp;
+	rxq->mbuf_size = rxq->mem_pool->elt_size;
+	rxq->mbuf_size -= (sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM);
+	hw->flbufsz = rxq->mbuf_size;
+
+	rxq->rx_count = nb_desc;
+	rxq->port_id = dev->data->port_id;
+	rxq->rx_free_thresh = rx_conf->rx_free_thresh;
+	rxq->drop_en = rx_conf->rx_drop_en;
+
+	/*
+	 * Allocate RX ring hardware descriptors. A memzone large enough to
+	 * handle the maximum ring size is allocated in order to allow for
+	 * resizing in later calls to the queue setup function.
+	 */
+	tz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
+				   sizeof(struct nfp_net_rx_desc) *
+				   NFP_NET_MAX_RX_DESC, NFP_MEMZONE_ALIGN,
+				   socket_id);
+
+	if (tz == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating rx dma");
+		nfp_net_rx_queue_release(rxq);
+		return -ENOMEM;
+	}
+
+	/* Saving physical and virtual addresses for the RX ring */
+	rxq->dma = (uint64_t)tz->iova;
+	rxq->rxds = (struct nfp_net_rx_desc *)tz->addr;
+
+	/* mbuf pointers array for referencing mbufs linked to RX descriptors */
+	rxq->rxbufs = rte_zmalloc_socket("rxq->rxbufs",
+					 sizeof(*rxq->rxbufs) * nb_desc,
+					 RTE_CACHE_LINE_SIZE, socket_id);
+	if (rxq->rxbufs == NULL) {
+		nfp_net_rx_queue_release(rxq);
+		return -ENOMEM;
+	}
+
+	PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
+		   rxq->rxbufs, rxq->rxds, (unsigned long)rxq->dma);
+
+	nfp_net_reset_rx_queue(rxq);
+
+	dev->data->rx_queues[queue_idx] = rxq;
+	rxq->hw = hw;
+
+	/*
+	 * Telling the HW about the physical address of the RX ring and number
+	 * of descriptors in log2 format
+	 */
+	nn_cfg_writeq(hw, NFP_NET_CFG_RXR_ADDR(queue_idx), rxq->dma);
+	nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), rte_log2_u32(nb_desc));
+
+	return 0;
+}
+
+/*
+ * nfp_net_tx_free_bufs - Check for descriptors with a complete
+ * status
+ * @txq: TX queue to work with
+ * Returns number of descriptors freed
+ */
+static int
+nfp_net_tx_free_bufs(struct nfp_net_txq *txq)
+{
+	uint32_t qcp_rd_p;
+	int todo;
+
+	PMD_TX_LOG(DEBUG, "queue %u. Check for descriptor with a complete"
+		   " status", txq->qidx);
+
+	/* Work out how many packets have been sent */
+	qcp_rd_p = nfp_qcp_read(txq->qcp_q, NFP_QCP_READ_PTR);
+
+	if (qcp_rd_p == txq->rd_p) {
+		PMD_TX_LOG(DEBUG, "queue %u: It seems harrier is not sending "
+			   "packets (%u, %u)", txq->qidx,
+			   qcp_rd_p, txq->rd_p);
+		return 0;
+	}
+
+	if (qcp_rd_p > txq->rd_p)
+		todo = qcp_rd_p - txq->rd_p;
+	else
+		todo = qcp_rd_p + txq->tx_count - txq->rd_p;
+
+	PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->rd_p: %u, qcp->rd_p: %u",
+		   qcp_rd_p, txq->rd_p, txq->rd_p);
+
+	if (todo == 0)
+		return todo;
+
+	txq->rd_p += todo;
+	if (unlikely(txq->rd_p >= txq->tx_count))
+		txq->rd_p -= txq->tx_count;
+
+	return todo;
+}
+
+static void
+nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq)
+{
+	unsigned int i;
+
+	if (txq->txbufs == NULL)
+		return;
+
+	for (i = 0; i < txq->tx_count; i++) {
+		if (txq->txbufs[i].mbuf) {
+			rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
+			txq->txbufs[i].mbuf = NULL;
+		}
+	}
+}
+
+void
+nfp_net_tx_queue_release(void *tx_queue)
+{
+	struct nfp_net_txq *txq = tx_queue;
+
+	if (txq) {
+		nfp_net_tx_queue_release_mbufs(txq);
+		rte_free(txq->txbufs);
+		rte_free(txq);
+	}
+}
+
+void
+nfp_net_reset_tx_queue(struct nfp_net_txq *txq)
+{
+	nfp_net_tx_queue_release_mbufs(txq);
+	txq->wr_p = 0;
+	txq->rd_p = 0;
+}
+
+int
+nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+		       uint16_t nb_desc, unsigned int socket_id,
+		       const struct rte_eth_txconf *tx_conf)
+{
+	const struct rte_memzone *tz;
+	struct nfp_net_txq *txq;
+	uint16_t tx_free_thresh;
+	struct nfp_net_hw *hw;
+	uint32_t tx_desc_sz;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_FUNC_TRACE();
+
+	/* Validating number of descriptors */
+	tx_desc_sz = nb_desc * sizeof(struct nfp_net_tx_desc);
+	if (tx_desc_sz % NFP_ALIGN_RING_DESC != 0 ||
+	    nb_desc > NFP_NET_MAX_TX_DESC ||
+	    nb_desc < NFP_NET_MIN_TX_DESC) {
+		PMD_DRV_LOG(ERR, "Wrong nb_desc value");
+		return -EINVAL;
+	}
+
+	tx_free_thresh = (uint16_t)((tx_conf->tx_free_thresh) ?
+				    tx_conf->tx_free_thresh :
+				    DEFAULT_TX_FREE_THRESH);
+
+	if (tx_free_thresh > (nb_desc)) {
+		PMD_DRV_LOG(ERR,
+			"tx_free_thresh must be less than the number of TX "
+			"descriptors. (tx_free_thresh=%u port=%d "
+			"queue=%d)", (unsigned int)tx_free_thresh,
+			dev->data->port_id, (int)queue_idx);
+		return -(EINVAL);
+	}
+
+	/*
+	 * Free memory prior to re-allocation if needed. This is the case after
+	 * calling nfp_net_stop
+	 */
+	if (dev->data->tx_queues[queue_idx]) {
+		PMD_TX_LOG(DEBUG, "Freeing memory prior to re-allocation %d",
+			   queue_idx);
+		nfp_net_tx_queue_release(dev->data->tx_queues[queue_idx]);
+		dev->data->tx_queues[queue_idx] = NULL;
+	}
+
+	/* Allocating tx queue data structure */
+	txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct nfp_net_txq),
+				 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating tx dma");
+		return -ENOMEM;
+	}
+
+	/*
+	 * Allocate TX ring hardware descriptors. A memzone large enough to
+	 * handle the maximum ring size is allocated in order to allow for
+	 * resizing in later calls to the queue setup function.
+	 */
+	tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
+				   sizeof(struct nfp_net_tx_desc) *
+				   NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN,
+				   socket_id);
+	if (tz == NULL) {
+		PMD_DRV_LOG(ERR, "Error allocating tx dma");
+		nfp_net_tx_queue_release(txq);
+		return -ENOMEM;
+	}
+
+	txq->tx_count = nb_desc;
+	txq->tx_free_thresh = tx_free_thresh;
+	txq->tx_pthresh = tx_conf->tx_thresh.pthresh;
+	txq->tx_hthresh = tx_conf->tx_thresh.hthresh;
+	txq->tx_wthresh = tx_conf->tx_thresh.wthresh;
+
+	/* queue mapping based on firmware configuration */
+	txq->qidx = queue_idx;
+	txq->tx_qcidx = queue_idx * hw->stride_tx;
+	txq->qcp_q = hw->tx_bar + NFP_QCP_QUEUE_OFF(txq->tx_qcidx);
+
+	txq->port_id = dev->data->port_id;
+
+	/* Saving physical and virtual addresses for the TX ring */
+	txq->dma = (uint64_t)tz->iova;
+	txq->txds = (struct nfp_net_tx_desc *)tz->addr;
+
+	/* mbuf pointers array for referencing mbufs linked to TX descriptors */
+	txq->txbufs = rte_zmalloc_socket("txq->txbufs",
+					 sizeof(*txq->txbufs) * nb_desc,
+					 RTE_CACHE_LINE_SIZE, socket_id);
+	if (txq->txbufs == NULL) {
+		nfp_net_tx_queue_release(txq);
+		return -ENOMEM;
+	}
+	PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64,
+		   txq->txbufs, txq->txds, (unsigned long)txq->dma);
+
+	nfp_net_reset_tx_queue(txq);
+
+	dev->data->tx_queues[queue_idx] = txq;
+	txq->hw = hw;
+
+	/*
+	 * Telling the HW about the physical address of the TX ring and number
+	 * of descriptors in log2 format
+	 */
+	nn_cfg_writeq(hw, NFP_NET_CFG_TXR_ADDR(queue_idx), txq->dma);
+	nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), rte_log2_u32(nb_desc));
+
+	return 0;
+}
+
+/* Leaving always free descriptors for avoiding wrapping confusion */
+static inline
+uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq)
+{
+	if (txq->wr_p >= txq->rd_p)
+		return txq->tx_count - (txq->wr_p - txq->rd_p) - 8;
+	else
+		return txq->rd_p - txq->wr_p - 8;
+}
+
+/*
+ * nfp_net_txq_full - Check if the TX queue free descriptors
+ * is below tx_free_threshold
+ *
+ * @txq: TX queue to check
+ *
+ * This function uses the host copy* of read/write pointers
+ */
+static inline
+uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
+{
+	return (nfp_free_tx_desc(txq) < txq->tx_free_thresh);
+}
+
+/* nfp_net_tx_tso - Set TX descriptor for TSO */
+static inline void
+nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+	       struct rte_mbuf *mb)
+{
+	uint64_t ol_flags;
+	struct nfp_net_hw *hw = txq->hw;
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY))
+		goto clean_txd;
+
+	ol_flags = mb->ol_flags;
+
+	if (!(ol_flags & PKT_TX_TCP_SEG))
+		goto clean_txd;
+
+	txd->l3_offset = mb->l2_len;
+	txd->l4_offset = mb->l2_len + mb->l3_len;
+	txd->lso_hdrlen = mb->l2_len + mb->l3_len + mb->l4_len;
+	txd->mss = rte_cpu_to_le_16(mb->tso_segsz);
+	txd->flags = PCIE_DESC_TX_LSO;
+	return;
+
+clean_txd:
+	txd->flags = 0;
+	txd->l3_offset = 0;
+	txd->l4_offset = 0;
+	txd->lso_hdrlen = 0;
+	txd->mss = 0;
+}
+
+/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
+static inline void
+nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+		 struct rte_mbuf *mb)
+{
+	uint64_t ol_flags;
+	struct nfp_net_hw *hw = txq->hw;
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_TXCSUM))
+		return;
+
+	ol_flags = mb->ol_flags;
+
+	/* IPv6 does not need checksum */
+	if (ol_flags & PKT_TX_IP_CKSUM)
+		txd->flags |= PCIE_DESC_TX_IP4_CSUM;
+
+	switch (ol_flags & PKT_TX_L4_MASK) {
+	case PKT_TX_UDP_CKSUM:
+		txd->flags |= PCIE_DESC_TX_UDP_CSUM;
+		break;
+	case PKT_TX_TCP_CKSUM:
+		txd->flags |= PCIE_DESC_TX_TCP_CSUM;
+		break;
+	}
+
+	if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
+		txd->flags |= PCIE_DESC_TX_CSUM;
+}
+
+uint16_t
+nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+	struct nfp_net_txq *txq;
+	struct nfp_net_hw *hw;
+	struct nfp_net_tx_desc *txds, txd;
+	struct rte_mbuf *pkt;
+	uint64_t dma_addr;
+	int pkt_size, dma_size;
+	uint16_t free_descs, issued_descs;
+	struct rte_mbuf **lmbuf;
+	int i;
+
+	txq = tx_queue;
+	hw = txq->hw;
+	txds = &txq->txds[txq->wr_p];
+
+	PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets",
+		   txq->qidx, txq->wr_p, nb_pkts);
+
+	if ((nfp_free_tx_desc(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
+		nfp_net_tx_free_bufs(txq);
+
+	free_descs = (uint16_t)nfp_free_tx_desc(txq);
+	if (unlikely(free_descs == 0))
+		return 0;
+
+	pkt = *tx_pkts;
+
+	i = 0;
+	issued_descs = 0;
+	PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets",
+		   txq->qidx, nb_pkts);
+	/* Sending packets */
+	while ((i < nb_pkts) && free_descs) {
+		/* Grabbing the mbuf linked to the current descriptor */
+		lmbuf = &txq->txbufs[txq->wr_p].mbuf;
+		/* Warming the cache for releasing the mbuf later on */
+		RTE_MBUF_PREFETCH_TO_FREE(*lmbuf);
+
+		pkt = *(tx_pkts + i);
+
+		if (unlikely(pkt->nb_segs > 1 &&
+			     !(hw->cap & NFP_NET_CFG_CTRL_GATHER))) {
+			PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set");
+			rte_panic("Multisegment packet unsupported\n");
+		}
+
+		/* Checking if we have enough descriptors */
+		if (unlikely(pkt->nb_segs > free_descs))
+			goto xmit_end;
+
+		/*
+		 * Checksum and VLAN flags just in the first descriptor for a
+		 * multisegment packet, but TSO info needs to be in all of them.
+		 */
+		txd.data_len = pkt->pkt_len;
+		nfp_net_tx_tso(txq, &txd, pkt);
+		nfp_net_tx_cksum(txq, &txd, pkt);
+
+		if ((pkt->ol_flags & PKT_TX_VLAN_PKT) &&
+		    (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)) {
+			txd.flags |= PCIE_DESC_TX_VLAN;
+			txd.vlan = pkt->vlan_tci;
+		}
+
+		/*
+		 * mbuf data_len is the data in one segment and pkt_len data
+		 * in the whole packet. When the packet is just one segment,
+		 * then data_len = pkt_len
+		 */
+		pkt_size = pkt->pkt_len;
+
+		while (pkt) {
+			/* Copying TSO, VLAN and cksum info */
+			*txds = txd;
+
+			/* Releasing mbuf used by this descriptor previously*/
+			if (*lmbuf)
+				rte_pktmbuf_free_seg(*lmbuf);
+
+			/*
+			 * Linking mbuf with descriptor for being released
+			 * next time descriptor is used
+			 */
+			*lmbuf = pkt;
+
+			dma_size = pkt->data_len;
+			dma_addr = rte_mbuf_data_iova(pkt);
+			PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
+				   "%" PRIx64 "", dma_addr);
+
+			/* Filling descriptors fields */
+			txds->dma_len = dma_size;
+			txds->data_len = txd.data_len;
+			txds->dma_addr_hi = (dma_addr >> 32) & 0xff;
+			txds->dma_addr_lo = (dma_addr & 0xffffffff);
+			ASSERT(free_descs > 0);
+			free_descs--;
+
+			txq->wr_p++;
+			if (unlikely(txq->wr_p == txq->tx_count)) /* wrapping?*/
+				txq->wr_p = 0;
+
+			pkt_size -= dma_size;
+
+			/*
+			 * Making the EOP, packets with just one segment
+			 * the priority
+			 */
+			if (likely(!pkt_size))
+				txds->offset_eop = PCIE_DESC_TX_EOP;
+			else
+				txds->offset_eop = 0;
+
+			pkt = pkt->next;
+			/* Referencing next free TX descriptor */
+			txds = &txq->txds[txq->wr_p];
+			lmbuf = &txq->txbufs[txq->wr_p].mbuf;
+			issued_descs++;
+		}
+		i++;
+	}
+
+xmit_end:
+	/* Increment write pointers. Force memory write before we let HW know */
+	rte_wmb();
+	nfp_qcp_ptr_add(txq->qcp_q, NFP_QCP_WRITE_PTR, issued_descs);
+
+	return i;
+}
diff --git a/drivers/net/nfp/nfp_rxtx.h b/drivers/net/nfp/nfp_rxtx.h
index 41a3a4b4e7..d2d0f3f175 100644
--- a/drivers/net/nfp/nfp_rxtx.h
+++ b/drivers/net/nfp/nfp_rxtx.h
@@ -17,6 +17,14 @@
 #include <linux/types.h>
 #include <rte_io.h>
 
+#define NFP_DESC_META_LEN(d) ((d)->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
+
+#define NFP_HASH_OFFSET      ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 4)
+#define NFP_HASH_TYPE_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 8)
+
+#define RTE_MBUF_DMA_ADDR_DEFAULT(mb) \
+	((uint64_t)((mb)->buf_iova + RTE_PKTMBUF_HEADROOM))
+
 /*
  * The maximum number of descriptors is limited by design as
  * DPDK uses uint16_t variables for these values
@@ -266,6 +274,25 @@ struct nfp_net_rxq {
 	int rx_qcidx;
 } __rte_aligned(64);
 
+int nfp_net_rx_freelist_setup(struct rte_eth_dev *dev);
+uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
+				       uint16_t queue_idx);
+uint16_t nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+				  uint16_t nb_pkts);
+void nfp_net_rx_queue_release(void *rxq);
+void nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq);
+int nfp_net_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+				  uint16_t nb_desc, unsigned int socket_id,
+				  const struct rte_eth_rxconf *rx_conf,
+				  struct rte_mempool *mp);
+void nfp_net_tx_queue_release(void *txq);
+void nfp_net_reset_tx_queue(struct nfp_net_txq *txq);
+int nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+				  uint16_t nb_desc, unsigned int socket_id,
+				  const struct rte_eth_txconf *tx_conf);
+uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
+				  uint16_t nb_pkts);
+
 #endif /* _NFP_RXTX_H_ */
 /*
  * Local variables:
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 3/7] net/nfp: move CPP bridge to a separate file
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
@ 2021-07-29 13:47     ` Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
                       ` (5 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

This commit moves the CPP bridge logic to a separate file. A new
corresponding header file is also created.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build      |   1 +
 drivers/net/nfp/nfp_cpp_bridge.c | 392 +++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_cpp_bridge.h |  36 +++
 drivers/net/nfp/nfp_net.c        | 367 +----------------------------
 4 files changed, 430 insertions(+), 366 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.c
 create mode 100644 drivers/net/nfp/nfp_cpp_bridge.h

