net/dpaa2: support more than 16 burst size in Rx
[dpdk.git] / drivers / net / dpaa2 / dpaa2_rxtx.c
index c5d49cb..97d0f45 100644 (file)
@@ -1,34 +1,8 @@
-/*-
- *   BSD LICENSE
+/* SPDX-License-Identifier: BSD-3-Clause
  *
  *   Copyright (c) 2016 Freescale Semiconductor, Inc. All rights reserved.
- *   Copyright (c) 2016 NXP. All rights reserved.
+ *   Copyright 2016 NXP
  *
- *   Redistribution and use in source and binary forms, with or without
- *   modification, are permitted provided that the following conditions
- *   are met:
- *
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in
- *       the documentation and/or other materials provided with the
- *       distribution.
- *     * Neither the name of Freescale Semiconductor, Inc nor the names of its
- *       contributors may be used to endorse or promote products derived
- *       from this software without specific prior written permission.
- *
- *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
 #include <time.h>
 #include <rte_memcpy.h>
 #include <rte_string_fns.h>
 #include <rte_dev.h>
-#include <rte_ethdev.h>
 
+#include <rte_fslmc.h>
 #include <fslmc_logs.h>
 #include <fslmc_vfio.h>
 #include <dpaa2_hw_pvt.h>
 #include <dpaa2_hw_dpio.h>
 #include <dpaa2_hw_mempool.h>
+#include <dpaa2_eventdev.h>
 
 #include "dpaa2_ethdev.h"
 #include "base/dpaa2_hw_dpni_annot.h"
 
+#define DPAA2_MBUF_TO_CONTIG_FD(_mbuf, _fd, _bpid)  do { \
+       DPAA2_SET_FD_ADDR(_fd, DPAA2_MBUF_VADDR_TO_IOVA(_mbuf)); \
+       DPAA2_SET_FD_LEN(_fd, _mbuf->data_len); \
+       DPAA2_SET_ONLY_FD_BPID(_fd, _bpid); \
+       DPAA2_SET_FD_OFFSET(_fd, _mbuf->data_off); \
+       DPAA2_SET_FD_ASAL(_fd, DPAA2_ASAL_VAL); \
+} while (0)
+
+static inline void __attribute__((hot))
+dpaa2_dev_rx_parse_frc(struct rte_mbuf *m, uint16_t frc)
+{
+       PMD_RX_LOG(DEBUG, "frc = 0x%x   ", frc);
+
+       m->packet_type = RTE_PTYPE_UNKNOWN;
+       switch (frc) {
+       case DPAA2_PKT_TYPE_ETHER:
+               m->packet_type = RTE_PTYPE_L2_ETHER;
+               break;
+       case DPAA2_PKT_TYPE_IPV4:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4;
+               break;
+       case DPAA2_PKT_TYPE_IPV6:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6;
+               break;
+       case DPAA2_PKT_TYPE_IPV4_EXT:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4_EXT;
+               break;
+       case DPAA2_PKT_TYPE_IPV6_EXT:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6_EXT;
+               break;
+       case DPAA2_PKT_TYPE_IPV4_TCP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_TCP;
+               break;
+       case DPAA2_PKT_TYPE_IPV6_TCP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_TCP;
+               break;
+       case DPAA2_PKT_TYPE_IPV4_UDP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_UDP;
+               break;
+       case DPAA2_PKT_TYPE_IPV6_UDP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_UDP;
+               break;
+       case DPAA2_PKT_TYPE_IPV4_SCTP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_SCTP;
+               break;
+       case DPAA2_PKT_TYPE_IPV6_SCTP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_SCTP;
+               break;
+       case DPAA2_PKT_TYPE_IPV4_ICMP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_ICMP;
+               break;
+       case DPAA2_PKT_TYPE_IPV6_ICMP:
+               m->packet_type = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_ICMP;
+               break;
+       case DPAA2_PKT_TYPE_VLAN_1:
+       case DPAA2_PKT_TYPE_VLAN_2:
+               m->ol_flags |= PKT_RX_VLAN;
+               break;
+       /* More switch cases can be added */
+       /* TODO: Add handling for checksum error check from FRC */
+       default:
+               m->packet_type = RTE_PTYPE_UNKNOWN;
+       }
+}
+
 static inline uint32_t __attribute__((hot))
-dpaa2_dev_rx_parse(uint64_t hw_annot_addr)
+dpaa2_dev_rx_parse_slow(uint64_t hw_annot_addr)
 {
        uint32_t pkt_type = RTE_PTYPE_UNKNOWN;
        struct dpaa2_annot_hdr *annotation =
                        (struct dpaa2_annot_hdr *)hw_annot_addr;
 