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 1b289e2354..b46ac2d40f 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -20,4 +20,5 @@ sources = files(
         'nfpcore/nfp_hwinfo.c',
         'nfp_net.c',
         'nfp_rxtx.c',
+        'nfp_cpp_bridge.c',
 )
diff --git a/drivers/net/nfp/nfp_cpp_bridge.c b/drivers/net/nfp/nfp_cpp_bridge.c
new file mode 100644
index 0000000000..d916793338
--- /dev/null
+++ b/drivers/net/nfp/nfp_cpp_bridge.c
@@ -0,0 +1,392 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_cpp_bridge.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: CPP Bridge
+ */
+
+#include <rte_service_component.h>
+
+#include "nfpcore/nfp_cpp.h"
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_nsp.h"
+
+#include "nfp_net_logs.h"
+#include "nfp_cpp_bridge.h"
+
+#include <sys/ioctl.h>
+
+/* Prototypes */
+static int nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp);
+static int nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp);
+static int nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp);
+
+void nfp_register_cpp_service(struct nfp_cpp *cpp)
+{
+	uint32_t *cpp_service_id = NULL;
+	struct rte_service_spec service;
+
+	memset(&service, 0, sizeof(struct rte_service_spec));
+	snprintf(service.name, sizeof(service.name), "nfp_cpp_service");
+	service.callback = nfp_cpp_bridge_service_func;
+	service.callback_userdata = (void *)cpp;
+
+	if (rte_service_component_register(&service,
+					   cpp_service_id))
+		RTE_LOG(WARNING, PMD, "NFP CPP bridge service register() failed");
+	else
+		RTE_LOG(DEBUG, PMD, "NFP CPP bridge service registered");
+}
+
+/*
+ * Serving a write request to NFP from host programs. The request
+ * sends the write size and the CPP target. The bridge makes use
+ * of CPP interface handler configured by the PMD setup.
+ */
+static int
+nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp)
+{
+	struct nfp_cpp_area *area;
+	off_t offset, nfp_offset;
+	uint32_t cpp_id, pos, len;
+	uint32_t tmpbuf[16];
+	size_t count, curlen, totlen = 0;
+	int err = 0;
+
+	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
+		sizeof(off_t), sizeof(size_t));
+
+	/* Reading the count param */
+	err = recv(sockfd, &count, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	curlen = count;
+
+	/* Reading the offset param */
+	err = recv(sockfd, &offset, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	/* Obtain target's CPP ID and offset in target */
+	cpp_id = (offset >> 40) << 8;
+	nfp_offset = offset & ((1ull << 40) - 1);
+
+	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
+		offset);
+	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
+		cpp_id, nfp_offset);
+
+	/* Adjust length if not aligned */
+	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
+	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
+		curlen = NFP_CPP_MEMIO_BOUNDARY -
+			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
+	}
+
+	while (count > 0) {
+		/* configure a CPP PCIe2CPP BAR for mapping the CPP target */
+		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
+						    nfp_offset, curlen);
+		if (!area) {
+			RTE_LOG(ERR, PMD, "%s: area alloc fail\n", __func__);
+			return -EIO;
+		}
+
+		/* mapping the target */
+		err = nfp_cpp_area_acquire(area);
+		if (err < 0) {
+			RTE_LOG(ERR, PMD, "area acquire failed\n");
+			nfp_cpp_area_free(area);
+			return -EIO;
+		}
+
+		for (pos = 0; pos < curlen; pos += len) {
+			len = curlen - pos;
+			if (len > sizeof(tmpbuf))
+				len = sizeof(tmpbuf);
+
+			PMD_CPP_LOG(DEBUG, "%s: Receive %u of %zu\n", __func__,
+					   len, count);
+			err = recv(sockfd, tmpbuf, len, MSG_WAITALL);
+			if (err != (int)len) {
+				RTE_LOG(ERR, PMD,
+					"%s: error when receiving, %d of %zu\n",
+					__func__, err, count);
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+			err = nfp_cpp_area_write(area, pos, tmpbuf, len);
+			if (err < 0) {
+				RTE_LOG(ERR, PMD, "nfp_cpp_area_write error\n");
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+		}
+
+		nfp_offset += pos;
+		totlen += pos;
+		nfp_cpp_area_release(area);
+		nfp_cpp_area_free(area);
+
+		count -= pos;
+		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
+			 NFP_CPP_MEMIO_BOUNDARY : count;
+	}
+
+	return 0;
+}
+
+/*
+ * Serving a read request to NFP from host programs. The request
+ * sends the read size and the CPP target. The bridge makes use
+ * of CPP interface handler configured by the PMD setup. The read
+ * data is sent to the requester using the same socket.
+ */
+static int
+nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp)
+{
+	struct nfp_cpp_area *area;
+	off_t offset, nfp_offset;
+	uint32_t cpp_id, pos, len;
+	uint32_t tmpbuf[16];
+	size_t count, curlen, totlen = 0;
+	int err = 0;
+
+	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
+		sizeof(off_t), sizeof(size_t));
+
+	/* Reading the count param */
+	err = recv(sockfd, &count, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	curlen = count;
+
+	/* Reading the offset param */
+	err = recv(sockfd, &offset, sizeof(off_t), 0);
+	if (err != sizeof(off_t))
+		return -EINVAL;
+
+	/* Obtain target's CPP ID and offset in target */
+	cpp_id = (offset >> 40) << 8;
+	nfp_offset = offset & ((1ull << 40) - 1);
+
+	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
+			   offset);
+	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
+			   cpp_id, nfp_offset);
+
+	/* Adjust length if not aligned */
+	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
+	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
+		curlen = NFP_CPP_MEMIO_BOUNDARY -
+			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
+	}
+
+	while (count > 0) {
+		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
+						    nfp_offset, curlen);
+		if (!area) {
+			RTE_LOG(ERR, PMD, "%s: area alloc failed\n", __func__);
+			return -EIO;
+		}
+
+		err = nfp_cpp_area_acquire(area);
+		if (err < 0) {
+			RTE_LOG(ERR, PMD, "area acquire failed\n");
+			nfp_cpp_area_free(area);
+			return -EIO;
+		}
+
+		for (pos = 0; pos < curlen; pos += len) {
+			len = curlen - pos;
+			if (len > sizeof(tmpbuf))
+				len = sizeof(tmpbuf);
+
+			err = nfp_cpp_area_read(area, pos, tmpbuf, len);
+			if (err < 0) {
+				RTE_LOG(ERR, PMD, "nfp_cpp_area_read error\n");
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+			PMD_CPP_LOG(DEBUG, "%s: sending %u of %zu\n", __func__,
+					   len, count);
+
+			err = send(sockfd, tmpbuf, len, 0);
+			if (err != (int)len) {
+				RTE_LOG(ERR, PMD,
+					"%s: error when sending: %d of %zu\n",
+					__func__, err, count);
+				nfp_cpp_area_release(area);
+				nfp_cpp_area_free(area);
+				return -EIO;
+			}
+		}
+
+		nfp_offset += pos;
+		totlen += pos;
+		nfp_cpp_area_release(area);
+		nfp_cpp_area_free(area);
+
+		count -= pos;
+		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
+			NFP_CPP_MEMIO_BOUNDARY : count;
+	}
+	return 0;
+}
+
+/*
+ * Serving a ioctl command from host NFP tools. This usually goes to
+ * a kernel driver char driver but it is not available when the PF is
+ * bound to the PMD. Currently just one ioctl command is served and it
+ * does not require any CPP access at all.
+ */
+static int
+nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp)
+{
+	uint32_t cmd, ident_size, tmp;
+	int err;
+
+	/* Reading now the IOCTL command */
+	err = recv(sockfd, &cmd, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
+		return -EIO;
+	}
+
+	/* Only supporting NFP_IOCTL_CPP_IDENTIFICATION */
+	if (cmd != NFP_IOCTL_CPP_IDENTIFICATION) {
+		RTE_LOG(ERR, PMD, "%s: unknown cmd %d\n", __func__, cmd);
+		return -EINVAL;
+	}
+
+	err = recv(sockfd, &ident_size, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
+		return -EIO;
+	}
+
+	tmp = nfp_cpp_model(cpp);
+
+	PMD_CPP_LOG(DEBUG, "%s: sending NFP model %08x\n", __func__, tmp);
+
+	err = send(sockfd, &tmp, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
+		return -EIO;
+	}
+
+	tmp = cpp->interface;
+
+	PMD_CPP_LOG(DEBUG, "%s: sending NFP interface %08x\n", __func__, tmp);
+
+	err = send(sockfd, &tmp, 4, 0);
+	if (err != 4) {
+		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
+		return -EIO;
+	}
+
+	return 0;
+}
+
+/*
+ * This is the code to be executed by a service core. The CPP bridge interface
+ * is based on a unix socket and requests usually received by a kernel char
+ * driver, read, write and ioctl, are handled by the CPP bridge. NFP host tools
+ * can be executed with a wrapper library and LD_LIBRARY being completely
+ * unaware of the CPP bridge performing the NFP kernel char driver for CPP
+ * accesses.
+ */
+int32_t
+nfp_cpp_bridge_service_func(void *args)
+{
+	struct sockaddr address;
+	struct nfp_cpp *cpp = args;
+	int sockfd, datafd, op, ret;
+
+	unlink("/tmp/nfp_cpp");
+	sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
+	if (sockfd < 0) {
+		RTE_LOG(ERR, PMD, "%s: socket creation error. Service failed\n",
+			__func__);
+		return -EIO;
+	}
+
+	memset(&address, 0, sizeof(struct sockaddr));
+
+	address.sa_family = AF_UNIX;
+	strcpy(address.sa_data, "/tmp/nfp_cpp");
+
+	ret = bind(sockfd, (const struct sockaddr *)&address,
+		   sizeof(struct sockaddr));
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "%s: bind error (%d). Service failed\n",
+				  __func__, errno);
+		close(sockfd);
+		return ret;
+	}
+
+	ret = listen(sockfd, 20);
+	if (ret < 0) {
+		RTE_LOG(ERR, PMD, "%s: listen error(%d). Service failed\n",
+				  __func__, errno);
+		close(sockfd);
+		return ret;
+	}
+
+	for (;;) {
+		datafd = accept(sockfd, NULL, NULL);
+		if (datafd < 0) {
+			RTE_LOG(ERR, PMD, "%s: accept call error (%d)\n",
+					  __func__, errno);
+			RTE_LOG(ERR, PMD, "%s: service failed\n", __func__);
+			close(sockfd);
+			return -EIO;
+		}
+
+		while (1) {
+			ret = recv(datafd, &op, 4, 0);
+			if (ret <= 0) {
+				PMD_CPP_LOG(DEBUG, "%s: socket close\n",
+						   __func__);
+				break;
+			}
+
+			PMD_CPP_LOG(DEBUG, "%s: getting op %u\n", __func__, op);
+
+			if (op == NFP_BRIDGE_OP_READ)
+				nfp_cpp_bridge_serve_read(datafd, cpp);
+
+			if (op == NFP_BRIDGE_OP_WRITE)
+				nfp_cpp_bridge_serve_write(datafd, cpp);
+
+			if (op == NFP_BRIDGE_OP_IOCTL)
+				nfp_cpp_bridge_serve_ioctl(datafd, cpp);
+
+			if (op == 0)
+				break;
+		}
+		close(datafd);
+	}
+	close(sockfd);
+
+	return 0;
+}
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_cpp_bridge.h b/drivers/net/nfp/nfp_cpp_bridge.h
new file mode 100644
index 0000000000..aea5fdc784
--- /dev/null
+++ b/drivers/net/nfp/nfp_cpp_bridge.h
@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_cpp_bridge.h
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: CPP Bridge header file
+ */
+
+#ifndef _NFP_CPP_BRIDGE_H_
+#define _NFP_CPP_BRIDGE_H_
+
+#define NFP_CPP_MEMIO_BOUNDARY	(1 << 20)
+#define NFP_BRIDGE_OP_READ	20
+#define NFP_BRIDGE_OP_WRITE	30
+#define NFP_BRIDGE_OP_IOCTL	40
+
+#define NFP_IOCTL 'n'
+#define NFP_IOCTL_CPP_IDENTIFICATION _IOW(NFP_IOCTL, 0x8f, uint32_t)
+
+void nfp_register_cpp_service(struct nfp_cpp *cpp);
+int32_t nfp_cpp_bridge_service_func(void *args);
+
+#endif /* _NFP_CPP_BRIDGE_H_ */
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index de0de808f4..af94219729 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -43,6 +43,7 @@
 #include "nfp_rxtx.h"
 #include "nfp_net_logs.h"
 #include "nfp_net_ctrl.h"
+#include "nfp_cpp_bridge.h"
 
 #include <sys/types.h>
 #include <sys/socket.h>
@@ -83,8 +84,6 @@ static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf);
 static int nfp_set_mac_addr(struct rte_eth_dev *dev,
 			     struct rte_ether_addr *mac_addr);
-static int32_t nfp_cpp_bridge_service_func(void *args);
-static void nfp_register_cpp_service(struct nfp_cpp *cpp);
 static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_cpp *cpp,
 			struct nfp_eth_table *nfp_eth_table,
@@ -1928,353 +1927,6 @@ nfp_net_init(struct rte_eth_dev *eth_dev)
 	return err;
 }
 
-#define NFP_CPP_MEMIO_BOUNDARY		(1 << 20)
-
-/*
- * Serving a write request to NFP from host programs. The request
- * sends the write size and the CPP target. The bridge makes use
- * of CPP interface handler configured by the PMD setup.
- */
-static int
-nfp_cpp_bridge_serve_write(int sockfd, struct nfp_cpp *cpp)
-{
-	struct nfp_cpp_area *area;
-	off_t offset, nfp_offset;
-	uint32_t cpp_id, pos, len;
-	uint32_t tmpbuf[16];
-	size_t count, curlen, totlen = 0;
-	int err = 0;
-
-	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
-		sizeof(off_t), sizeof(size_t));
-
-	/* Reading the count param */
-	err = recv(sockfd, &count, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	curlen = count;
-
-	/* Reading the offset param */
-	err = recv(sockfd, &offset, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	/* Obtain target's CPP ID and offset in target */
-	cpp_id = (offset >> 40) << 8;
-	nfp_offset = offset & ((1ull << 40) - 1);
-
-	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
-		offset);
-	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
-		cpp_id, nfp_offset);
-
-	/* Adjust length if not aligned */
-	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
-	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
-		curlen = NFP_CPP_MEMIO_BOUNDARY -
-			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
-	}
-
-	while (count > 0) {
-		/* configure a CPP PCIe2CPP BAR for mapping the CPP target */
-		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
-						    nfp_offset, curlen);
-		if (!area) {
-			RTE_LOG(ERR, PMD, "%s: area alloc fail\n", __func__);
-			return -EIO;
-		}
-
-		/* mapping the target */
-		err = nfp_cpp_area_acquire(area);
-		if (err < 0) {
-			RTE_LOG(ERR, PMD, "area acquire failed\n");
-			nfp_cpp_area_free(area);
-			return -EIO;
-		}
-
-		for (pos = 0; pos < curlen; pos += len) {
-			len = curlen - pos;
-			if (len > sizeof(tmpbuf))
-				len = sizeof(tmpbuf);
-
-			PMD_CPP_LOG(DEBUG, "%s: Receive %u of %zu\n", __func__,
-					   len, count);
-			err = recv(sockfd, tmpbuf, len, MSG_WAITALL);
-			if (err != (int)len) {
-				RTE_LOG(ERR, PMD,
-					"%s: error when receiving, %d of %zu\n",
-					__func__, err, count);
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-			err = nfp_cpp_area_write(area, pos, tmpbuf, len);
-			if (err < 0) {
-				RTE_LOG(ERR, PMD, "nfp_cpp_area_write error\n");
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-		}
-
-		nfp_offset += pos;
-		totlen += pos;
-		nfp_cpp_area_release(area);
-		nfp_cpp_area_free(area);
-
-		count -= pos;
-		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
-			 NFP_CPP_MEMIO_BOUNDARY : count;
-	}
-
-	return 0;
-}
-
-/*
- * Serving a read request to NFP from host programs. The request
- * sends the read size and the CPP target. The bridge makes use
- * of CPP interface handler configured by the PMD setup. The read
- * data is sent to the requester using the same socket.
- */
-static int
-nfp_cpp_bridge_serve_read(int sockfd, struct nfp_cpp *cpp)
-{
-	struct nfp_cpp_area *area;
-	off_t offset, nfp_offset;
-	uint32_t cpp_id, pos, len;
-	uint32_t tmpbuf[16];
-	size_t count, curlen, totlen = 0;
-	int err = 0;
-
-	PMD_CPP_LOG(DEBUG, "%s: offset size %zu, count_size: %zu\n", __func__,
-		sizeof(off_t), sizeof(size_t));
-
-	/* Reading the count param */
-	err = recv(sockfd, &count, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	curlen = count;
-
-	/* Reading the offset param */
-	err = recv(sockfd, &offset, sizeof(off_t), 0);
-	if (err != sizeof(off_t))
-		return -EINVAL;
-
-	/* Obtain target's CPP ID and offset in target */
-	cpp_id = (offset >> 40) << 8;
-	nfp_offset = offset & ((1ull << 40) - 1);
-
-	PMD_CPP_LOG(DEBUG, "%s: count %zu and offset %jd\n", __func__, count,
-			   offset);
-	PMD_CPP_LOG(DEBUG, "%s: cpp_id %08x and nfp_offset %jd\n", __func__,
-			   cpp_id, nfp_offset);
-
-	/* Adjust length if not aligned */
-	if (((nfp_offset + (off_t)count - 1) & ~(NFP_CPP_MEMIO_BOUNDARY - 1)) !=
-	    (nfp_offset & ~(NFP_CPP_MEMIO_BOUNDARY - 1))) {
-		curlen = NFP_CPP_MEMIO_BOUNDARY -
-			(nfp_offset & (NFP_CPP_MEMIO_BOUNDARY - 1));
-	}
-
-	while (count > 0) {
-		area = nfp_cpp_area_alloc_with_name(cpp, cpp_id, "nfp.cdev",
-						    nfp_offset, curlen);
-		if (!area) {
-			RTE_LOG(ERR, PMD, "%s: area alloc failed\n", __func__);
-			return -EIO;
-		}
-
-		err = nfp_cpp_area_acquire(area);
-		if (err < 0) {
-			RTE_LOG(ERR, PMD, "area acquire failed\n");
-			nfp_cpp_area_free(area);
-			return -EIO;
-		}
-
-		for (pos = 0; pos < curlen; pos += len) {
-			len = curlen - pos;
-			if (len > sizeof(tmpbuf))
-				len = sizeof(tmpbuf);
-
-			err = nfp_cpp_area_read(area, pos, tmpbuf, len);
-			if (err < 0) {
-				RTE_LOG(ERR, PMD, "nfp_cpp_area_read error\n");
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-			PMD_CPP_LOG(DEBUG, "%s: sending %u of %zu\n", __func__,
-					   len, count);
-
-			err = send(sockfd, tmpbuf, len, 0);
-			if (err != (int)len) {
-				RTE_LOG(ERR, PMD,
-					"%s: error when sending: %d of %zu\n",
-					__func__, err, count);
-				nfp_cpp_area_release(area);
-				nfp_cpp_area_free(area);
-				return -EIO;
-			}
-		}
-
-		nfp_offset += pos;
-		totlen += pos;
-		nfp_cpp_area_release(area);
-		nfp_cpp_area_free(area);
-
-		count -= pos;
-		curlen = (count > NFP_CPP_MEMIO_BOUNDARY) ?
-			NFP_CPP_MEMIO_BOUNDARY : count;
-	}
-	return 0;
-}
-
-#define NFP_IOCTL 'n'
-#define NFP_IOCTL_CPP_IDENTIFICATION _IOW(NFP_IOCTL, 0x8f, uint32_t)
-/*
- * Serving a ioctl command from host NFP tools. This usually goes to
- * a kernel driver char driver but it is not available when the PF is
- * bound to the PMD. Currently just one ioctl command is served and it
- * does not require any CPP access at all.
- */
-static int
-nfp_cpp_bridge_serve_ioctl(int sockfd, struct nfp_cpp *cpp)
-{
-	uint32_t cmd, ident_size, tmp;
-	int err;
-
-	/* Reading now the IOCTL command */
-	err = recv(sockfd, &cmd, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
-		return -EIO;
-	}
-
-	/* Only supporting NFP_IOCTL_CPP_IDENTIFICATION */
-	if (cmd != NFP_IOCTL_CPP_IDENTIFICATION) {
-		RTE_LOG(ERR, PMD, "%s: unknown cmd %d\n", __func__, cmd);
-		return -EINVAL;
-	}
-
-	err = recv(sockfd, &ident_size, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: read error from socket\n", __func__);
-		return -EIO;
-	}
-
-	tmp = nfp_cpp_model(cpp);
-
-	PMD_CPP_LOG(DEBUG, "%s: sending NFP model %08x\n", __func__, tmp);
-
-	err = send(sockfd, &tmp, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
-		return -EIO;
-	}
-
-	tmp = cpp->interface;
-
-	PMD_CPP_LOG(DEBUG, "%s: sending NFP interface %08x\n", __func__, tmp);
-
-	err = send(sockfd, &tmp, 4, 0);
-	if (err != 4) {
-		RTE_LOG(ERR, PMD, "%s: error writing to socket\n", __func__);
-		return -EIO;
-	}
-
-	return 0;
-}
-
-#define NFP_BRIDGE_OP_READ	20
-#define NFP_BRIDGE_OP_WRITE	30
-#define NFP_BRIDGE_OP_IOCTL	40
-
-/*
- * This is the code to be executed by a service core. The CPP bridge interface
- * is based on a unix socket and requests usually received by a kernel char
- * driver, read, write and ioctl, are handled by the CPP bridge. NFP host tools
- * can be executed with a wrapper library and LD_LIBRARY being completely
- * unaware of the CPP bridge performing the NFP kernel char driver for CPP
- * accesses.
- */
-static int32_t
-nfp_cpp_bridge_service_func(void *args)
-{
-	struct sockaddr address;
-	struct nfp_cpp *cpp = args;
-	int sockfd, datafd, op, ret;
-
-	unlink("/tmp/nfp_cpp");
-	sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
-	if (sockfd < 0) {
-		RTE_LOG(ERR, PMD, "%s: socket creation error. Service failed\n",
-			__func__);
-		return -EIO;
-	}
-
-	memset(&address, 0, sizeof(struct sockaddr));
-
-	address.sa_family = AF_UNIX;
-	strcpy(address.sa_data, "/tmp/nfp_cpp");
-
-	ret = bind(sockfd, (const struct sockaddr *)&address,
-		   sizeof(struct sockaddr));
-	if (ret < 0) {
-		RTE_LOG(ERR, PMD, "%s: bind error (%d). Service failed\n",
-				  __func__, errno);
-		close(sockfd);
-		return ret;
-	}
-
-	ret = listen(sockfd, 20);
-	if (ret < 0) {
-		RTE_LOG(ERR, PMD, "%s: listen error(%d). Service failed\n",
-				  __func__, errno);
-		close(sockfd);
-		return ret;
-	}
-
-	for (;;) {
-		datafd = accept(sockfd, NULL, NULL);
-		if (datafd < 0) {
-			RTE_LOG(ERR, PMD, "%s: accept call error (%d)\n",
-					  __func__, errno);
-			RTE_LOG(ERR, PMD, "%s: service failed\n", __func__);
-			close(sockfd);
-			return -EIO;
-		}
-
-		while (1) {
-			ret = recv(datafd, &op, 4, 0);
-			if (ret <= 0) {
-				PMD_CPP_LOG(DEBUG, "%s: socket close\n",
-						   __func__);
-				break;
-			}
-
-			PMD_CPP_LOG(DEBUG, "%s: getting op %u\n", __func__, op);
-
-			if (op == NFP_BRIDGE_OP_READ)
-				nfp_cpp_bridge_serve_read(datafd, cpp);
-
-			if (op == NFP_BRIDGE_OP_WRITE)
-				nfp_cpp_bridge_serve_write(datafd, cpp);
-
-			if (op == NFP_BRIDGE_OP_IOCTL)
-				nfp_cpp_bridge_serve_ioctl(datafd, cpp);
-
-			if (op == 0)
-				break;
-		}
-		close(datafd);
-	}
-	close(sockfd);
-
-	return 0;
-}
-
 #define DEFAULT_FW_PATH       "/lib/firmware/netronome"
 
 static int
@@ -2459,23 +2111,6 @@ static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
 	return ret;
 }
 
-static void nfp_register_cpp_service(struct nfp_cpp *cpp)
-{
-	uint32_t *cpp_service_id = NULL;
-	struct rte_service_spec service;
-
-	memset(&service, 0, sizeof(struct rte_service_spec));
-	snprintf(service.name, sizeof(service.name), "nfp_cpp_service");
-	service.callback = nfp_cpp_bridge_service_func;
-	service.callback_userdata = (void *)cpp;
-
-	if (rte_service_component_register(&service,
-					   cpp_service_id))
-		RTE_LOG(WARNING, PMD, "NFP CPP bridge service register() failed");
-	else
-		RTE_LOG(DEBUG, PMD, "NFP CPP bridge service registered");
-}
-
 static int nfp_pf_init(struct rte_pci_device *pci_dev)
 {
 	struct nfp_pf_dev *pf_dev = NULL;
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 4/7] net/nfp: prototype common functions in header file
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
                       ` (2 preceding siblings ...)
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
@ 2021-07-29 13:47     ` Heinrich Kuhn
  2021-08-17 14:44       ` Ferruh Yigit
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
                       ` (4 subsequent siblings)
  8 siblings, 1 reply; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

The majority of "ethdev" type functions are used for both PF devices and
VF devices. Prototype these functions in the nfp_net_pmd header file in
preparation of splitting PF and VF specific functions.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/nfp_net.c     | 87 +++++++++++++----------------------
 drivers/net/nfp/nfp_net_pmd.h | 49 ++++++++++++++++++++
 2 files changed, 81 insertions(+), 55 deletions(-)

diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index af94219729..d42d267c6a 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -55,35 +55,12 @@
 
 /* Prototypes */
 static int nfp_net_close(struct rte_eth_dev *dev);
-static int nfp_net_configure(struct rte_eth_dev *dev);
-static void nfp_net_dev_interrupt_handler(void *param);
-static void nfp_net_dev_interrupt_delayed_handler(void *param);
-static int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
-static int nfp_net_infos_get(struct rte_eth_dev *dev,
-			     struct rte_eth_dev_info *dev_info);
 static int nfp_net_init(struct rte_eth_dev *eth_dev);
 static int nfp_pf_init(struct rte_pci_device *pci_dev);
 static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
 static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
-static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
-static int nfp_net_promisc_enable(struct rte_eth_dev *dev);
-static int nfp_net_promisc_disable(struct rte_eth_dev *dev);
-static int nfp_net_start(struct rte_eth_dev *dev);
-static int nfp_net_stats_get(struct rte_eth_dev *dev,
-			      struct rte_eth_stats *stats);
-static int nfp_net_stats_reset(struct rte_eth_dev *dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
-static int nfp_net_rss_config_default(struct rte_eth_dev *dev);
-static int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
-				   struct rte_eth_rss_conf *rss_conf);
-static int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
-		    struct rte_eth_rss_reta_entry64 *reta_conf,
-		    uint16_t reta_size);
-static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
-			struct rte_eth_rss_conf *rss_conf);
-static int nfp_set_mac_addr(struct rte_eth_dev *dev,
-			     struct rte_ether_addr *mac_addr);
 static int nfp_fw_setup(struct rte_pci_device *dev,
 			struct nfp_cpp *cpp,
 			struct nfp_eth_table *nfp_eth_table,
@@ -138,7 +115,7 @@ __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
  * Write the update word to the BAR and ping the reconfig queue. Then poll
  * until the firmware has acknowledged the update by zeroing the update word.
  */
-static int
+int
 nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update)
 {
 	uint32_t err;
@@ -174,7 +151,7 @@ nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update)
  * before any other function in the Ethernet API. This function can
  * also be re-invoked when a device is in the stopped state.
  */
-static int
+int
 nfp_net_configure(struct rte_eth_dev *dev)
 {
 	struct rte_eth_conf *dev_conf;
@@ -217,7 +194,7 @@ nfp_net_configure(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static void
+void
 nfp_net_enable_queues(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -241,7 +218,7 @@ nfp_net_enable_queues(struct rte_eth_dev *dev)
 	nn_cfg_writeq(hw, NFP_NET_CFG_RXRS_ENABLE, enabled_queues);
 }
 
-static void
+void
 nfp_net_disable_queues(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -266,14 +243,14 @@ nfp_net_disable_queues(struct rte_eth_dev *dev)
 	hw->ctrl = new_ctrl;
 }
 
-static void
+void
 nfp_net_params_setup(struct nfp_net_hw *hw)
 {
 	nn_cfg_writel(hw, NFP_NET_CFG_MTU, hw->mtu);
 	nn_cfg_writel(hw, NFP_NET_CFG_FLBUFSZ, hw->flbufsz);
 }
 
-static void
+void
 nfp_net_cfg_queue_setup(struct nfp_net_hw *hw)
 {
 	hw->qcp_cfg = hw->tx_bar + NFP_QCP_QUEUE_ADDR_SZ;
@@ -281,7 +258,7 @@ nfp_net_cfg_queue_setup(struct nfp_net_hw *hw)
 
 #define ETH_ADDR_LEN	6
 
-static void
+void
 nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
 {
 	int i;
@@ -320,7 +297,7 @@ nfp_net_vf_read_mac(struct nfp_net_hw *hw)
 	memcpy(&hw->mac_addr[4], &tmp, 2);
 }
 
-static void
+void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
 	uint32_t mac0 = *(uint32_t *)mac;
@@ -368,7 +345,7 @@ nfp_set_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr)
 	return 0;
 }
 
-static int
+int
 nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
 			   struct rte_intr_handle *intr_handle)
 {
@@ -412,7 +389,7 @@ nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static uint32_t
+uint32_t
 nfp_check_offloads(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -748,7 +725,7 @@ nfp_net_close(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_promisc_enable(struct rte_eth_dev *dev)
 {
 	uint32_t new_ctrl, update = 0;
@@ -785,7 +762,7 @@ nfp_net_promisc_enable(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_promisc_disable(struct rte_eth_dev *dev)
 {
 	uint32_t new_ctrl, update = 0;
@@ -821,7 +798,7 @@ nfp_net_promisc_disable(struct rte_eth_dev *dev)
  * Wait to complete is needed as it can take up to 9 seconds to get the Link
  * status.
  */
-static int
+int
 nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
 {
 	struct nfp_net_hw *hw;
@@ -871,7 +848,7 @@ nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
 	return ret;
 }
 
-static int
+int
 nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 {
 	int i;
@@ -966,7 +943,7 @@ nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 	return -EINVAL;
 }
 
-static int
+int
 nfp_net_stats_reset(struct rte_eth_dev *dev)
 {
 	int i;
@@ -1031,7 +1008,7 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
 	return 0;
 }
 
-static int
+int
 nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 {
 	struct nfp_net_hw *hw;
@@ -1125,7 +1102,7 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 	return 0;
 }
 
-static const uint32_t *
+const uint32_t *
 nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 {
 	static const uint32_t ptypes[] = {
@@ -1142,7 +1119,7 @@ nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
 	return NULL;
 }
 
-static int
+int
 nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
 	struct rte_pci_device *pci_dev;
@@ -1162,7 +1139,7 @@ nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 	return 0;
 }
 
-static int
+int
 nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
 	struct rte_pci_device *pci_dev;
@@ -1181,7 +1158,7 @@ nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
 	return 0;
 }
 
-static void
+void
 nfp_net_dev_link_status_print(struct rte_eth_dev *dev)
 {
 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
@@ -1210,7 +1187,7 @@ nfp_net_dev_link_status_print(struct rte_eth_dev *dev)
  * If MSI-X auto-masking is enabled clear the mask bit, otherwise
  * clear the ICR for the entry.
  */
-static void
+void
 nfp_net_irq_unmask(struct rte_eth_dev *dev)
 {
 	struct nfp_net_hw *hw;
@@ -1231,7 +1208,7 @@ nfp_net_irq_unmask(struct rte_eth_dev *dev)
 	}
 }
 
-static void
+void
 nfp_net_dev_interrupt_handler(void *param)
 {
 	int64_t timeout;
@@ -1274,7 +1251,7 @@ nfp_net_dev_interrupt_handler(void *param)
  *
  * @return  void
  */
-static void
+void
 nfp_net_dev_interrupt_delayed_handler(void *param)
 {
 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
@@ -1288,7 +1265,7 @@ nfp_net_dev_interrupt_delayed_handler(void *param)
 	nfp_net_irq_unmask(dev);
 }
 
-static int
+int
 nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 {
 	struct nfp_net_hw *hw;
@@ -1323,7 +1300,7 @@ nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 	return 0;
 }
 
-static int
+int
 nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 {
 	uint32_t new_ctrl, update;
@@ -1355,7 +1332,7 @@ nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 	return ret;
 }
 
-static int
+int
 nfp_net_rss_reta_write(struct rte_eth_dev *dev,
 		    struct rte_eth_rss_reta_entry64 *reta_conf,
 		    uint16_t reta_size)
@@ -1406,7 +1383,7 @@ nfp_net_rss_reta_write(struct rte_eth_dev *dev,
 }
 
 /* Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device */
-static int
+int
 nfp_net_reta_update(struct rte_eth_dev *dev,
 		    struct rte_eth_rss_reta_entry64 *reta_conf,
 		    uint16_t reta_size)
@@ -1432,7 +1409,7 @@ nfp_net_reta_update(struct rte_eth_dev *dev,
 }
 
  /* Query Redirection Table(RETA) of Receive Side Scaling of Ethernet device. */
-static int
+int
 nfp_net_reta_query(struct rte_eth_dev *dev,
 		   struct rte_eth_rss_reta_entry64 *reta_conf,
 		   uint16_t reta_size)
@@ -1479,7 +1456,7 @@ nfp_net_reta_query(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf)
 {
@@ -1529,7 +1506,7 @@ nfp_net_rss_hash_write(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 			struct rte_eth_rss_conf *rss_conf)
 {
@@ -1565,7 +1542,7 @@ nfp_net_rss_hash_update(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
 			  struct rte_eth_rss_conf *rss_conf)
 {
@@ -1616,7 +1593,7 @@ nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
 	return 0;
 }
 
-static int
+int
 nfp_net_rss_config_default(struct rte_eth_dev *dev)
 {
 	struct rte_eth_conf *dev_conf;
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_net_pmd.h
index 9265496bf0..dc05e888df 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_net_pmd.h
@@ -349,6 +349,55 @@ nfp_qcp_read(uint8_t *q, enum nfp_qcp_ptr ptr)
 		return val & NFP_QCP_QUEUE_STS_HI_WRITEPTR_mask;
 }
 
+/* Prototypes for common NFP functions */
+int nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update);
+int nfp_net_configure(struct rte_eth_dev *dev);
+void nfp_net_enable_queues(struct rte_eth_dev *dev);
+void nfp_net_disable_queues(struct rte_eth_dev *dev);
+void nfp_net_params_setup(struct nfp_net_hw *hw);
+void nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src);
+void nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac);
+int nfp_set_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr);
+int nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
+			       struct rte_intr_handle *intr_handle);
+uint32_t nfp_check_offloads(struct rte_eth_dev *dev);
+int nfp_net_promisc_enable(struct rte_eth_dev *dev);
+int nfp_net_promisc_disable(struct rte_eth_dev *dev);
+int nfp_net_link_update(struct rte_eth_dev *dev,
+			__rte_unused int wait_to_complete);
+int nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats);
+int nfp_net_stats_reset(struct rte_eth_dev *dev);
+int nfp_net_infos_get(struct rte_eth_dev *dev,
+		      struct rte_eth_dev_info *dev_info);
+const uint32_t *nfp_net_supported_ptypes_get(struct rte_eth_dev *dev);
+int nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
+int nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
+void nfp_net_params_setup(struct nfp_net_hw *hw);
+void nfp_net_cfg_queue_setup(struct nfp_net_hw *hw);
+void nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src);
+void nfp_net_dev_link_status_print(struct rte_eth_dev *dev);
+void nfp_net_irq_unmask(struct rte_eth_dev *dev);
+void nfp_net_dev_interrupt_handler(void *param);
+void nfp_net_dev_interrupt_delayed_handler(void *param);
+int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
+int nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask);
+int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
+			   struct rte_eth_rss_reta_entry64 *reta_conf,
+			   uint16_t reta_size);
+int nfp_net_reta_update(struct rte_eth_dev *dev,
+			struct rte_eth_rss_reta_entry64 *reta_conf,
+			uint16_t reta_size);
+int nfp_net_reta_query(struct rte_eth_dev *dev,
+		       struct rte_eth_rss_reta_entry64 *reta_conf,
+		       uint16_t reta_size);
+int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
+			   struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
+			    struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_hash_conf_get(struct rte_eth_dev *dev,
+			      struct rte_eth_rss_conf *rss_conf);
+int nfp_net_rss_config_default(struct rte_eth_dev *dev);
+
 #define NFP_NET_DEV_PRIVATE_TO_HW(adapter)\
 	(&((struct nfp_net_adapter *)adapter)->hw)
 