        PMD_RX_LOG(DEBUG, "annotation = 0x%lx   ", annotation->word4);
-
        if (BIT_ISSET_AT_POS(annotation->word3, L2_ARP_PRESENT)) {
                pkt_type = RTE_PTYPE_L2_ETHER_ARP;
                goto parse_done;
@@ -115,6 +166,41 @@ parse_done:
        return pkt_type;
 }
 
+
+static inline uint32_t __attribute__((hot))
+dpaa2_dev_rx_parse(uint64_t hw_annot_addr)
+{
+       struct dpaa2_annot_hdr *annotation =
+                       (struct dpaa2_annot_hdr *)hw_annot_addr;
+
+       PMD_RX_LOG(DEBUG, "annotation = 0x%lx   ", annotation->word4);
+
+       /* Return some common types from parse processing */
+       switch (annotation->word4) {
+       case DPAA2_L3_IPv4:
+               return RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4;
+       case DPAA2_L3_IPv6:
+               return  RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6;
+       case DPAA2_L3_IPv4_TCP:
+               return  RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4 |
+                               RTE_PTYPE_L4_TCP;
+       case DPAA2_L3_IPv4_UDP:
+               return  RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4 |
+                               RTE_PTYPE_L4_UDP;
+       case DPAA2_L3_IPv6_TCP:
+               return  RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6 |
+                               RTE_PTYPE_L4_TCP;
+       case DPAA2_L3_IPv6_UDP:
+               return  RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6 |
+                               RTE_PTYPE_L4_UDP;
+       default:
+               PMD_RX_LOG(DEBUG, "Slow parse the parsing results\n");
+               break;
+       }
+
+       return dpaa2_dev_rx_parse_slow(hw_annot_addr);
+}
+
 static inline void __attribute__((hot))
 dpaa2_dev_rx_offload(uint64_t hw_annot_addr, struct rte_mbuf *mbuf)
 {
@@ -123,7 +209,7 @@ dpaa2_dev_rx_offload(uint64_t hw_annot_addr, struct rte_mbuf *mbuf)
 
        if (BIT_ISSET_AT_POS(annotation->word3,
                             L2_VLAN_1_PRESENT | L2_VLAN_N_PRESENT))
-               mbuf->ol_flags |= PKT_RX_VLAN_PKT;
+               mbuf->ol_flags |= PKT_RX_VLAN;
 
        if (BIT_ISSET_AT_POS(annotation->word8, DPAA2_ETH_FAS_L3CE))
                mbuf->ol_flags |= PKT_RX_IP_CKSUM_BAD;
@@ -132,6 +218,70 @@ dpaa2_dev_rx_offload(uint64_t hw_annot_addr, struct rte_mbuf *mbuf)
                mbuf->ol_flags |= PKT_RX_L4_CKSUM_BAD;
 }
 