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 5/7] net/nfp: move VF functions into new file
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
                       ` (3 preceding siblings ...)
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
@ 2021-07-29 13:47     ` Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 6/7] net/nfp: move PF " Heinrich Kuhn
                       ` (3 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Move any ethdev functionality specific to VF devices into a new file
called nfp_ethdev_vf.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build     |   1 +
 drivers/net/nfp/nfp_ethdev_vf.c | 504 ++++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_net.c       |  42 +--
 3 files changed, 506 insertions(+), 41 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_ethdev_vf.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index b46ac2d40f..34f4054b3c 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -21,4 +21,5 @@ sources = files(
         'nfp_net.c',
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
+        'nfp_ethdev_vf.c',
 )
diff --git a/drivers/net/nfp/nfp_ethdev_vf.c b/drivers/net/nfp/nfp_ethdev_vf.c
new file mode 100644
index 0000000000..223142c0ed
--- /dev/null
+++ b/drivers/net/nfp/nfp_ethdev_vf.c
@@ -0,0 +1,504 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_ethdev_vf.c
+ *
+ * Netronome vNIC  VF DPDK Poll-Mode Driver: Main entry point
+ */
+
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_rtsym.h"
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+
+static void nfp_netvf_read_mac(struct nfp_net_hw *hw);
+static int nfp_netvf_start(struct rte_eth_dev *dev);
+static int nfp_netvf_stop(struct rte_eth_dev *dev);
+static int nfp_netvf_set_link_up(struct rte_eth_dev *dev);
+static int nfp_netvf_set_link_down(struct rte_eth_dev *dev);
+static int nfp_netvf_close(struct rte_eth_dev *dev);
+static int nfp_netvf_init(struct rte_eth_dev *eth_dev);
+static int nfp_vf_pci_uninit(struct rte_eth_dev *eth_dev);
+static int eth_nfp_vf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+	struct rte_pci_device *pci_dev);
+static int eth_nfp_vf_pci_remove(struct rte_pci_device *pci_dev);
+
+static void
+nfp_netvf_read_mac(struct nfp_net_hw *hw)
+{
+	uint32_t tmp;
+
+	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR));
+	memcpy(&hw->mac_addr[0], &tmp, 4);
+
+	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR + 4));
+	memcpy(&hw->mac_addr[4], &tmp, 2);
+}
+
+static int
+nfp_netvf_start(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+	uint32_t new_ctrl, update = 0;
+	struct nfp_net_hw *hw;
+	struct rte_eth_conf *dev_conf;
+	struct rte_eth_rxmode *rxmode;
+	uint32_t intr_vector;
+	int ret;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	PMD_INIT_LOG(DEBUG, "Start");
+
+	/* Disabling queues just in case... */
+	nfp_net_disable_queues(dev);
+
+	/* Enabling the required queues in the device */
+	nfp_net_enable_queues(dev);
+
+	/* check and configure queue intr-vector mapping */
+	if (dev->data->dev_conf.intr_conf.rxq != 0) {
+		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+			/*
+			 * Better not to share LSC with RX interrupts.
+			 * Unregistering LSC interrupt handler
+			 */
+			rte_intr_callback_unregister(&pci_dev->intr_handle,
+				nfp_net_dev_interrupt_handler, (void *)dev);
+
+			if (dev->data->nb_rx_queues > 1) {
+				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+					     "supports 1 queue with UIO");
+				return -EIO;
+			}
+		}
+		intr_vector = dev->data->nb_rx_queues;
+		if (rte_intr_efd_enable(intr_handle, intr_vector))
+			return -1;
+
+		nfp_configure_rx_interrupt(dev, intr_handle);
+		update = NFP_NET_CFG_UPDATE_MSIX;
+	}
+
+	rte_intr_enable(intr_handle);
+
+	new_ctrl = nfp_check_offloads(dev);
+
+	/* Writing configuration parameters in the device */
+	nfp_net_params_setup(hw);
+
+	dev_conf = &dev->data->dev_conf;
+	rxmode = &dev_conf->rxmode;
+
+	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
+		nfp_net_rss_config_default(dev);
+		update |= NFP_NET_CFG_UPDATE_RSS;
+		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
+	}
+
+	/* Enable device */
+	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
+
+	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+
+	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
+		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
+
+	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
+	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+		return -EIO;
+
+	/*
+	 * Allocating rte mbufs for configured rx queues.
+	 * This requires queues being enabled before
+	 */
+	if (nfp_net_rx_freelist_setup(dev) < 0) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	hw->ctrl = new_ctrl;
+
+	return 0;
+
+error:
+	/*
+	 * An error returned by this function should mean the app
+	 * exiting and then the system releasing all the memory
+	 * allocated even memory coming from hugepages.
+	 *
+	 * The device could be enabled at this point with some queues
+	 * ready for getting packets. This is true if the call to
+	 * nfp_net_rx_freelist_setup() succeeds for some queues but
+	 * fails for subsequent queues.
+	 *
+	 * This should make the app exiting but better if we tell the
+	 * device first.
+	 */
+	nfp_net_disable_queues(dev);
+
+	return ret;
+}
+
+static int
+nfp_netvf_stop(struct rte_eth_dev *dev)
+{
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	PMD_INIT_LOG(DEBUG, "Stop");
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	return 0;
+}
+
+static int
+nfp_netvf_set_link_up(struct rte_eth_dev *dev __rte_unused)
+{
+	return -ENOTSUP;
+}
+
+/* Set the link down. */
+static int
+nfp_netvf_set_link_down(struct rte_eth_dev *dev __rte_unused)
+{
+	return -ENOTSUP;
+}
+
+/* Reset and stop device. The device can not be restarted. */
+static int
+nfp_netvf_close(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	PMD_INIT_LOG(DEBUG, "Close");
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+	/*
+	 * We assume that the DPDK application is stopping all the
+	 * threads/queues before calling the device close function.
+	 */
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q =  (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q =  (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	rte_intr_disable(&pci_dev->intr_handle);
+
+	/* unregister callback func from eal lib */
+	rte_intr_callback_unregister(&pci_dev->intr_handle,
+				     nfp_net_dev_interrupt_handler,
+				     (void *)dev);
+
+	/*
+	 * The ixgbe PMD driver disables the pcie master on the
+	 * device. The i40e does not...
+	 */
+
+	return 0;
+}
+
+/* Initialise and register VF driver with DPDK Application */
+static const struct eth_dev_ops nfp_netvf_eth_dev_ops = {
+	.dev_configure		= nfp_net_configure,
+	.dev_start		= nfp_netvf_start,
+	.dev_stop		= nfp_netvf_stop,
+	.dev_set_link_up	= nfp_netvf_set_link_up,
+	.dev_set_link_down	= nfp_netvf_set_link_down,
+	.dev_close		= nfp_netvf_close,
+	.promiscuous_enable	= nfp_net_promisc_enable,
+	.promiscuous_disable	= nfp_net_promisc_disable,
+	.link_update		= nfp_net_link_update,
+	.stats_get		= nfp_net_stats_get,
+	.stats_reset		= nfp_net_stats_reset,
+	.dev_infos_get		= nfp_net_infos_get,
+	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
+	.mtu_set		= nfp_net_dev_mtu_set,
+	.mac_addr_set           = nfp_set_mac_addr,
+	.vlan_offload_set	= nfp_net_vlan_offload_set,
+	.reta_update		= nfp_net_reta_update,
+	.reta_query		= nfp_net_reta_query,
+	.rss_hash_update	= nfp_net_rss_hash_update,
+	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
+	.rx_queue_setup		= nfp_net_rx_queue_setup,
+	.rx_queue_release	= nfp_net_rx_queue_release,
+	.tx_queue_setup		= nfp_net_tx_queue_setup,
+	.tx_queue_release	= nfp_net_tx_queue_release,
+	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
+	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
+};
+
+static int
+nfp_netvf_init(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_net_hw *hw;
+	struct rte_ether_addr *tmp_ether_addr;
+
+	uint64_t tx_bar_off = 0, rx_bar_off = 0;
+	uint32_t start_q;
+	int stride = 4;
+	int port = 0;
+	int err;
+
+	PMD_INIT_FUNC_TRACE();
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* NFP can not handle DMA addresses requiring more than 40 bits */
+	if (rte_mem_check_dma_mask(40)) {
+		RTE_LOG(ERR, PMD, "device %s can not be used:",
+				   pci_dev->device.name);
+		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
+		return -ENODEV;
+	};
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+
+	eth_dev->dev_ops = &nfp_netvf_eth_dev_ops;
+	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	rte_eth_copy_pci_info(eth_dev, pci_dev);
+
+	hw->device_id = pci_dev->id.device_id;
+	hw->vendor_id = pci_dev->id.vendor_id;
+	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
+	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
+
+	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
+		     pci_dev->id.vendor_id, pci_dev->id.device_id,
+		     pci_dev->addr.domain, pci_dev->addr.bus,
+		     pci_dev->addr.devid, pci_dev->addr.function);
+
+	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
+	if (hw->ctrl_bar == NULL) {
+		PMD_DRV_LOG(ERR,
+			"hw->ctrl_bar is NULL. BAR0 not configured");
+		return -ENODEV;
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+
+	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
+	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
+
+	/* Work out where in the BAR the queues start. */
+	switch (pci_dev->id.device_id) {
+	case PCI_DEVICE_ID_NFP6000_VF_NIC:
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
+		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
+		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
+		err = -ENODEV;
+		goto dev_err_ctrl_map;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
+	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
+
+	hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+		     tx_bar_off;
+	hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+		     rx_bar_off;
+
+	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
+		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
+
+	nfp_net_cfg_queue_setup(hw);
+
+	/* Get some of the read-only fields from the config BAR */
+	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
+	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
+	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
+	hw->mtu = RTE_ETHER_MTU;
+
+	/* VLAN insertion is incompatible with LSOv2 */
+	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
+		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
+
+	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
+		hw->rx_offset = NFP_NET_RX_OFFSET;
+	else
+		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
+
+	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
+			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
+			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
+
+	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
+		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
+
+	hw->ctrl = 0;
+
+	hw->stride_rx = stride;
+	hw->stride_tx = stride;
+
+	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
+		     hw->max_rx_queues, hw->max_tx_queues);
+
+	/* Initializing spinlock for reconfigs */
+	rte_spinlock_init(&hw->reconfig_lock);
+
+	/* Allocating memory for mac addr */
+	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
+					       RTE_ETHER_ADDR_LEN, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
+		err = -ENOMEM;
+		goto dev_err_queues_map;
+	}
+
+	nfp_netvf_read_mac(hw);
+
+	tmp_ether_addr = (struct rte_ether_addr *)&hw->mac_addr;
+	if (!rte_is_valid_assigned_ether_addr(tmp_ether_addr)) {
+		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
+				   port);
+		/* Using random mac addresses for VFs */
+		rte_eth_random_addr(&hw->mac_addr[0]);
+		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+	}
+
+	/* Copying mac address to DPDK eth_dev struct */
+	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
+			&eth_dev->data->mac_addrs[0]);
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
+
+	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
+
+	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
+		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
+		     eth_dev->data->port_id, pci_dev->id.vendor_id,
+		     pci_dev->id.device_id,
+		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
+		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		/* Registering LSC interrupt handler */
+		rte_intr_callback_register(&pci_dev->intr_handle,
+					   nfp_net_dev_interrupt_handler,
+					   (void *)eth_dev);
+		/* Telling the firmware about the LSC interrupt entry */
+		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
+		/* Recording current stats counters values */
+		nfp_net_stats_reset(eth_dev);
+	}
+
+	return 0;
+
+dev_err_queues_map:
+		nfp_cpp_area_free(hw->hwqueues_area);
+dev_err_ctrl_map:
+		nfp_cpp_area_free(hw->ctrl_area);
+
+	return err;
+}
+
+static const struct rte_pci_id pci_id_nfp_vf_net_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP6000_VF_NIC)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static int nfp_vf_pci_uninit(struct rte_eth_dev *eth_dev)
+{
+	/* VF cleanup, just free private port data */
+	return nfp_netvf_close(eth_dev);
+}
+
+static int eth_nfp_vf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+	struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_probe(pci_dev,
+		sizeof(struct nfp_net_adapter), nfp_netvf_init);
+}
+
+static int eth_nfp_vf_pci_remove(struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_vf_pci_uninit);
+}
+
+static struct rte_pci_driver rte_nfp_net_vf_pmd = {
+	.id_table = pci_id_nfp_vf_net_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	.probe = eth_nfp_vf_pci_probe,
+	.remove = eth_nfp_vf_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_nfp_vf, rte_nfp_net_vf_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_vf, pci_id_nfp_vf_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_vf, "* igb_uio | uio_pci_generic | vfio");
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index d42d267c6a..3c71b513ac 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -58,6 +58,7 @@ static int nfp_net_close(struct rte_eth_dev *dev);
 static int nfp_net_init(struct rte_eth_dev *eth_dev);
 static int nfp_pf_init(struct rte_pci_device *pci_dev);
 static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
+static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
 static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
 static int nfp_net_stop(struct rte_eth_dev *dev);
@@ -285,18 +286,6 @@ nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
 	return 0;
 }
 
-static void
-nfp_net_vf_read_mac(struct nfp_net_hw *hw)
-{
-	uint32_t tmp;
-
-	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR));
-	memcpy(&hw->mac_addr[0], &tmp, 4);
-
-	tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR + 4));
-	memcpy(&hw->mac_addr[4], &tmp, 2);
-}
-
 void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
@@ -1854,8 +1843,6 @@ nfp_net_init(struct rte_eth_dev *eth_dev)
 	if (hw->is_phyport) {
 		nfp_net_pf_read_mac(pf_dev, port);
 		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	} else {
-		nfp_net_vf_read_mac(hw);
 	}
 
 	if (!rte_is_valid_assigned_ether_addr(
@@ -2337,16 +2324,6 @@ static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
 	},
 };
 
-static const struct rte_pci_id pci_id_nfp_vf_net_map[] = {
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP6000_VF_NIC)
-	},
-	{
-		.vendor_id = 0,
-	},
-};
-
 static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
 {
 	struct rte_pci_device *pci_dev;
@@ -2370,13 +2347,6 @@ static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
 	return nfp_net_close(eth_dev);
 }
 
-static int eth_nfp_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
-	struct rte_pci_device *pci_dev)
-{
-	return rte_eth_dev_pci_generic_probe(pci_dev,
-		sizeof(struct nfp_net_adapter), nfp_net_init);
-}
-
 static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
 {
 	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
@@ -2389,19 +2359,9 @@ static struct rte_pci_driver rte_nfp_net_pf_pmd = {
 	.remove = eth_nfp_pci_remove,
 };
 
-static struct rte_pci_driver rte_nfp_net_vf_pmd = {
-	.id_table = pci_id_nfp_vf_net_map,
-	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
-	.probe = eth_nfp_pci_probe,
-	.remove = eth_nfp_pci_remove,
-};
-
 RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
-RTE_PMD_REGISTER_PCI(net_nfp_vf, rte_nfp_net_vf_pmd);
 RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
-RTE_PMD_REGISTER_PCI_TABLE(net_nfp_vf, pci_id_nfp_vf_net_map);
 RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
-RTE_PMD_REGISTER_KMOD_DEP(net_nfp_vf, "* igb_uio | uio_pci_generic | vfio");
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_init, init, NOTICE);
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_driver, driver, NOTICE);
 /*
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 6/7] net/nfp: move PF functions into new file
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
                       ` (4 preceding siblings ...)
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
@ 2021-07-29 13:47     ` Heinrich Kuhn
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
                       ` (2 subsequent siblings)
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Similar to the last commit, this changeset moves all the PF specific
functions to a new file called nfp_ethdev.c.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build  |    1 +
 drivers/net/nfp/nfp_ethdev.c | 1067 ++++++++++++++++++++++++++++++++++
 drivers/net/nfp/nfp_net.c    | 1056 +--------------------------------
 3 files changed, 1071 insertions(+), 1053 deletions(-)
 create mode 100644 drivers/net/nfp/nfp_ethdev.c

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index 34f4054b3c..ab64d0cac3 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -22,4 +22,5 @@ sources = files(
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
         'nfp_ethdev_vf.c',
+        'nfp_ethdev.c',
 )
diff --git a/drivers/net/nfp/nfp_ethdev.c b/drivers/net/nfp/nfp_ethdev.c
new file mode 100644
index 0000000000..056d962306
--- /dev/null
+++ b/drivers/net/nfp/nfp_ethdev.c
@@ -0,0 +1,1067 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright (c) 2014-2021 Netronome Systems, Inc.
+ * All rights reserved.
+ *
+ * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation.
+ */
+
+/*
+ * vim:shiftwidth=8:noexpandtab
+ *
+ * @file dpdk/pmd/nfp_ethdev.c
+ *
+ * Netronome vNIC DPDK Poll-Mode Driver: Main entry point
+ */
+
+#include <rte_common.h>
+#include <ethdev_driver.h>
+#include <ethdev_pci.h>
+#include <rte_dev.h>
+#include <rte_ether.h>
+#include <rte_malloc.h>
+#include <rte_memzone.h>
+#include <rte_mempool.h>
+#include <rte_service_component.h>
+#include "eal_firmware.h"
+
+#include "nfpcore/nfp_cpp.h"
+#include "nfpcore/nfp_nffw.h"
+#include "nfpcore/nfp_hwinfo.h"
+#include "nfpcore/nfp_mip.h"
+#include "nfpcore/nfp_rtsym.h"
+#include "nfpcore/nfp_nsp.h"
+
+#include "nfp_net_pmd.h"
+#include "nfp_rxtx.h"
+#include "nfp_net_logs.h"
+#include "nfp_net_ctrl.h"
+#include "nfp_cpp_bridge.h"
+
+
+static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
+static int nfp_net_start(struct rte_eth_dev *dev);
+static int nfp_net_stop(struct rte_eth_dev *dev);
+static int nfp_net_set_link_up(struct rte_eth_dev *dev);
+static int nfp_net_set_link_down(struct rte_eth_dev *dev);
+static int nfp_net_close(struct rte_eth_dev *dev);
+static int nfp_net_init(struct rte_eth_dev *eth_dev);
+static int nfp_fw_upload(struct rte_pci_device *dev,
+			 struct nfp_nsp *nsp, char *card);
+static int nfp_fw_setup(struct rte_pci_device *dev,
+			struct nfp_cpp *cpp,
+			struct nfp_eth_table *nfp_eth_table,
+			struct nfp_hwinfo *hwinfo);
+static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
+static int nfp_pf_init(struct rte_pci_device *pci_dev);
+static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
+static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+			    struct rte_pci_device *dev);
+static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
+static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev);
+
+static int
+nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
+{
+	struct nfp_eth_table *nfp_eth_table;
+	struct nfp_net_hw *hw = NULL;
+
+	/* Grab a pointer to the correct physical port */
+	hw = pf_dev->ports[port];
+
+	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
+
+	nfp_eth_copy_mac((uint8_t *)&hw->mac_addr,
+			 (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
+
+	free(nfp_eth_table);
+	return 0;
+}
+
+static int
+nfp_net_start(struct rte_eth_dev *dev)
+{
+	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+	uint32_t new_ctrl, update = 0;
+	struct nfp_net_hw *hw;
+	struct nfp_pf_dev *pf_dev;
+	struct rte_eth_conf *dev_conf;
+	struct rte_eth_rxmode *rxmode;
+	uint32_t intr_vector;
+	int ret;
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+
+	PMD_INIT_LOG(DEBUG, "Start");
+
+	/* Disabling queues just in case... */
+	nfp_net_disable_queues(dev);
+
+	/* Enabling the required queues in the device */
+	nfp_net_enable_queues(dev);
+
+	/* check and configure queue intr-vector mapping */
+	if (dev->data->dev_conf.intr_conf.rxq != 0) {
+		if (pf_dev->multiport) {
+			PMD_INIT_LOG(ERR, "PMD rx interrupt is not supported "
+					  "with NFP multiport PF");
+				return -EINVAL;
+		}
+		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+			/*
+			 * Better not to share LSC with RX interrupts.
+			 * Unregistering LSC interrupt handler
+			 */
+			rte_intr_callback_unregister(&pci_dev->intr_handle,
+				nfp_net_dev_interrupt_handler, (void *)dev);
+
+			if (dev->data->nb_rx_queues > 1) {
+				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+					     "supports 1 queue with UIO");
+				return -EIO;
+			}
+		}
+		intr_vector = dev->data->nb_rx_queues;
+		if (rte_intr_efd_enable(intr_handle, intr_vector))
+			return -1;
+
+		nfp_configure_rx_interrupt(dev, intr_handle);
+		update = NFP_NET_CFG_UPDATE_MSIX;
+	}
+
+	rte_intr_enable(intr_handle);
+
+	new_ctrl = nfp_check_offloads(dev);
+
+	/* Writing configuration parameters in the device */
+	nfp_net_params_setup(hw);
+
+	dev_conf = &dev->data->dev_conf;
+	rxmode = &dev_conf->rxmode;
+
+	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
+		nfp_net_rss_config_default(dev);
+		update |= NFP_NET_CFG_UPDATE_RSS;
+		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
+	}
+
+	/* Enable device */
+	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
+
+	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+
+	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
+		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
+
+	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
+	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+		return -EIO;
+
+	/*
+	 * Allocating rte mbufs for configured rx queues.
+	 * This requires queues being enabled before
+	 */
+	if (nfp_net_rx_freelist_setup(dev) < 0) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port up */
+		nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
+	else
+		nfp_eth_set_configured(dev->process_private,
+				       hw->nfp_idx, 1);
+
+	hw->ctrl = new_ctrl;
+
+	return 0;
+
+error:
+	/*
+	 * An error returned by this function should mean the app
+	 * exiting and then the system releasing all the memory
+	 * allocated even memory coming from hugepages.
+	 *
+	 * The device could be enabled at this point with some queues
+	 * ready for getting packets. This is true if the call to
+	 * nfp_net_rx_freelist_setup() succeeds for some queues but
+	 * fails for subsequent queues.
+	 *
+	 * This should make the app exiting but better if we tell the
+	 * device first.
+	 */
+	nfp_net_disable_queues(dev);
+
+	return ret;
+}
+
+/* Stop device: disable rx and tx functions to allow for reconfiguring. */
+static int
+nfp_net_stop(struct rte_eth_dev *dev)
+{
+	int i;
+	struct nfp_net_hw *hw;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+
+	PMD_INIT_LOG(DEBUG, "Stop");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
+	else
+		nfp_eth_set_configured(dev->process_private,
+				       hw->nfp_idx, 0);
+
+	return 0;
+}
+
+/* Set the link up. */
+static int
+nfp_net_set_link_up(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+
+	PMD_DRV_LOG(DEBUG, "Set link up");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
+	else
+		return nfp_eth_set_configured(dev->process_private,
+					      hw->nfp_idx, 1);
+}
+
+/* Set the link down. */
+static int
+nfp_net_set_link_down(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+
+	PMD_DRV_LOG(DEBUG, "Set link down");
+
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		/* Configure the physical port down */
+		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
+	else
+		return nfp_eth_set_configured(dev->process_private,
+					      hw->nfp_idx, 0);
+}
+
+/* Reset and stop device. The device can not be restarted. */
+static int
+nfp_net_close(struct rte_eth_dev *dev)
+{
+	struct nfp_net_hw *hw;
+	struct rte_pci_device *pci_dev;
+	struct nfp_pf_dev *pf_dev;
+	struct nfp_net_txq *this_tx_q;
+	struct nfp_net_rxq *this_rx_q;
+	int i;
+
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	PMD_INIT_LOG(DEBUG, "Close");
+
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
+	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+	/*
+	 * We assume that the DPDK application is stopping all the
+	 * threads/queues before calling the device close function.
+	 */
+
+	nfp_net_disable_queues(dev);
+
+	/* Clear queues */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		this_tx_q = (struct nfp_net_txq *)dev->data->tx_queues[i];
+		nfp_net_reset_tx_queue(this_tx_q);
+	}
+
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		this_rx_q = (struct nfp_net_rxq *)dev->data->rx_queues[i];
+		nfp_net_reset_rx_queue(this_rx_q);
+	}
+
+	/* Only free PF resources after all physical ports have been closed */
+	/* Mark this port as unused and free device priv resources*/
+	nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
+	pf_dev->ports[hw->idx] = NULL;
+	rte_eth_dev_release_port(dev);
+
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		/* Check to see if ports are still in use */
+		if (pf_dev->ports[i])
+			return 0;
+	}
+
+	/* Now it is safe to free all PF resources */
+	PMD_INIT_LOG(INFO, "Freeing PF resources");
+	nfp_cpp_area_free(pf_dev->ctrl_area);
+	nfp_cpp_area_free(pf_dev->hwqueues_area);
+	free(pf_dev->hwinfo);
+	free(pf_dev->sym_tbl);
+	nfp_cpp_free(pf_dev->cpp);
+	rte_free(pf_dev);
+
+	rte_intr_disable(&pci_dev->intr_handle);
+
+	/* unregister callback func from eal lib */
+	rte_intr_callback_unregister(&pci_dev->intr_handle,
+				     nfp_net_dev_interrupt_handler,
+				     (void *)dev);
+
+	/*
+	 * The ixgbe PMD driver disables the pcie master on the
+	 * device. The i40e does not...
+	 */
+
+	return 0;
+}
+
+/* Initialise and register driver with DPDK Application */
+static const struct eth_dev_ops nfp_net_eth_dev_ops = {
+	.dev_configure		= nfp_net_configure,
+	.dev_start		= nfp_net_start,
+	.dev_stop		= nfp_net_stop,
+	.dev_set_link_up	= nfp_net_set_link_up,
+	.dev_set_link_down	= nfp_net_set_link_down,
+	.dev_close		= nfp_net_close,
+	.promiscuous_enable	= nfp_net_promisc_enable,
+	.promiscuous_disable	= nfp_net_promisc_disable,
+	.link_update		= nfp_net_link_update,
+	.stats_get		= nfp_net_stats_get,
+	.stats_reset		= nfp_net_stats_reset,
+	.dev_infos_get		= nfp_net_infos_get,
+	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
+	.mtu_set		= nfp_net_dev_mtu_set,
+	.mac_addr_set           = nfp_set_mac_addr,
+	.vlan_offload_set	= nfp_net_vlan_offload_set,
+	.reta_update		= nfp_net_reta_update,
+	.reta_query		= nfp_net_reta_query,
+	.rss_hash_update	= nfp_net_rss_hash_update,
+	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
+	.rx_queue_setup		= nfp_net_rx_queue_setup,
+	.rx_queue_release	= nfp_net_rx_queue_release,
+	.tx_queue_setup		= nfp_net_tx_queue_setup,
+	.tx_queue_release	= nfp_net_tx_queue_release,
+	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
+	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
+};
+
+static int
+nfp_net_init(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	struct nfp_pf_dev *pf_dev;
+	struct nfp_net_hw *hw;
+	struct rte_ether_addr *tmp_ether_addr;
+
+	uint64_t tx_bar_off = 0, rx_bar_off = 0;
+	uint32_t start_q;
+	int stride = 4;
+	int port = 0;
+	int err;
+
+	PMD_INIT_FUNC_TRACE();
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* Use backpointer here to the PF of this eth_dev */
+	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(eth_dev->data->dev_private);
+
+	/* NFP can not handle DMA addresses requiring more than 40 bits */
+	if (rte_mem_check_dma_mask(40)) {
+		RTE_LOG(ERR, PMD, "device %s can not be used:",
+				   pci_dev->device.name);
+		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
+		return -ENODEV;
+	};
+
+	port = ((struct nfp_net_hw *)eth_dev->data->dev_private)->idx;
+	if (port < 0 || port > 7) {
+		PMD_DRV_LOG(ERR, "Port value is wrong");
+		return -ENODEV;
+	}
+
+	/* Use PF array of physical ports to get pointer to
+	 * this specific port
+	 */
+	hw = pf_dev->ports[port];
+
+	PMD_INIT_LOG(DEBUG, "Working with physical port number: %d, "
+			    "NFP internal port number: %d",
+			    port, hw->nfp_idx);
+
+	eth_dev->dev_ops = &nfp_net_eth_dev_ops;
+	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+
+	/* For secondary processes, the primary has done all the work */
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
+
+	rte_eth_copy_pci_info(eth_dev, pci_dev);
+
+	hw->device_id = pci_dev->id.device_id;
+	hw->vendor_id = pci_dev->id.vendor_id;
+	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
+	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
+
+	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
+		     pci_dev->id.vendor_id, pci_dev->id.device_id,
+		     pci_dev->addr.domain, pci_dev->addr.bus,
+		     pci_dev->addr.devid, pci_dev->addr.function);
+
+	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
+	if (hw->ctrl_bar == NULL) {
+		PMD_DRV_LOG(ERR,
+			"hw->ctrl_bar is NULL. BAR0 not configured");
+		return -ENODEV;
+	}
+
+	if (port == 0) {
+		hw->ctrl_bar = pf_dev->ctrl_bar;
+	} else {
+		if (!pf_dev->ctrl_bar)
+			return -ENODEV;
+		/* Use port offset in pf ctrl_bar for this
+		 * ports control bar
+		 */
+		hw->ctrl_bar = pf_dev->ctrl_bar +
+			       (port * NFP_PF_CSR_SLICE_SIZE);
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+
+	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
+	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
+
+	/* Work out where in the BAR the queues start. */
+	switch (pci_dev->id.device_id) {
+	case PCI_DEVICE_ID_NFP4000_PF_NIC:
+	case PCI_DEVICE_ID_NFP6000_PF_NIC:
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
+		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
+		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
+		err = -ENODEV;
+		goto dev_err_ctrl_map;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
+	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
+
+	hw->tx_bar = pf_dev->hw_queues + tx_bar_off;
+	hw->rx_bar = pf_dev->hw_queues + rx_bar_off;
+	eth_dev->data->dev_private = hw;
+
+	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
+		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
+
+	nfp_net_cfg_queue_setup(hw);
+
+	/* Get some of the read-only fields from the config BAR */
+	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
+	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
+	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
+	hw->mtu = RTE_ETHER_MTU;
+
+	/* VLAN insertion is incompatible with LSOv2 */
+	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
+		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
+
+	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
+		hw->rx_offset = NFP_NET_RX_OFFSET;
+	else
+		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
+
+	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
+			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
+			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
+
+	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
+		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
+		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
+		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
+
+	hw->ctrl = 0;
+
+	hw->stride_rx = stride;
+	hw->stride_tx = stride;
+
+	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
+		     hw->max_rx_queues, hw->max_tx_queues);
+
+	/* Initializing spinlock for reconfigs */
+	rte_spinlock_init(&hw->reconfig_lock);
+
+	/* Allocating memory for mac addr */
+	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
+					       RTE_ETHER_ADDR_LEN, 0);
+	if (eth_dev->data->mac_addrs == NULL) {
+		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
+		err = -ENOMEM;
+		goto dev_err_queues_map;
+	}
+
+	nfp_net_pf_read_mac(pf_dev, port);
+	nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+
+	tmp_ether_addr = (struct rte_ether_addr *)&hw->mac_addr;
+	if (!rte_is_valid_assigned_ether_addr(tmp_ether_addr)) {
+		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
+				   port);
+		/* Using random mac addresses for VFs */
+		rte_eth_random_addr(&hw->mac_addr[0]);
+		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+	}
+
+	/* Copying mac address to DPDK eth_dev struct */
+	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
+			&eth_dev->data->mac_addrs[0]);
+
+	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
+		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
+
+	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
+
+	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
+		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
+		     eth_dev->data->port_id, pci_dev->id.vendor_id,
+		     pci_dev->id.device_id,
+		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
+		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
+
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
+		/* Registering LSC interrupt handler */
+		rte_intr_callback_register(&pci_dev->intr_handle,
+					   nfp_net_dev_interrupt_handler,
+					   (void *)eth_dev);
+		/* Telling the firmware about the LSC interrupt entry */
+		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
+		/* Recording current stats counters values */
+		nfp_net_stats_reset(eth_dev);
+	}
+
+	return 0;
+
+dev_err_queues_map:
+		nfp_cpp_area_free(hw->hwqueues_area);
+dev_err_ctrl_map:
+		nfp_cpp_area_free(hw->ctrl_area);
+
+	return err;
+}
+
+#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
+
+static int
+nfp_fw_upload(struct rte_pci_device *dev, struct nfp_nsp *nsp, char *card)
+{
+	struct nfp_cpp *cpp = nsp->cpp;
+	void *fw_buf;
+	char fw_name[125];
+	char serial[40];
+	size_t fsize;
+
+	/* Looking for firmware file in order of priority */
+
+	/* First try to find a firmware image specific for this device */
+	snprintf(serial, sizeof(serial),
+			"serial-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x",
+		cpp->serial[0], cpp->serial[1], cpp->serial[2], cpp->serial[3],
+		cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
+		cpp->interface & 0xff);
+
+	snprintf(fw_name, sizeof(fw_name), "%s/%s.nffw", DEFAULT_FW_PATH,
+			serial);
+
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	if (rte_firmware_read(fw_name, &fw_buf, &fsize) == 0)
+		goto load_fw;
+	/* Then try the PCI name */
+	snprintf(fw_name, sizeof(fw_name), "%s/pci-%s.nffw", DEFAULT_FW_PATH,
+			dev->device.name);
+
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	if (rte_firmware_read(fw_name, &fw_buf, &fsize) == 0)
+		goto load_fw;
+
+	/* Finally try the card type and media */
+	snprintf(fw_name, sizeof(fw_name), "%s/%s", DEFAULT_FW_PATH, card);
+	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
+	if (rte_firmware_read(fw_name, &fw_buf, &fsize) < 0) {
+		PMD_DRV_LOG(INFO, "Firmware file %s not found.", fw_name);
+		return -ENOENT;
+	}
+
+load_fw:
+	PMD_DRV_LOG(INFO, "Firmware file found at %s with size: %zu",
+		fw_name, fsize);
+	PMD_DRV_LOG(INFO, "Uploading the firmware ...");
+	nfp_nsp_load_fw(nsp, fw_buf, fsize);
+	PMD_DRV_LOG(INFO, "Done");
+
+	free(fw_buf);
+
+	return 0;
+}
+
+static int
+nfp_fw_setup(struct rte_pci_device *dev, struct nfp_cpp *cpp,
+	     struct nfp_eth_table *nfp_eth_table, struct nfp_hwinfo *hwinfo)
+{
+	struct nfp_nsp *nsp;
+	const char *nfp_fw_model;
+	char card_desc[100];
+	int err = 0;
+
+	nfp_fw_model = nfp_hwinfo_lookup(hwinfo, "assembly.partno");
+
+	if (nfp_fw_model) {
+		PMD_DRV_LOG(INFO, "firmware model found: %s", nfp_fw_model);
+	} else {
+		PMD_DRV_LOG(ERR, "firmware model NOT found");
+		return -EIO;
+	}
+
+	if (nfp_eth_table->count == 0 || nfp_eth_table->count > 8) {
+		PMD_DRV_LOG(ERR, "NFP ethernet table reports wrong ports: %u",
+		       nfp_eth_table->count);
+		return -EIO;
+	}
+
+	PMD_DRV_LOG(INFO, "NFP ethernet port table reports %u ports",
+			   nfp_eth_table->count);
+
+	PMD_DRV_LOG(INFO, "Port speed: %u", nfp_eth_table->ports[0].speed);
+
+	snprintf(card_desc, sizeof(card_desc), "nic_%s_%dx%d.nffw",
+			nfp_fw_model, nfp_eth_table->count,
+			nfp_eth_table->ports[0].speed / 1000);
+
+	nsp = nfp_nsp_open(cpp);
+	if (!nsp) {
+		PMD_DRV_LOG(ERR, "NFP error when obtaining NSP handle");
+		return -EIO;
+	}
+
+	nfp_nsp_device_soft_reset(nsp);
+	err = nfp_fw_upload(dev, nsp, card_desc);
+
+	nfp_nsp_close(nsp);
+	return err;
+}
+
+static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
+{
+	struct nfp_net_hw *hw;
+	struct rte_eth_dev *eth_dev;
+	struct nfp_eth_table *nfp_eth_table = NULL;
+	int ret = 0;
+	int i;
+
+	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
+	if (!nfp_eth_table) {
+		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
+		ret = -EIO;
+		goto error;
+	}
+
+	/* Loop through all physical ports on PF */
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		const unsigned int numa_node = rte_socket_id();
+		char port_name[RTE_ETH_NAME_MAX_LEN];
+
+		snprintf(port_name, sizeof(port_name), "%s_port%d",
+			 pf_dev->pci_dev->device.name, i);
+
+		/* Allocate a eth_dev for this phyport */
+		eth_dev = rte_eth_dev_allocate(port_name);
+		if (!eth_dev) {
+			ret = -ENODEV;
+			goto port_cleanup;
+		}
+
+		/* Allocate memory for this phyport */
+		eth_dev->data->dev_private =
+			rte_zmalloc_socket(port_name, sizeof(struct nfp_net_hw),
+					   RTE_CACHE_LINE_SIZE, numa_node);
+		if (!eth_dev->data->dev_private) {
+			ret = -ENOMEM;
+			rte_eth_dev_release_port(eth_dev);
+			goto port_cleanup;
+		}
+
+		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+
+		/* Add this device to the PF's array of physical ports */
+		pf_dev->ports[i] = hw;
+
+		hw->pf_dev = pf_dev;
+		hw->cpp = pf_dev->cpp;
+		hw->eth_dev = eth_dev;
+		hw->idx = i;
+		hw->nfp_idx = nfp_eth_table->ports[i].index;
+		hw->is_phyport = true;
+
+		eth_dev->device = &pf_dev->pci_dev->device;
+
+		/* ctrl/tx/rx BAR mappings and remaining init happens in
+		 * nfp_net_init
+		 */
+		ret = nfp_net_init(eth_dev);
+
+		if (ret) {
+			ret = -ENODEV;
+			goto port_cleanup;
+		}
+
+		rte_eth_dev_probing_finish(eth_dev);
+
+	} /* End loop, all ports on this PF */
+	ret = 0;
+	goto eth_table_cleanup;
+
+port_cleanup:
+	for (i = 0; i < pf_dev->total_phyports; i++) {
+		if (pf_dev->ports[i] && pf_dev->ports[i]->eth_dev) {
+			struct rte_eth_dev *tmp_dev;
+			tmp_dev = pf_dev->ports[i]->eth_dev;
+			rte_eth_dev_release_port(tmp_dev);
+			pf_dev->ports[i] = NULL;
+		}
+	}
+eth_table_cleanup:
+	free(nfp_eth_table);
+error:
+	return ret;
+}
+
+static int nfp_pf_init(struct rte_pci_device *pci_dev)
+{
+	struct nfp_pf_dev *pf_dev = NULL;
+	struct nfp_cpp *cpp;
+	struct nfp_hwinfo *hwinfo;
+	struct nfp_rtsym_table *sym_tbl;
+	struct nfp_eth_table *nfp_eth_table = NULL;
+	char name[RTE_ETH_NAME_MAX_LEN];
+	int total_ports;
+	int ret = -ENODEV;
+	int err;
+
+	if (!pci_dev)
+		return ret;
+
+	/*
+	 * When device bound to UIO, the device could be used, by mistake,
+	 * by two DPDK apps, and the UIO driver does not avoid it. This
+	 * could lead to a serious problem when configuring the NFP CPP
+	 * interface. Here we avoid this telling to the CPP init code to
+	 * use a lock file if UIO is being used.
+	 */
+	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
+		cpp = nfp_cpp_from_device_name(pci_dev, 0);
+	else
+		cpp = nfp_cpp_from_device_name(pci_dev, 1);
+
+	if (!cpp) {
+		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
+		ret = -EIO;
+		goto error;
+	}
+
+	hwinfo = nfp_hwinfo_read(cpp);
+	if (!hwinfo) {
+		PMD_INIT_LOG(ERR, "Error reading hwinfo table");
+		ret = -EIO;
+		goto error;
+	}
+
+	nfp_eth_table = nfp_eth_read_ports(cpp);
+	if (!nfp_eth_table) {
+		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
+		ret = -EIO;
+		goto hwinfo_cleanup;
+	}
+
+	if (nfp_fw_setup(pci_dev, cpp, nfp_eth_table, hwinfo)) {
+		PMD_INIT_LOG(ERR, "Error when uploading firmware");
+		ret = -EIO;
+		goto eth_table_cleanup;
+	}
+
+	/* Now the symbol table should be there */
+	sym_tbl = nfp_rtsym_table_read(cpp);
+	if (!sym_tbl) {
+		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
+				" symbol table");
+		ret = -EIO;
+		goto eth_table_cleanup;
+	}
+
+	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
+	if (total_ports != (int)nfp_eth_table->count) {
+		PMD_DRV_LOG(ERR, "Inconsistent number of ports");
+		ret = -EIO;
+		goto sym_tbl_cleanup;
+	}
+
+	PMD_INIT_LOG(INFO, "Total physical ports: %d", total_ports);
+
+	if (total_ports <= 0 || total_ports > 8) {
+		PMD_INIT_LOG(ERR, "nfd_cfg_pf0_num_ports symbol with wrong value");
+		ret = -ENODEV;
+		goto sym_tbl_cleanup;
+	}
+	/* Allocate memory for the PF "device" */
+	snprintf(name, sizeof(name), "nfp_pf%d", 0);
+	pf_dev = rte_zmalloc(name, sizeof(*pf_dev), 0);
+	if (!pf_dev) {
+		ret = -ENOMEM;
+		goto sym_tbl_cleanup;
+	}
+
+	/* Populate the newly created PF device */
+	pf_dev->cpp = cpp;
+	pf_dev->hwinfo = hwinfo;
+	pf_dev->sym_tbl = sym_tbl;
+	pf_dev->total_phyports = total_ports;
+
+	if (total_ports > 1)
+		pf_dev->multiport = true;
+
+	pf_dev->pci_dev = pci_dev;
+
+	/* Map the symbol table */
+	pf_dev->ctrl_bar = nfp_rtsym_map(pf_dev->sym_tbl, "_pf0_net_bar0",
+				     pf_dev->total_phyports * 32768,
+				     &pf_dev->ctrl_area);
+	if (!pf_dev->ctrl_bar) {
+		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for _pf0_net_ctrl_bar");
+		ret = -EIO;
+		goto pf_cleanup;
+	}
+
+	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", pf_dev->ctrl_bar);
+
+	/* configure access to tx/rx vNIC BARs */
+	pf_dev->hw_queues = nfp_cpp_map_area(pf_dev->cpp, 0, 0,
+					      NFP_PCIE_QUEUE(0),
+					      NFP_QCP_QUEUE_AREA_SZ,
+					      &pf_dev->hwqueues_area);
+	if (!pf_dev->hw_queues) {
+		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for net.qc");
+		ret = -EIO;
+		goto ctrl_area_cleanup;
+	}
+
+	PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p", pf_dev->hw_queues);
+
+	/* Initialize and prep physical ports now
+	 * This will loop through all physical ports
+	 */
+	ret = nfp_init_phyports(pf_dev);
+	if (ret) {
+		PMD_INIT_LOG(ERR, "Could not create physical ports");
+		goto hwqueues_cleanup;
+	}
+
+	/* register the CPP bridge service here for primary use */
+	nfp_register_cpp_service(pf_dev->cpp);
+
+	return 0;
+
+hwqueues_cleanup:
+	nfp_cpp_area_free(pf_dev->hwqueues_area);
+ctrl_area_cleanup:
+	nfp_cpp_area_free(pf_dev->ctrl_area);
+pf_cleanup:
+	rte_free(pf_dev);
+sym_tbl_cleanup:
+	free(sym_tbl);
+eth_table_cleanup:
+	free(nfp_eth_table);
+hwinfo_cleanup:
+	free(hwinfo);
+error:
+	return ret;
+}
+
+/*
+ * When attaching to the NFP4000/6000 PF on a secondary process there
+ * is no need to initilise the PF again. Only minimal work is required
+ * here
+ */
+static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev)
+{
+	struct nfp_cpp *cpp;
+	struct nfp_rtsym_table *sym_tbl;
+	int total_ports;
+	int i;
+	int err;
+
+	if (!pci_dev)
+		return -ENODEV;
+
+	/*
+	 * When device bound to UIO, the device could be used, by mistake,
+	 * by two DPDK apps, and the UIO driver does not avoid it. This
+	 * could lead to a serious problem when configuring the NFP CPP
+	 * interface. Here we avoid this telling to the CPP init code to
+	 * use a lock file if UIO is being used.
+	 */
+	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
+		cpp = nfp_cpp_from_device_name(pci_dev, 0);
+	else
+		cpp = nfp_cpp_from_device_name(pci_dev, 1);
+
+	if (!cpp) {
+		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
+		return -EIO;
+	}
+
+	/*
+	 * We don't have access to the PF created in the primary process
+	 * here so we have to read the number of ports from firmware
+	 */
+	sym_tbl = nfp_rtsym_table_read(cpp);
+	if (!sym_tbl) {
+		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
+				" symbol table");
+		return -EIO;
+	}
+
+	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
+
+	for (i = 0; i < total_ports; i++) {
+		struct rte_eth_dev *eth_dev;
+		char port_name[RTE_ETH_NAME_MAX_LEN];
+
+		snprintf(port_name, sizeof(port_name), "%s_port%d",
+			 pci_dev->device.name, i);
+
+		PMD_DRV_LOG(DEBUG, "Secondary attaching to port %s",
+		    port_name);
+		eth_dev = rte_eth_dev_attach_secondary(port_name);
+		if (!eth_dev) {
+			RTE_LOG(ERR, EAL,
+			"secondary process attach failed, "
+			"ethdev doesn't exist");
+			return -ENODEV;
+		}
+		eth_dev->process_private = cpp;
+		eth_dev->dev_ops = &nfp_net_eth_dev_ops;
+		eth_dev->rx_queue_count = nfp_net_rx_queue_count;
+		eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
+		eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
+		rte_eth_dev_probing_finish(eth_dev);
+	}
+
+	/* Register the CPP bridge service for the secondary too */
+	nfp_register_cpp_service(cpp);
+
+	return 0;
+}
+
+static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+			    struct rte_pci_device *dev)
+{
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		return nfp_pf_init(dev);
+	else
+		return nfp_pf_secondary_init(dev);
+}
+
+static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP4000_PF_NIC)
+	},
+	{
+		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+			       PCI_DEVICE_ID_NFP6000_PF_NIC)
+	},
+	{
+		.vendor_id = 0,
+	},
+};
+
+static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
+{
+	struct rte_pci_device *pci_dev;
+	uint16_t port_id;
+
+	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+	/* Free up all physical ports under PF */
+	RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
+		rte_eth_dev_close(port_id);
+	/*
+	 * Ports can be closed and freed but hotplugging is not
+	 * currently supported
+	 */
+	return -ENOTSUP;
+}
+
+static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
+{
+	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
+}
+
+static struct rte_pci_driver rte_nfp_net_pf_pmd = {
+	.id_table = pci_id_nfp_pf_net_map,
+	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+	.probe = nfp_pf_pci_probe,
+	.remove = eth_nfp_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
+/*
+ * Local variables:
+ * c-file-style: "Linux"
+ * indent-tabs-mode: t
+ * End:
+ */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 3c71b513ac..a6097eaab0 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -30,8 +30,6 @@
 #include <rte_spinlock.h>
 #include <rte_service_component.h>
 