+static inline struct rte_mbuf *__attribute__((hot))
+eth_sg_fd_to_mbuf(const struct qbman_fd *fd)
+{
+       struct qbman_sge *sgt, *sge;
+       dma_addr_t sg_addr;
+       int i = 0;
+       uint64_t fd_addr;
+       struct rte_mbuf *first_seg, *next_seg, *cur_seg, *temp;
+
+       fd_addr = (uint64_t)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd));
+
+       /* Get Scatter gather table address */
+       sgt = (struct qbman_sge *)(fd_addr + DPAA2_GET_FD_OFFSET(fd));
+
+       sge = &sgt[i++];
+       sg_addr = (uint64_t)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FLE_ADDR(sge));
+
+       /* First Scatter gather entry */
+       first_seg = DPAA2_INLINE_MBUF_FROM_BUF(sg_addr,
+               rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size);
+       /* Prepare all the metadata for first segment */
+       first_seg->buf_addr = (uint8_t *)sg_addr;
+       first_seg->ol_flags = 0;
+       first_seg->data_off = DPAA2_GET_FLE_OFFSET(sge);
+       first_seg->data_len = sge->length  & 0x1FFFF;
+       first_seg->pkt_len = DPAA2_GET_FD_LEN(fd);
+       first_seg->nb_segs = 1;
+       first_seg->next = NULL;
+       if (dpaa2_svr_family == SVR_LX2160A)
+               dpaa2_dev_rx_parse_frc(first_seg,
+                               DPAA2_GET_FD_FRC_PARSE_SUM(fd));
+       else {
+               first_seg->packet_type = dpaa2_dev_rx_parse(
+                        (uint64_t)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd))
+                        + DPAA2_FD_PTA_SIZE);
+               dpaa2_dev_rx_offload((uint64_t)DPAA2_IOVA_TO_VADDR(
+                       DPAA2_GET_FD_ADDR(fd)) +
+                       DPAA2_FD_PTA_SIZE, first_seg);
+       }
+       rte_mbuf_refcnt_set(first_seg, 1);
+       cur_seg = first_seg;
+       while (!DPAA2_SG_IS_FINAL(sge)) {
+               sge = &sgt[i++];
+               sg_addr = (uint64_t)DPAA2_IOVA_TO_VADDR(
+                               DPAA2_GET_FLE_ADDR(sge));
+               next_seg = DPAA2_INLINE_MBUF_FROM_BUF(sg_addr,
+                       rte_dpaa2_bpid_info[DPAA2_GET_FLE_BPID(sge)].meta_data_size);
+               next_seg->buf_addr  = (uint8_t *)sg_addr;
+               next_seg->data_off  = DPAA2_GET_FLE_OFFSET(sge);
+               next_seg->data_len  = sge->length  & 0x1FFFF;
+               first_seg->nb_segs += 1;
+               rte_mbuf_refcnt_set(next_seg, 1);
+               cur_seg->next = next_seg;
+               next_seg->next = NULL;
+               cur_seg = next_seg;
+       }
+       temp = DPAA2_INLINE_MBUF_FROM_BUF(fd_addr,
+               rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size);
+       rte_mbuf_refcnt_set(temp, 1);
+       rte_pktmbuf_free_seg(temp);
+
+       return (void *)first_seg;
+}
+
 static inline struct rte_mbuf *__attribute__((hot))
 eth_fd_to_mbuf(const struct qbman_fd *fd)
 {
@@ -147,19 +297,25 @@ eth_fd_to_mbuf(const struct qbman_fd *fd)
        mbuf->data_off = DPAA2_GET_FD_OFFSET(fd);
        mbuf->data_len = DPAA2_GET_FD_LEN(fd);
        mbuf->pkt_len = mbuf->data_len;
+       mbuf->next = NULL;
+       rte_mbuf_refcnt_set(mbuf, 1);
 
        /* Parse the packet */
-       /* parse results are after the private - sw annotation area */
-       mbuf->packet_type = dpaa2_dev_rx_parse(
+       /* parse results for LX2 are there in FRC field of FD.
+        * For other DPAA2 platforms , parse results are after
+        * the private - sw annotation area
+        */
+
+       if (dpaa2_svr_family == SVR_LX2160A)
+               dpaa2_dev_rx_parse_frc(mbuf, DPAA2_GET_FD_FRC_PARSE_SUM(fd));
+       else {
+               mbuf->packet_type = dpaa2_dev_rx_parse(
                        (uint64_t)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd))
                         + DPAA2_FD_PTA_SIZE);
-
-       dpaa2_dev_rx_offload((uint64_t)DPAA2_IOVA_TO_VADDR(
+               dpaa2_dev_rx_offload((uint64_t)DPAA2_IOVA_TO_VADDR(
                             DPAA2_GET_FD_ADDR(fd)) +
                             DPAA2_FD_PTA_SIZE, mbuf);
-
-       mbuf->next = NULL;
-       rte_mbuf_refcnt_set(mbuf, 1);
+       }
 
        PMD_RX_LOG(DEBUG, "to mbuf - mbuf =%p, mbuf->buf_addr =%p, off = %d,"
                "fd_off=%d fd =%lx, meta = %d  bpid =%d, len=%d\n",
@@ -171,18 +327,100 @@ eth_fd_to_mbuf(const struct qbman_fd *fd)
        return mbuf;
 }
 