-#include "eal_firmware.h"
-
 #include "nfpcore/nfp_cpp.h"
 #include "nfpcore/nfp_nffw.h"
 #include "nfpcore/nfp_hwinfo.h"
@@ -39,10 +37,10 @@
 #include "nfpcore/nfp_rtsym.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 #include "nfp_cpp_bridge.h"
 
 #include <sys/types.h>
@@ -53,20 +51,6 @@
 #include <sys/ioctl.h>
 #include <errno.h>
 
-/* Prototypes */
-static int nfp_net_close(struct rte_eth_dev *dev);
-static int nfp_net_init(struct rte_eth_dev *eth_dev);
-static int nfp_pf_init(struct rte_pci_device *pci_dev);
-static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev);
-static int nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port);
-static int nfp_pci_uninit(struct rte_eth_dev *eth_dev);
-static int nfp_init_phyports(struct nfp_pf_dev *pf_dev);
-static int nfp_net_stop(struct rte_eth_dev *dev);
-static int nfp_fw_setup(struct rte_pci_device *dev,
-			struct nfp_cpp *cpp,
-			struct nfp_eth_table *nfp_eth_table,
-			struct nfp_hwinfo *hwinfo);
-
 static int
 __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update)
 {
@@ -268,24 +252,6 @@ nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
 		dst[i] = src[i];
 }
 
-static int
-nfp_net_pf_read_mac(struct nfp_pf_dev *pf_dev, int port)
-{
-	struct nfp_eth_table *nfp_eth_table;
-	struct nfp_net_hw *hw = NULL;
-
-	/* Grab a pointer to the correct physical port */
-	hw = pf_dev->ports[port];
-
-	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
-
-	nfp_eth_copy_mac((uint8_t *)&hw->mac_addr,
-			 (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
-
-	free(nfp_eth_table);
-	return 0;
-}
-
 void
 nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
 {
@@ -438,282 +404,6 @@ nfp_check_offloads(struct rte_eth_dev *dev)
 	return ctrl;
 }
 
-static int
-nfp_net_start(struct rte_eth_dev *dev)
-{
-	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
-	uint32_t new_ctrl, update = 0;
-	struct nfp_net_hw *hw;
-	struct nfp_pf_dev *pf_dev;
-	struct rte_eth_conf *dev_conf;
-	struct rte_eth_rxmode *rxmode;
-	uint32_t intr_vector;
-	int ret;
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
-
-	PMD_INIT_LOG(DEBUG, "Start");
-
-	/* Disabling queues just in case... */
-	nfp_net_disable_queues(dev);
-
-	/* Enabling the required queues in the device */
-	nfp_net_enable_queues(dev);
-
-	/* check and configure queue intr-vector mapping */
-	if (dev->data->dev_conf.intr_conf.rxq != 0) {
-		if (pf_dev->multiport) {
-			PMD_INIT_LOG(ERR, "PMD rx interrupt is not supported "
-					  "with NFP multiport PF");
-				return -EINVAL;
-		}
-		if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
-			/*
-			 * Better not to share LSC with RX interrupts.
-			 * Unregistering LSC interrupt handler
-			 */
-			rte_intr_callback_unregister(&pci_dev->intr_handle,
-				nfp_net_dev_interrupt_handler, (void *)dev);
-
-			if (dev->data->nb_rx_queues > 1) {
-				PMD_INIT_LOG(ERR, "PMD rx interrupt only "
-					     "supports 1 queue with UIO");
-				return -EIO;
-			}
-		}
-		intr_vector = dev->data->nb_rx_queues;
-		if (rte_intr_efd_enable(intr_handle, intr_vector))
-			return -1;
-
-		nfp_configure_rx_interrupt(dev, intr_handle);
-		update = NFP_NET_CFG_UPDATE_MSIX;
-	}
-
-	rte_intr_enable(intr_handle);
-
-	new_ctrl = nfp_check_offloads(dev);
-
-	/* Writing configuration parameters in the device */
-	nfp_net_params_setup(hw);
-
-	dev_conf = &dev->data->dev_conf;
-	rxmode = &dev_conf->rxmode;
-
-	if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
-		nfp_net_rss_config_default(dev);
-		update |= NFP_NET_CFG_UPDATE_RSS;
-		new_ctrl |= NFP_NET_CFG_CTRL_RSS;
-	}
-
-	/* Enable device */
-	new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
-
-	update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
-
-	if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
-		new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
-
-	nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl);
-	if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
-		return -EIO;
-
-	/*
-	 * Allocating rte mbufs for configured rx queues.
-	 * This requires queues being enabled before
-	 */
-	if (nfp_net_rx_freelist_setup(dev) < 0) {
-		ret = -ENOMEM;
-		goto error;
-	}
-
-	if (hw->is_phyport) {
-		if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-			/* Configure the physical port up */
-			nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
-		else
-			nfp_eth_set_configured(dev->process_private,
-					       hw->nfp_idx, 1);
-	}
-
-	hw->ctrl = new_ctrl;
-
-	return 0;
-
-error:
-	/*
-	 * An error returned by this function should mean the app
-	 * exiting and then the system releasing all the memory
-	 * allocated even memory coming from hugepages.
-	 *
-	 * The device could be enabled at this point with some queues
-	 * ready for getting packets. This is true if the call to
-	 * nfp_net_rx_freelist_setup() succeeds for some queues but
-	 * fails for subsequent queues.
-	 *
-	 * This should make the app exiting but better if we tell the
-	 * device first.
-	 */
-	nfp_net_disable_queues(dev);
-
-	return ret;
-}
-
-/* Stop device: disable rx and tx functions to allow for reconfiguring. */
-static int
-nfp_net_stop(struct rte_eth_dev *dev)
-{
-	int i;
-	struct nfp_net_hw *hw;
-
-	PMD_INIT_LOG(DEBUG, "Stop");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	nfp_net_disable_queues(dev);
-
-	/* Clear queues */
-	for (i = 0; i < dev->data->nb_tx_queues; i++) {
-		nfp_net_reset_tx_queue(
-			(struct nfp_net_txq *)dev->data->tx_queues[i]);
-	}
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		nfp_net_reset_rx_queue(
-			(struct nfp_net_rxq *)dev->data->rx_queues[i]);
-	}
-
-	if (hw->is_phyport) {
-		if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-			/* Configure the physical port down */
-			nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
-		else
-			nfp_eth_set_configured(dev->process_private,
-					       hw->nfp_idx, 0);
-	}
-
-	return 0;
-}
-
-/* Set the link up. */
-static int
-nfp_net_set_link_up(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-
-	PMD_DRV_LOG(DEBUG, "Set link up");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	if (!hw->is_phyport)
-		return -ENOTSUP;
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		/* Configure the physical port down */
-		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 1);
-	else
-		return nfp_eth_set_configured(dev->process_private,
-					      hw->nfp_idx, 1);
-}
-
-/* Set the link down. */
-static int
-nfp_net_set_link_down(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-
-	PMD_DRV_LOG(DEBUG, "Set link down");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-	if (!hw->is_phyport)
-		return -ENOTSUP;
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		/* Configure the physical port down */
-		return nfp_eth_set_configured(hw->cpp, hw->nfp_idx, 0);
-	else
-		return nfp_eth_set_configured(dev->process_private,
-					      hw->nfp_idx, 0);
-}
-
-/* Reset and stop device. The device can not be restarted. */
-static int
-nfp_net_close(struct rte_eth_dev *dev)
-{
-	struct nfp_net_hw *hw;
-	struct rte_pci_device *pci_dev;
-	int i;
-
-	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
-		return 0;
-
-	PMD_INIT_LOG(DEBUG, "Close");
-
-	hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-
-	/*
-	 * We assume that the DPDK application is stopping all the
-	 * threads/queues before calling the device close function.
-	 */
-
-	nfp_net_disable_queues(dev);
-
-	/* Clear queues */
-	for (i = 0; i < dev->data->nb_tx_queues; i++) {
-		nfp_net_reset_tx_queue(
-			(struct nfp_net_txq *)dev->data->tx_queues[i]);
-	}
-
-	for (i = 0; i < dev->data->nb_rx_queues; i++) {
-		nfp_net_reset_rx_queue(
-			(struct nfp_net_rxq *)dev->data->rx_queues[i]);
-	}
-
-	/* Only free PF resources after all physical ports have been closed */
-	if (pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC ||
-	    pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC) {
-		struct nfp_pf_dev *pf_dev;
-		pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(dev->data->dev_private);
-
-		/* Mark this port as unused and free device priv resources*/
-		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
-		pf_dev->ports[hw->idx] = NULL;
-		rte_eth_dev_release_port(dev);
-
-		for (i = 0; i < pf_dev->total_phyports; i++) {
-			/* Check to see if ports are still in use */
-			if (pf_dev->ports[i])
-				return 0;
-		}
-
-		/* Now it is safe to free all PF resources */
-		PMD_INIT_LOG(INFO, "Freeing PF resources");
-		nfp_cpp_area_free(pf_dev->ctrl_area);
-		nfp_cpp_area_free(pf_dev->hwqueues_area);
-		free(pf_dev->hwinfo);
-		free(pf_dev->sym_tbl);
-		nfp_cpp_free(pf_dev->cpp);
-		rte_free(pf_dev);
-	}
-
-	rte_intr_disable(&pci_dev->intr_handle);
-
-	/* unregister callback func from eal lib */
-	rte_intr_callback_unregister(&pci_dev->intr_handle,
-				     nfp_net_dev_interrupt_handler,
-				     (void *)dev);
-
-	/*
-	 * The ixgbe PMD driver disables the pcie master on the
-	 * device. The i40e does not...
-	 */
-
-	return 0;
-}
-
 int
 nfp_net_promisc_enable(struct rte_eth_dev *dev)
 {
@@ -1622,746 +1312,6 @@ nfp_net_rss_config_default(struct rte_eth_dev *dev)
 	return ret;
 }
 
-
-/* Initialise and register driver with DPDK Application */
-static const struct eth_dev_ops nfp_net_eth_dev_ops = {
-	.dev_configure		= nfp_net_configure,
-	.dev_start		= nfp_net_start,
-	.dev_stop		= nfp_net_stop,
-	.dev_set_link_up	= nfp_net_set_link_up,
-	.dev_set_link_down	= nfp_net_set_link_down,
-	.dev_close		= nfp_net_close,
-	.promiscuous_enable	= nfp_net_promisc_enable,
-	.promiscuous_disable	= nfp_net_promisc_disable,
-	.link_update		= nfp_net_link_update,
-	.stats_get		= nfp_net_stats_get,
-	.stats_reset		= nfp_net_stats_reset,
-	.dev_infos_get		= nfp_net_infos_get,
-	.dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
-	.mtu_set		= nfp_net_dev_mtu_set,
-	.mac_addr_set           = nfp_set_mac_addr,
-	.vlan_offload_set	= nfp_net_vlan_offload_set,
-	.reta_update		= nfp_net_reta_update,
-	.reta_query		= nfp_net_reta_query,
-	.rss_hash_update	= nfp_net_rss_hash_update,
-	.rss_hash_conf_get	= nfp_net_rss_hash_conf_get,
-	.rx_queue_setup		= nfp_net_rx_queue_setup,
-	.rx_queue_release	= nfp_net_rx_queue_release,
-	.tx_queue_setup		= nfp_net_tx_queue_setup,
-	.tx_queue_release	= nfp_net_tx_queue_release,
-	.rx_queue_intr_enable   = nfp_rx_queue_intr_enable,
-	.rx_queue_intr_disable  = nfp_rx_queue_intr_disable,
-};
-
-
-static int
-nfp_net_init(struct rte_eth_dev *eth_dev)
-{
-	struct rte_pci_device *pci_dev;
-	struct nfp_pf_dev *pf_dev;
-	struct nfp_net_hw *hw;
-
-	uint64_t tx_bar_off = 0, rx_bar_off = 0;
-	uint32_t start_q;
-	int stride = 4;
-	int port = 0;
-	int err;
-
-	PMD_INIT_FUNC_TRACE();
-
-	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
-
-	/* Use backpointer here to the PF of this eth_dev */
-	pf_dev = NFP_NET_DEV_PRIVATE_TO_PF(eth_dev->data->dev_private);
-
-	/* NFP can not handle DMA addresses requiring more than 40 bits */
-	if (rte_mem_check_dma_mask(40)) {
-		RTE_LOG(ERR, PMD, "device %s can not be used:",
-				   pci_dev->device.name);
-		RTE_LOG(ERR, PMD, "\trestricted dma mask to 40 bits!\n");
-		return -ENODEV;
-	};
-
-	if ((pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC) ||
-	    (pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC)) {
-		port = ((struct nfp_net_hw *)eth_dev->data->dev_private)->idx;
-		if (port < 0 || port > 7) {
-			PMD_DRV_LOG(ERR, "Port value is wrong");
-			return -ENODEV;
-		}
-
-		/* Use PF array of physical ports to get pointer to
-		 * this specific port
-		 */
-		hw = pf_dev->ports[port];
-
-		PMD_INIT_LOG(DEBUG, "Working with physical port number: %d, "
-				    "NFP internal port number: %d",
-				    port, hw->nfp_idx);
-
-	} else {
-		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
-	}
-
-	eth_dev->dev_ops = &nfp_net_eth_dev_ops;
-	eth_dev->rx_queue_count = nfp_net_rx_queue_count;
-	eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
-	eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
-
-	/* For secondary processes, the primary has done all the work */
-	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
-		return 0;
-
-	rte_eth_copy_pci_info(eth_dev, pci_dev);
-
-	hw->device_id = pci_dev->id.device_id;
-	hw->vendor_id = pci_dev->id.vendor_id;
-	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
-	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
-
-	PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
-		     pci_dev->id.vendor_id, pci_dev->id.device_id,
-		     pci_dev->addr.domain, pci_dev->addr.bus,
-		     pci_dev->addr.devid, pci_dev->addr.function);
-
-	hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr;
-	if (hw->ctrl_bar == NULL) {
-		PMD_DRV_LOG(ERR,
-			"hw->ctrl_bar is NULL. BAR0 not configured");
-		return -ENODEV;
-	}
-
-	if (hw->is_phyport) {
-		if (port == 0) {
-			hw->ctrl_bar = pf_dev->ctrl_bar;
-		} else {
-			if (!pf_dev->ctrl_bar)
-				return -ENODEV;
-			/* Use port offset in pf ctrl_bar for this
-			 * ports control bar
-			 */
-			hw->ctrl_bar = pf_dev->ctrl_bar +
-				       (port * NFP_PF_CSR_SLICE_SIZE);
-		}
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
-
-	hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
-	hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
-
-	/* Work out where in the BAR the queues start. */
-	switch (pci_dev->id.device_id) {
-	case PCI_DEVICE_ID_NFP4000_PF_NIC:
-	case PCI_DEVICE_ID_NFP6000_PF_NIC:
-	case PCI_DEVICE_ID_NFP6000_VF_NIC:
-		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
-		tx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
-		start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ);
-		rx_bar_off = (uint64_t)start_q * NFP_QCP_QUEUE_ADDR_SZ;
-		break;
-	default:
-		PMD_DRV_LOG(ERR, "nfp_net: no device ID matching");
-		err = -ENODEV;
-		goto dev_err_ctrl_map;
-	}
-
-	PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%" PRIx64 "", tx_bar_off);
-	PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "", rx_bar_off);
-
-	if (hw->is_phyport) {
-		hw->tx_bar = pf_dev->hw_queues + tx_bar_off;
-		hw->rx_bar = pf_dev->hw_queues + rx_bar_off;
-		eth_dev->data->dev_private = hw;
-	} else {
-		hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
-			     tx_bar_off;
-		hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
-			     rx_bar_off;
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
-		     hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
-
-	nfp_net_cfg_queue_setup(hw);
-
-	/* Get some of the read-only fields from the config BAR */
-	hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
-	hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
-	hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
-	hw->mtu = RTE_ETHER_MTU;
-
-	/* VLAN insertion is incompatible with LSOv2 */
-	if (hw->cap & NFP_NET_CFG_CTRL_LSO2)
-		hw->cap &= ~NFP_NET_CFG_CTRL_TXVLAN;
-
-	if (NFD_CFG_MAJOR_VERSION_of(hw->ver) < 2)
-		hw->rx_offset = NFP_NET_RX_OFFSET;
-	else
-		hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
-
-	PMD_INIT_LOG(INFO, "VER: %u.%u, Maximum supported MTU: %d",
-			   NFD_CFG_MAJOR_VERSION_of(hw->ver),
-			   NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
-
-	PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
-		     hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_L2BC    ? "L2BCFILT " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_L2MC    ? "L2MCFILT " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RXCSUM  ? "RXCSUM "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_TXCSUM  ? "TXCSUM "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RXVLAN  ? "RXVLAN "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_TXVLAN  ? "TXVLAN "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
-		     hw->cap & NFP_NET_CFG_CTRL_GATHER  ? "GATHER "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR "  : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LSO     ? "TSO "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_LSO2     ? "TSOv2 "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RSS     ? "RSS "     : "",
-		     hw->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSSv2 "     : "");
-
-	hw->ctrl = 0;
-
-	hw->stride_rx = stride;
-	hw->stride_tx = stride;
-
-	PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
-		     hw->max_rx_queues, hw->max_tx_queues);
-
-	/* Initializing spinlock for reconfigs */
-	rte_spinlock_init(&hw->reconfig_lock);
-
-	/* Allocating memory for mac addr */
-	eth_dev->data->mac_addrs = rte_zmalloc("mac_addr",
-					       RTE_ETHER_ADDR_LEN, 0);
-	if (eth_dev->data->mac_addrs == NULL) {
-		PMD_INIT_LOG(ERR, "Failed to space for MAC address");
-		err = -ENOMEM;
-		goto dev_err_queues_map;
-	}
-
-	if (hw->is_phyport) {
-		nfp_net_pf_read_mac(pf_dev, port);
-		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	}
-
-	if (!rte_is_valid_assigned_ether_addr(
-		    (struct rte_ether_addr *)&hw->mac_addr)) {
-		PMD_INIT_LOG(INFO, "Using random mac address for port %d",
-				   port);
-		/* Using random mac addresses for VFs */
-		rte_eth_random_addr(&hw->mac_addr[0]);
-		nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
-	}
-
-	/* Copying mac address to DPDK eth_dev struct */
-	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac_addr,
-			&eth_dev->data->mac_addrs[0]);
-
-	if (!(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR))
-		eth_dev->data->dev_flags |= RTE_ETH_DEV_NOLIVE_MAC_ADDR;
-
-	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
-
-	PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
-		     "mac=%02x:%02x:%02x:%02x:%02x:%02x",
-		     eth_dev->data->port_id, pci_dev->id.vendor_id,
-		     pci_dev->id.device_id,
-		     hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
-		     hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
-
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
-		/* Registering LSC interrupt handler */
-		rte_intr_callback_register(&pci_dev->intr_handle,
-					   nfp_net_dev_interrupt_handler,
-					   (void *)eth_dev);
-		/* Telling the firmware about the LSC interrupt entry */
-		nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
-		/* Recording current stats counters values */
-		nfp_net_stats_reset(eth_dev);
-	}
-
-	return 0;
-
-dev_err_queues_map:
-		nfp_cpp_area_free(hw->hwqueues_area);
-dev_err_ctrl_map:
-		nfp_cpp_area_free(hw->ctrl_area);
-
-	return err;
-}
-
-#define DEFAULT_FW_PATH       "/lib/firmware/netronome"
-
-static int
-nfp_fw_upload(struct rte_pci_device *dev, struct nfp_nsp *nsp, char *card)
-{
-	struct nfp_cpp *cpp = nsp->cpp;
-	void *fw_buf;
-	char fw_name[125];
-	char serial[40];
-	size_t fsize;
-
-	/* Looking for firmware file in order of priority */
-
-	/* First try to find a firmware image specific for this device */
-	snprintf(serial, sizeof(serial),
-			"serial-%02x-%02x-%02x-%02x-%02x-%02x-%02x-%02x",
-		cpp->serial[0], cpp->serial[1], cpp->serial[2], cpp->serial[3],
-		cpp->serial[4], cpp->serial[5], cpp->interface >> 8,
-		cpp->interface & 0xff);
-
-	snprintf(fw_name, sizeof(fw_name), "%s/%s.nffw", DEFAULT_FW_PATH,
-			serial);
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	if (rte_firmware_read(fw_name, &fw_buf, &fsize) == 0)
-		goto load_fw;
-
-	/* Then try the PCI name */
-	snprintf(fw_name, sizeof(fw_name), "%s/pci-%s.nffw", DEFAULT_FW_PATH,
-			dev->device.name);
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	if (rte_firmware_read(fw_name, &fw_buf, &fsize) == 0)
-		goto load_fw;
-
-	/* Finally try the card type and media */
-	snprintf(fw_name, sizeof(fw_name), "%s/%s", DEFAULT_FW_PATH, card);
-	PMD_DRV_LOG(DEBUG, "Trying with fw file: %s", fw_name);
-	if (rte_firmware_read(fw_name, &fw_buf, &fsize) < 0) {
-		PMD_DRV_LOG(INFO, "Firmware file %s not found.", fw_name);
-		return -ENOENT;
-	}
-
-load_fw:
-	PMD_DRV_LOG(INFO, "Firmware file found at %s with size: %zu",
-		fw_name, fsize);
-
-	PMD_DRV_LOG(INFO, "Uploading the firmware ...");
-	nfp_nsp_load_fw(nsp, fw_buf, fsize);
-	PMD_DRV_LOG(INFO, "Done");
-
-	free(fw_buf);
-	return 0;
-}
-
-static int
-nfp_fw_setup(struct rte_pci_device *dev, struct nfp_cpp *cpp,
-	     struct nfp_eth_table *nfp_eth_table, struct nfp_hwinfo *hwinfo)
-{
-	struct nfp_nsp *nsp;
-	const char *nfp_fw_model;
-	char card_desc[100];
-	int err = 0;
-
-	nfp_fw_model = nfp_hwinfo_lookup(hwinfo, "assembly.partno");
-
-	if (nfp_fw_model) {
-		PMD_DRV_LOG(INFO, "firmware model found: %s", nfp_fw_model);
-	} else {
-		PMD_DRV_LOG(ERR, "firmware model NOT found");
-		return -EIO;
-	}
-
-	if (nfp_eth_table->count == 0 || nfp_eth_table->count > 8) {
-		PMD_DRV_LOG(ERR, "NFP ethernet table reports wrong ports: %u",
-		       nfp_eth_table->count);
-		return -EIO;
-	}
-
-	PMD_DRV_LOG(INFO, "NFP ethernet port table reports %u ports",
-			   nfp_eth_table->count);
-
-	PMD_DRV_LOG(INFO, "Port speed: %u", nfp_eth_table->ports[0].speed);
-
-	snprintf(card_desc, sizeof(card_desc), "nic_%s_%dx%d.nffw",
-			nfp_fw_model, nfp_eth_table->count,
-			nfp_eth_table->ports[0].speed / 1000);
-
-	nsp = nfp_nsp_open(cpp);
-	if (!nsp) {
-		PMD_DRV_LOG(ERR, "NFP error when obtaining NSP handle");
-		return -EIO;
-	}
-
-	nfp_nsp_device_soft_reset(nsp);
-	err = nfp_fw_upload(dev, nsp, card_desc);
-
-	nfp_nsp_close(nsp);
-	return err;
-}
-
-static int nfp_init_phyports(struct nfp_pf_dev *pf_dev)
-{
-	struct nfp_net_hw *hw;
-	struct rte_eth_dev *eth_dev;
-	struct nfp_eth_table *nfp_eth_table = NULL;
-	int ret = 0;
-	int i;
-
-	nfp_eth_table = nfp_eth_read_ports(pf_dev->cpp);
-	if (!nfp_eth_table) {
-		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
-		ret = -EIO;
-		goto error;
-	}
-
-	/* Loop through all physical ports on PF */
-	for (i = 0; i < pf_dev->total_phyports; i++) {
-		const unsigned int numa_node = rte_socket_id();
-		char port_name[RTE_ETH_NAME_MAX_LEN];
-
-		snprintf(port_name, sizeof(port_name), "%s_port%d",
-			 pf_dev->pci_dev->device.name, i);
-
-		/* Allocate a eth_dev for this phyport */
-		eth_dev = rte_eth_dev_allocate(port_name);
-		if (!eth_dev) {
-			ret = -ENODEV;
-			goto port_cleanup;
-		}
-
-		/* Allocate memory for this phyport */
-		eth_dev->data->dev_private =
-			rte_zmalloc_socket(port_name, sizeof(struct nfp_net_hw),
-					   RTE_CACHE_LINE_SIZE, numa_node);
-		if (!eth_dev->data->dev_private) {
-			ret = -ENOMEM;
-			rte_eth_dev_release_port(eth_dev);
-			goto port_cleanup;
-		}
-
-		hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
-
-		/* Add this device to the PF's array of physical ports */
-		pf_dev->ports[i] = hw;
-
-		hw->pf_dev = pf_dev;
-		hw->cpp = pf_dev->cpp;
-		hw->eth_dev = eth_dev;
-		hw->idx = i;
-		hw->nfp_idx = nfp_eth_table->ports[i].index;
-		hw->is_phyport = true;
-
-		eth_dev->device = &pf_dev->pci_dev->device;
-
-		/* ctrl/tx/rx BAR mappings and remaining init happens in
-		 * nfp_net_init
-		 */
-		ret = nfp_net_init(eth_dev);
-
-		if (ret) {
-			ret = -ENODEV;
-			goto port_cleanup;
-		}
-
-		rte_eth_dev_probing_finish(eth_dev);
-
-	} /* End loop, all ports on this PF */
-	ret = 0;
-	goto eth_table_cleanup;
-
-port_cleanup:
-	for (i = 0; i < pf_dev->total_phyports; i++) {
-		if (pf_dev->ports[i] && pf_dev->ports[i]->eth_dev) {
-			struct rte_eth_dev *tmp_dev;
-			tmp_dev = pf_dev->ports[i]->eth_dev;
-			rte_eth_dev_release_port(tmp_dev);
-			pf_dev->ports[i] = NULL;
-		}
-	}
-eth_table_cleanup:
-	free(nfp_eth_table);
-error:
-	return ret;
-}
-
-static int nfp_pf_init(struct rte_pci_device *pci_dev)
-{
-	struct nfp_pf_dev *pf_dev = NULL;
-	struct nfp_cpp *cpp;
-	struct nfp_hwinfo *hwinfo;
-	struct nfp_rtsym_table *sym_tbl;
-	struct nfp_eth_table *nfp_eth_table = NULL;
-	char name[RTE_ETH_NAME_MAX_LEN];
-	int total_ports;
-	int ret = -ENODEV;
-	int err;
-
-	if (!pci_dev)
-		return ret;
-
-	/*
-	 * When device bound to UIO, the device could be used, by mistake,
-	 * by two DPDK apps, and the UIO driver does not avoid it. This
-	 * could lead to a serious problem when configuring the NFP CPP
-	 * interface. Here we avoid this telling to the CPP init code to
-	 * use a lock file if UIO is being used.
-	 */
-	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
-		cpp = nfp_cpp_from_device_name(pci_dev, 0);
-	else
-		cpp = nfp_cpp_from_device_name(pci_dev, 1);
-
-	if (!cpp) {
-		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
-		ret = -EIO;
-		goto error;
-	}
-
-	hwinfo = nfp_hwinfo_read(cpp);
-	if (!hwinfo) {
-		PMD_INIT_LOG(ERR, "Error reading hwinfo table");
-		ret = -EIO;
-		goto error;
-	}
-
-	nfp_eth_table = nfp_eth_read_ports(cpp);
-	if (!nfp_eth_table) {
-		PMD_INIT_LOG(ERR, "Error reading NFP ethernet table");
-		ret = -EIO;
-		goto hwinfo_cleanup;
-	}
-
-	if (nfp_fw_setup(pci_dev, cpp, nfp_eth_table, hwinfo)) {
-		PMD_INIT_LOG(ERR, "Error when uploading firmware");
-		ret = -EIO;
-		goto eth_table_cleanup;
-	}
-
-	/* Now the symbol table should be there */
-	sym_tbl = nfp_rtsym_table_read(cpp);
-	if (!sym_tbl) {
-		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
-				" symbol table");
-		ret = -EIO;
-		goto eth_table_cleanup;
-	}
-
-	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
-	if (total_ports != (int)nfp_eth_table->count) {
-		PMD_DRV_LOG(ERR, "Inconsistent number of ports");
-		ret = -EIO;
-		goto sym_tbl_cleanup;
-	}
-
-	PMD_INIT_LOG(INFO, "Total physical ports: %d", total_ports);
-
-	if (total_ports <= 0 || total_ports > 8) {
-		PMD_INIT_LOG(ERR, "nfd_cfg_pf0_num_ports symbol with wrong value");
-		ret = -ENODEV;
-		goto sym_tbl_cleanup;
-	}
-	/* Allocate memory for the PF "device" */
-	snprintf(name, sizeof(name), "nfp_pf%d", 0);
-	pf_dev = rte_zmalloc(name, sizeof(*pf_dev), 0);
-	if (!pf_dev) {
-		ret = -ENOMEM;
-		goto sym_tbl_cleanup;
-	}
-
-	/* Populate the newly created PF device */
-	pf_dev->cpp = cpp;
-	pf_dev->hwinfo = hwinfo;
-	pf_dev->sym_tbl = sym_tbl;
-	pf_dev->total_phyports = total_ports;
-
-	if (total_ports > 1)
-		pf_dev->multiport = true;
-
-	pf_dev->pci_dev = pci_dev;
-
-	/* Map the symbol table */
-	pf_dev->ctrl_bar = nfp_rtsym_map(pf_dev->sym_tbl, "_pf0_net_bar0",
-				     pf_dev->total_phyports * 32768,
-				     &pf_dev->ctrl_area);
-	if (!pf_dev->ctrl_bar) {
-		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for _pf0_net_ctrl_bar");
-		ret = -EIO;
-		goto pf_cleanup;
-	}
-
-	PMD_INIT_LOG(DEBUG, "ctrl bar: %p", pf_dev->ctrl_bar);
-
-	/* configure access to tx/rx vNIC BARs */
-	pf_dev->hw_queues = nfp_cpp_map_area(pf_dev->cpp, 0, 0,
-					      NFP_PCIE_QUEUE(0),
-					      NFP_QCP_QUEUE_AREA_SZ,
-					      &pf_dev->hwqueues_area);
-	if (!pf_dev->hw_queues) {
-		PMD_INIT_LOG(ERR, "nfp_rtsym_map fails for net.qc");
-		ret = -EIO;
-		goto ctrl_area_cleanup;
-	}
-
-	PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p", pf_dev->hw_queues);
-
-	/* Initialize and prep physical ports now
-	 * This will loop through all physical ports
-	 */
-	ret = nfp_init_phyports(pf_dev);
-	if (ret) {
-		PMD_INIT_LOG(ERR, "Could not create physical ports");
-		goto hwqueues_cleanup;
-	}
-
-	/* register the CPP bridge service here for primary use */
-	nfp_register_cpp_service(pf_dev->cpp);
-
-	return 0;
-
-hwqueues_cleanup:
-	nfp_cpp_area_free(pf_dev->hwqueues_area);
-ctrl_area_cleanup:
-	nfp_cpp_area_free(pf_dev->ctrl_area);
-pf_cleanup:
-	rte_free(pf_dev);
-sym_tbl_cleanup:
-	free(sym_tbl);
-eth_table_cleanup:
-	free(nfp_eth_table);
-hwinfo_cleanup:
-	free(hwinfo);
-error:
-	return ret;
-}
-
-/*
- * When attaching to the NFP4000/6000 PF on a secondary process there
- * is no need to initialize the PF again. Only minimal work is required
- * here
- */
-static int nfp_pf_secondary_init(struct rte_pci_device *pci_dev)
-{
-	struct nfp_cpp *cpp;
-	struct nfp_rtsym_table *sym_tbl;
-	int total_ports;
-	int i;
-	int err;
-
-	if (!pci_dev)
-		return -ENODEV;
-
-	/*
-	 * When device bound to UIO, the device could be used, by mistake,
-	 * by two DPDK apps, and the UIO driver does not avoid it. This
-	 * could lead to a serious problem when configuring the NFP CPP
-	 * interface. Here we avoid this telling to the CPP init code to
-	 * use a lock file if UIO is being used.
-	 */
-	if (pci_dev->kdrv == RTE_PCI_KDRV_VFIO)
-		cpp = nfp_cpp_from_device_name(pci_dev, 0);
-	else
-		cpp = nfp_cpp_from_device_name(pci_dev, 1);
-
-	if (!cpp) {
-		PMD_INIT_LOG(ERR, "A CPP handle can not be obtained");
-		return -EIO;
-	}
-
-	/*
-	 * We don't have access to the PF created in the primary process
-	 * here so we have to read the number of ports from firmware
-	 */
-	sym_tbl = nfp_rtsym_table_read(cpp);
-	if (!sym_tbl) {
-		PMD_INIT_LOG(ERR, "Something is wrong with the firmware"
-				" symbol table");
-		return -EIO;
-	}
-
-	total_ports = nfp_rtsym_read_le(sym_tbl, "nfd_cfg_pf0_num_ports", &err);
-
-	for (i = 0; i < total_ports; i++) {
-		struct rte_eth_dev *eth_dev;
-		char port_name[RTE_ETH_NAME_MAX_LEN];
-
-		snprintf(port_name, sizeof(port_name), "%s_port%d",
-			 pci_dev->device.name, i);
-
-		PMD_DRV_LOG(DEBUG, "Secondary attaching to port %s",
-		    port_name);
-		eth_dev = rte_eth_dev_attach_secondary(port_name);
-		if (!eth_dev) {
-			RTE_LOG(ERR, EAL,
-			"secondary process attach failed, "
-			"ethdev doesn't exist");
-			return -ENODEV;
-		}
-		eth_dev->process_private = cpp;
-		eth_dev->dev_ops = &nfp_net_eth_dev_ops;
-		eth_dev->rx_queue_count = nfp_net_rx_queue_count;
-		eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
-		eth_dev->tx_pkt_burst = &nfp_net_xmit_pkts;
-		rte_eth_dev_probing_finish(eth_dev);
-	}
-
-	/* Register the CPP bridge service for the secondary too */
-	nfp_register_cpp_service(cpp);
-
-	return 0;
-}
-
-static int nfp_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
-			    struct rte_pci_device *dev)
-{
-	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
-		return nfp_pf_init(dev);
-	else
-		return nfp_pf_secondary_init(dev);
-}
-
-static const struct rte_pci_id pci_id_nfp_pf_net_map[] = {
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP4000_PF_NIC)
-	},
-	{
-		RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
-			       PCI_DEVICE_ID_NFP6000_PF_NIC)
-	},
-	{
-		.vendor_id = 0,
-	},
-};
-
-static int nfp_pci_uninit(struct rte_eth_dev *eth_dev)
-{
-	struct rte_pci_device *pci_dev;
-	uint16_t port_id;
-
-	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
-
-	if (pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC ||
-	    pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC) {
-		/* Free up all physical ports under PF */
-		RTE_ETH_FOREACH_DEV_OF(port_id, &pci_dev->device)
-			rte_eth_dev_close(port_id);
-		/*
-		 * Ports can be closed and freed but hotplugging is not
-		 * currently supported
-		 */
-		return -ENOTSUP;
-	}
-
-	/* VF cleanup, just free private port data */
-	return nfp_net_close(eth_dev);
-}
-
-static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
-{
-	return rte_eth_dev_pci_generic_remove(pci_dev, nfp_pci_uninit);
-}
-
-static struct rte_pci_driver rte_nfp_net_pf_pmd = {
-	.id_table = pci_id_nfp_pf_net_map,
-	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
-	.probe = nfp_pf_pci_probe,
-	.remove = eth_nfp_pci_remove,
-};
-
-RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
-RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
-RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_init, init, NOTICE);
 RTE_LOG_REGISTER_SUFFIX(nfp_logtype_driver, driver, NOTICE);
 /*
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* [dpdk-dev] [PATCH v3 7/7] net/nfp: batch file rename for consistency
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
                       ` (5 preceding siblings ...)
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 6/7] net/nfp: move PF " Heinrich Kuhn
@ 2021-07-29 13:47     ` Heinrich Kuhn
  2021-08-13 12:46     ` [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD Heinrich Kuhn
  2021-08-17 16:29     ` Ferruh Yigit
  8 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:47 UTC (permalink / raw)
  To: dev; +Cc: Heinrich Kuhn, Simon Horman

Rename the nfp_net.c file to nfp_common as it now contains functions
common to VF and PF functionality. Rename the header file too to be
consistent. Also remove the "net" naming from the _ctrl and _logs files
for consistency across the PMD.

Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
Signed-off-by: Simon Horman <simon.horman@corigine.com>
---
 drivers/net/nfp/meson.build                     | 2 +-
 drivers/net/nfp/{nfp_net.c => nfp_common.c}     | 4 ++--
 drivers/net/nfp/{nfp_net_pmd.h => nfp_common.h} | 6 +++---
 drivers/net/nfp/nfp_cpp_bridge.c                | 2 +-
 drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h}  | 6 +++---
 drivers/net/nfp/nfp_ethdev.c                    | 6 +++---
 drivers/net/nfp/nfp_ethdev_vf.c                 | 6 +++---
 drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h}  | 6 +++---
 drivers/net/nfp/nfp_rxtx.c                      | 6 +++---
 9 files changed, 22 insertions(+), 22 deletions(-)
 rename drivers/net/nfp/{nfp_net.c => nfp_common.c} (99%)
 rename drivers/net/nfp/{nfp_net_pmd.h => nfp_common.h} (99%)
 rename drivers/net/nfp/{nfp_net_ctrl.h => nfp_ctrl.h} (99%)
 rename drivers/net/nfp/{nfp_net_logs.h => nfp_logs.h} (94%)

diff --git a/drivers/net/nfp/meson.build b/drivers/net/nfp/meson.build
index ab64d0cac3..810f02ae5b 100644
--- a/drivers/net/nfp/meson.build
+++ b/drivers/net/nfp/meson.build
@@ -18,7 +18,7 @@ sources = files(
         'nfpcore/nfp_mutex.c',
         'nfpcore/nfp_nsp_eth.c',
         'nfpcore/nfp_hwinfo.c',
-        'nfp_net.c',
+        'nfp_common.c',
         'nfp_rxtx.c',
         'nfp_cpp_bridge.c',
         'nfp_ethdev_vf.c',
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_common.c
similarity index 99%
rename from drivers/net/nfp/nfp_net.c
rename to drivers/net/nfp/nfp_common.c
index a6097eaab0..87e8f5f333 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_common.c
@@ -8,9 +8,9 @@
 /*
  * vim:shiftwidth=8:noexpandtab
  *
- * @file dpdk/pmd/nfp_net.c
+ * @file dpdk/pmd/nfp_common.c
  *
- * Netronome vNIC DPDK Poll-Mode Driver: Main entry point
+ * Netronome vNIC DPDK Poll-Mode Driver: Common files
  */
 
 #include <rte_byteorder.h>
diff --git a/drivers/net/nfp/nfp_net_pmd.h b/drivers/net/nfp/nfp_common.h
similarity index 99%
rename from drivers/net/nfp/nfp_net_pmd.h
rename to drivers/net/nfp/nfp_common.h
index dc05e888df..54ac937bd2 100644
--- a/drivers/net/nfp/nfp_net_pmd.h
+++ b/drivers/net/nfp/nfp_common.h
@@ -11,8 +11,8 @@
  * Netronome NFP_NET PMD driver
  */
 
-#ifndef _NFP_NET_PMD_H_
-#define _NFP_NET_PMD_H_
+#ifndef _NFP_COMMON_H_
+#define _NFP_COMMON_H_
 
 #define NFP_NET_PMD_VERSION "0.1"
 #define PCI_VENDOR_ID_NETRONOME         0x19ee
@@ -404,7 +404,7 @@ int nfp_net_rss_config_default(struct rte_eth_dev *dev);
 #define NFP_NET_DEV_PRIVATE_TO_PF(dev_priv)\
 	(((struct nfp_net_hw *)dev_priv)->pf_dev)
 
-#endif /* _NFP_NET_PMD_H_ */
+#endif /* _NFP_COMMON_H_ */
 /*
  * Local variables:
  * c-file-style: "Linux"
diff --git a/drivers/net/nfp/nfp_cpp_bridge.c b/drivers/net/nfp/nfp_cpp_bridge.c
index d916793338..74a0eacb3f 100644
--- a/drivers/net/nfp/nfp_cpp_bridge.c
+++ b/drivers/net/nfp/nfp_cpp_bridge.c
@@ -19,7 +19,7 @@
 #include "nfpcore/nfp_mip.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_logs.h"
+#include "nfp_logs.h"
 #include "nfp_cpp_bridge.h"
 
 #include <sys/ioctl.h>
diff --git a/drivers/net/nfp/nfp_net_ctrl.h b/drivers/net/nfp/nfp_ctrl.h
similarity index 99%
rename from drivers/net/nfp/nfp_net_ctrl.h
rename to drivers/net/nfp/nfp_ctrl.h
index 4f26ccf483..4dd62ef194 100644
--- a/drivers/net/nfp/nfp_net_ctrl.h
+++ b/drivers/net/nfp/nfp_ctrl.h
@@ -8,8 +8,8 @@
  *
  * Netronome network device driver: Control BAR layout
  */
-#ifndef _NFP_NET_CTRL_H_
-#define _NFP_NET_CTRL_H_
+#ifndef _NFP_CTRL_H_
+#define _NFP_CTRL_H_
 
 /*
  * Configuration BAR size.
@@ -317,7 +317,7 @@
 /* PF multiport offset */
 #define NFP_PF_CSR_SLICE_SIZE	(32 * 1024)
 
-#endif /* _NFP_NET_CTRL_H_ */
+#endif /* _NFP_CTRL_H_ */
 /*
  * Local variables:
  * c-file-style: "Linux"
diff --git a/drivers/net/nfp/nfp_ethdev.c b/drivers/net/nfp/nfp_ethdev.c
index 056d962306..dbdac413ae 100644
--- a/drivers/net/nfp/nfp_ethdev.c
+++ b/drivers/net/nfp/nfp_ethdev.c
@@ -31,10 +31,10 @@
 #include "nfpcore/nfp_rtsym.h"
 #include "nfpcore/nfp_nsp.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 #include "nfp_cpp_bridge.h"
 
 
diff --git a/drivers/net/nfp/nfp_ethdev_vf.c b/drivers/net/nfp/nfp_ethdev_vf.c
index 223142c0ed..b697b55865 100644
--- a/drivers/net/nfp/nfp_ethdev_vf.c
+++ b/drivers/net/nfp/nfp_ethdev_vf.c
@@ -16,10 +16,10 @@
 #include "nfpcore/nfp_mip.h"
 #include "nfpcore/nfp_rtsym.h"
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 
 static void nfp_netvf_read_mac(struct nfp_net_hw *hw);
 static int nfp_netvf_start(struct rte_eth_dev *dev);
diff --git a/drivers/net/nfp/nfp_net_logs.h b/drivers/net/nfp/nfp_logs.h
similarity index 94%
rename from drivers/net/nfp/nfp_net_logs.h
rename to drivers/net/nfp/nfp_logs.h
index 27dd87611b..bd5a5e1ec5 100644
--- a/drivers/net/nfp/nfp_net_logs.h
+++ b/drivers/net/nfp/nfp_logs.h
@@ -3,8 +3,8 @@
  * All rights reserved.
  */
 
-#ifndef _NFP_NET_LOGS_H_
-#define _NFP_NET_LOGS_H_
+#ifndef _NFP_LOGS_H_
+#define _NFP_LOGS_H_
 
 #include <rte_log.h>
 
@@ -44,4 +44,4 @@ extern int nfp_logtype_driver;
 	rte_log(RTE_LOG_ ## level, nfp_logtype_driver, \
 		"%s(): " fmt "\n", __func__, ## args)
 
-#endif /* _NFP_NET_LOGS_H_ */
+#endif /* _NFP_LOGS_H_ */
diff --git a/drivers/net/nfp/nfp_rxtx.c b/drivers/net/nfp/nfp_rxtx.c
index 9ee9e5c9a3..1402c5f84a 100644
--- a/drivers/net/nfp/nfp_rxtx.c
+++ b/drivers/net/nfp/nfp_rxtx.c
@@ -16,10 +16,10 @@
 #include <ethdev_driver.h>
 #include <ethdev_pci.h>
 
-#include "nfp_net_pmd.h"
+#include "nfp_common.h"
 #include "nfp_rxtx.h"
-#include "nfp_net_logs.h"
-#include "nfp_net_ctrl.h"
+#include "nfp_logs.h"
+#include "nfp_ctrl.h"
 
 /* Prototypes */
 static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
-- 
2.30.1 (Apple Git-130)


^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD
  2021-07-23  9:18   ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Thomas Monjalon
@ 2021-07-29 13:48     ` Heinrich Kuhn
  0 siblings, 0 replies; 31+ messages in thread
From: Heinrich Kuhn @ 2021-07-29 13:48 UTC (permalink / raw)
  To: Thomas Monjalon, Simon Horman; +Cc: dev, david.marchand



On 2021/07/23 11:18, Thomas Monjalon wrote:
> 16/07/2021 10:35, Heinrich Kuhn:
>> This patch set restructures the NFP PMD, aligning it more with the
>> common layout adopted by most other PMD's. Although the changes look
>> fairly large, functionally nothing is added or removed from the driver
>> and the existing code is mostly just reorganized into the familiar
>> structure seen in other PMD's.
> 
> It seems this refactoring is destroying this firmware improvement:
> 	https://git.dpdk.org/dpdk/commit/?id=40edb9c0d36b781
> 
> Please rebase carefully, keeping the behaviour really intact.
> 
>> Apart form adopting the common PMD layout
>> this change should also aid in future feature development to the NFP
>> PMD. The previous layout where most of the logic resided in a single
>> file (nfp_net.c) would have become tedious to support going forward.
> 
> Thanks Thomas, I actually finished this work earlier in the month and
just missed your firmware helper addition. I posted v3 of this rework

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
                       ` (6 preceding siblings ...)
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
@ 2021-08-13 12:46     ` Heinrich Kuhn
  2021-08-13 14:47       ` Ferruh Yigit
  2021-08-17 16:29     ` Ferruh Yigit
  8 siblings, 1 reply; 31+ messages in thread
From: Heinrich Kuhn @ 2021-08-13 12:46 UTC (permalink / raw)
  To: dev



On 2021/07/29 15:47, Heinrich Kuhn wrote:
> This patch set restructures the NFP PMD, aligning it more with the
> common layout adopted by most other PMD's. Although the changes look
> fairly large, functionally nothing is added or removed from the driver
> and the existing code is mostly just reorganized into the familiar
> structure seen in other PMD's. Apart form adopting the common PMD layout
> this change should also aid in future feature development to the NFP
> PMD. The previous layout where most of the logic resided in a single
> file (nfp_net.c) would have become tedious to support going forward.
> 
> v3:
> * Avoid squashing the new firmware loader helper added in: 
>   https://git.dpdk.org/dpdk/commit/?id=40edb9c0d36b781
> * Add dependency to patch-93299
> 
> v2:
> * Added missing sign-off's
> 
> ---

I think this refactor is a step in the right direction for the NFP PMD.
I do have a question/concern regarding future bug fixes. If this is
merged, back-porting any bug fixes will require a little bit more effort
since the code base will differ quite substantially for some time.

If there is a strong preference to avoid a situation like this we can
certainly live without this refactor

Regards,
Heinrich

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD
  2021-08-13 12:46     ` [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD Heinrich Kuhn
@ 2021-08-13 14:47       ` Ferruh Yigit
  0 siblings, 0 replies; 31+ messages in thread
From: Ferruh Yigit @ 2021-08-13 14:47 UTC (permalink / raw)
  To: Heinrich Kuhn, dev; +Cc: Luca Boccassi, Christian Ehrhardt

On 8/13/2021 1:46 PM, Heinrich Kuhn wrote:
> 
> 
> On 2021/07/29 15:47, Heinrich Kuhn wrote:
>> This patch set restructures the NFP PMD, aligning it more with the
>> common layout adopted by most other PMD's. Although the changes look
>> fairly large, functionally nothing is added or removed from the driver
>> and the existing code is mostly just reorganized into the familiar
>> structure seen in other PMD's. Apart form adopting the common PMD layout
>> this change should also aid in future feature development to the NFP
>> PMD. The previous layout where most of the logic resided in a single
>> file (nfp_net.c) would have become tedious to support going forward.
>>
>> v3:
>> * Avoid squashing the new firmware loader helper added in: 
>>   https://git.dpdk.org/dpdk/commit/?id=40edb9c0d36b781
>> * Add dependency to patch-93299
>>
>> v2:
>> * Added missing sign-off's
>>
>> ---
> 
> I think this refactor is a step in the right direction for the NFP PMD.
> I do have a question/concern regarding future bug fixes. If this is
> merged, back-porting any bug fixes will require a little bit more effort
> since the code base will differ quite substantially for some time.
> 
> If there is a strong preference to avoid a situation like this we can
> certainly live without this refactor
> 

I didn't check the set yet, but as generic rule when there is a conflict in
backporting, stable tree maintainers ask support from patch authors, for this
case from you.
If you are OK to provide this support when needed, there should be no problem.

We have rejected only a few refactoring patches in the past, they were doing
batch syntax updates or batch renames etc.. The benefit was small comparing the
conflict/noise it brings. But I am for having meaningful refactoring.

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [dpdk-dev] [PATCH v3 4/7] net/nfp: prototype common functions in header file
  2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
@ 2021-08-17 14:44       ` Ferruh Yigit
  2021-08-17 16:29         ` Ferruh Yigit
  0 siblings, 1 reply; 31+ messages in thread
From: Ferruh Yigit @ 2021-08-17 14:44 UTC (permalink / raw)
  To: Heinrich Kuhn, dev; +Cc: Simon Horman

On 7/29/2021 2:47 PM, Heinrich Kuhn wrote:
> The majority of "ethdev" type functions are used for both PF devices and
> VF devices. Prototype these functions in the nfp_net_pmd header file in
> preparation of splitting PF and VF specific functions.
> 
> Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
> Signed-off-by: Simon Horman <simon.horman@corigine.com>
> ---
>  drivers/net/nfp/nfp_net.c     | 87 +++++++++++++----------------------
>  drivers/net/nfp/nfp_net_pmd.h | 49 ++++++++++++++++++++
>  2 files changed, 81 insertions(+), 55 deletions(-)
> 

<...>


Following functions seems only used in this file, and you may want to keep them
'static':
nfp_net_dev_link_status_print
nfp_net_irq_unmask
nfp_net_dev_interrupt_delayed_handler
nfp_net_rss_reta_write
nfp_net_rss_hash_write

(I may be missed some, you may want to double check all.)

Rest of the set looks good to me.

Thanks,
ferruh

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [dpdk-dev] [PATCH v3 4/7] net/nfp: prototype common functions in header file
  2021-08-17 14:44       ` Ferruh Yigit
@ 2021-08-17 16:29         ` Ferruh Yigit
  0 siblings, 0 replies; 31+ messages in thread
From: Ferruh Yigit @ 2021-08-17 16:29 UTC (permalink / raw)
  To: Heinrich Kuhn, dev; +Cc: Simon Horman

On 8/17/2021 3:44 PM, Ferruh Yigit wrote:
> On 7/29/2021 2:47 PM, Heinrich Kuhn wrote:
>> The majority of "ethdev" type functions are used for both PF devices and
>> VF devices. Prototype these functions in the nfp_net_pmd header file in
>> preparation of splitting PF and VF specific functions.
>>
>> Signed-off-by: Heinrich Kuhn <heinrich.kuhn@netronome.com>
>> Signed-off-by: Simon Horman <simon.horman@corigine.com>
>> ---
>>  drivers/net/nfp/nfp_net.c     | 87 +++++++++++++----------------------
>>  drivers/net/nfp/nfp_net_pmd.h | 49 ++++++++++++++++++++
>>  2 files changed, 81 insertions(+), 55 deletions(-)
>>
> 
> <...>
> 
> 
> Following functions seems only used in this file, and you may want to keep them
> 'static':
> nfp_net_dev_link_status_print
> nfp_net_irq_unmask
> nfp_net_dev_interrupt_delayed_handler
> nfp_net_rss_reta_write
> nfp_net_rss_hash_write
> 
> (I may be missed some, you may want to double check all.)
> 
> Rest of the set looks good to me.
> 

Since rest is OK, I will make above functions static while merging, if they
needs to be used by other files they can always be updated again.

^ permalink raw reply	[flat|nested] 31+ messages in thread

* Re: [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD
  2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
                       ` (7 preceding siblings ...)
  2021-08-13 12:46     ` [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD Heinrich Kuhn
@ 2021-08-17 16:29     ` Ferruh Yigit
  8 siblings, 0 replies; 31+ messages in thread
From: Ferruh Yigit @ 2021-08-17 16:29 UTC (permalink / raw)
  To: Heinrich Kuhn, dev

On 7/29/2021 2:47 PM, Heinrich Kuhn wrote:
> This patch set restructures the NFP PMD, aligning it more with the
> common layout adopted by most other PMD's. Although the changes look
> fairly large, functionally nothing is added or removed from the driver
> and the existing code is mostly just reorganized into the familiar
> structure seen in other PMD's. Apart form adopting the common PMD layout
> this change should also aid in future feature development to the NFP
> PMD. The previous layout where most of the logic resided in a single
> file (nfp_net.c) would have become tedious to support going forward.
> 
> v3:
> * Avoid squashing the new firmware loader helper added in: 
>   https://git.dpdk.org/dpdk/commit/?id=40edb9c0d36b781
> * Add dependency to patch-93299
> 
> v2:
> * Added missing sign-off's
> 
> ---
> Depends-on: patch-93299 ("net/nfp: remove compile time log")
> 
> Heinrich Kuhn (7):
>   net/nfp: split rxtx headers into separate file
>   net/nfp: move rxtx functions to their own file
>   net/nfp: move CPP bridge to a separate file
>   net/nfp: prototype common functions in header file
>   net/nfp: move VF functions into new file
>   net/nfp: move PF functions into new file
>   net/nfp: batch file rename for consistency
> 

Series applied to dpdk-next-net/main, thanks.


^ permalink raw reply	[flat|nested] 31+ messages in thread

end of thread, other threads:[~2021-08-17 16:30 UTC | newest]

Thread overview: 31+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-16  8:23 [dpdk-dev] [PATCH 0/7] Refactor the NFP PMD Heinrich Kuhn
2021-07-16  8:23 ` [dpdk-dev] [PATCH 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
2021-07-16  8:23 ` [dpdk-dev] [PATCH 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
2021-07-16  8:23 ` [dpdk-dev] [PATCH 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
2021-07-16  8:23 ` [dpdk-dev] [PATCH 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
2021-07-16  8:23 ` [dpdk-dev] [PATCH 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
2021-07-16  8:23 ` [dpdk-dev] [PATCH 6/7] net/nfp: move PF " Heinrich Kuhn
2021-07-16  8:23 ` [dpdk-dev] [PATCH 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
2021-07-16  8:35 ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Heinrich Kuhn
2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 6/7] net/nfp: move PF " Heinrich Kuhn
2021-07-16  8:35   ` [dpdk-dev] [PATCH v2 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
2021-07-23  9:18   ` [dpdk-dev] [PATCH v2 0/7] Refactor the NFP PMD Thomas Monjalon
2021-07-29 13:48     ` Heinrich Kuhn
2021-07-29 13:47   ` [dpdk-dev] [PATCH v3 " Heinrich Kuhn
2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 1/7] net/nfp: split rxtx headers into separate file Heinrich Kuhn
2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 2/7] net/nfp: move rxtx functions to their own file Heinrich Kuhn
2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 3/7] net/nfp: move CPP bridge to a separate file Heinrich Kuhn
2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 4/7] net/nfp: prototype common functions in header file Heinrich Kuhn
2021-08-17 14:44       ` Ferruh Yigit
2021-08-17 16:29         ` Ferruh Yigit
2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 5/7] net/nfp: move VF functions into new file Heinrich Kuhn
2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 6/7] net/nfp: move PF " Heinrich Kuhn
2021-07-29 13:47     ` [dpdk-dev] [PATCH v3 7/7] net/nfp: batch file rename for consistency Heinrich Kuhn
2021-08-13 12:46     ` [dpdk-dev] [PATCH v3 0/7] Refactor the NFP PMD Heinrich Kuhn
2021-08-13 14:47       ` Ferruh Yigit
2021-08-17 16:29     ` Ferruh Yigit

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).