+static int __attribute__ ((noinline)) __attribute__((hot))
+eth_mbuf_to_sg_fd(struct rte_mbuf *mbuf,
+                 struct qbman_fd *fd, uint16_t bpid)
+{
+       struct rte_mbuf *cur_seg = mbuf, *prev_seg, *mi, *temp;
+       struct qbman_sge *sgt, *sge = NULL;
+       int i;
+
+       /* First Prepare FD to be transmited*/
+       /* Resetting the buffer pool id and offset field*/
+       fd->simple.bpid_offset = 0;
+
+       if (unlikely(mbuf->ol_flags & PKT_TX_VLAN_PKT)) {
+               int ret = rte_vlan_insert(&mbuf);
+               if (ret)
+                       return ret;
+       }
+
+       temp = rte_pktmbuf_alloc(mbuf->pool);
+       if (temp == NULL) {
+               PMD_TX_LOG(ERR, "No memory to allocate S/G table");
+               return -ENOMEM;
+       }
+
+       DPAA2_SET_FD_ADDR(fd, DPAA2_MBUF_VADDR_TO_IOVA(temp));
+       DPAA2_SET_FD_LEN(fd, mbuf->pkt_len);
+       DPAA2_SET_FD_OFFSET(fd, temp->data_off);
+       DPAA2_SET_FD_BPID(fd, bpid);
+       DPAA2_SET_FD_ASAL(fd, DPAA2_ASAL_VAL);
+       DPAA2_FD_SET_FORMAT(fd, qbman_fd_sg);
+       /*Set Scatter gather table and Scatter gather entries*/
+       sgt = (struct qbman_sge *)(
+                       (uint64_t)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd))
+                       + DPAA2_GET_FD_OFFSET(fd));
+
+       for (i = 0; i < mbuf->nb_segs; i++) {
+               sge = &sgt[i];
+               /*Resetting the buffer pool id and offset field*/
+               sge->fin_bpid_offset = 0;
+               DPAA2_SET_FLE_ADDR(sge, DPAA2_MBUF_VADDR_TO_IOVA(cur_seg));
+               DPAA2_SET_FLE_OFFSET(sge, cur_seg->data_off);
+               sge->length = cur_seg->data_len;
+               if (RTE_MBUF_DIRECT(cur_seg)) {
+                       if (rte_mbuf_refcnt_read(cur_seg) > 1) {
+                               /* If refcnt > 1, invalid bpid is set to ensure
+                                * buffer is not freed by HW
+                                */
+                               DPAA2_SET_FLE_IVP(sge);
+                               rte_mbuf_refcnt_update(cur_seg, -1);
+                       } else
+                               DPAA2_SET_FLE_BPID(sge,
+                                               mempool_to_bpid(cur_seg->pool));
+                       cur_seg = cur_seg->next;
+               } else {
+                       /* Get owner MBUF from indirect buffer */
+                       mi = rte_mbuf_from_indirect(cur_seg);
+                       if (rte_mbuf_refcnt_read(mi) > 1) {
+                               /* If refcnt > 1, invalid bpid is set to ensure
+                                * owner buffer is not freed by HW
+                                */
+                               DPAA2_SET_FLE_IVP(sge);
+                       } else {
+                               DPAA2_SET_FLE_BPID(sge,
+                                                  mempool_to_bpid(mi->pool));
+                               rte_mbuf_refcnt_update(mi, 1);
+                       }
+                       prev_seg = cur_seg;
+                       cur_seg = cur_seg->next;
+                       prev_seg->next = NULL;
+                       rte_pktmbuf_free(prev_seg);
+               }
+       }
+       DPAA2_SG_SET_FINAL(sge, true);
+       return 0;
+}
+
+static void
+eth_mbuf_to_fd(struct rte_mbuf *mbuf,
+              struct qbman_fd *fd, uint16_t bpid) __attribute__((unused));
+
 static void __attribute__ ((noinline)) __attribute__((hot))
 eth_mbuf_to_fd(struct rte_mbuf *mbuf,
               struct qbman_fd *fd, uint16_t bpid)
 {
+       if (unlikely(mbuf->ol_flags & PKT_TX_VLAN_PKT)) {
+               if (rte_vlan_insert(&mbuf)) {
+                       rte_pktmbuf_free(mbuf);
+                       return;
+               }
+       }
        /*Resetting the buffer pool id and offset field*/
        fd->simple.bpid_offset = 0;
 
-       DPAA2_SET_FD_ADDR(fd, DPAA2_MBUF_VADDR_TO_IOVA(mbuf));
-       DPAA2_SET_FD_LEN(fd, mbuf->data_len);
-       DPAA2_SET_FD_BPID(fd, bpid);
-       DPAA2_SET_FD_OFFSET(fd, mbuf->data_off);
-       DPAA2_SET_FD_ASAL(fd, DPAA2_ASAL_VAL);
+       DPAA2_MBUF_TO_CONTIG_FD(mbuf, fd, bpid);
 
        PMD_TX_LOG(DEBUG, "mbuf =%p, mbuf->buf_addr =%p, off = %d,"
                "fd_off=%d fd =%lx, meta = %d  bpid =%d, len=%d\n",
@@ -190,8 +428,22 @@ eth_mbuf_to_fd(struct rte_mbuf *mbuf,
                DPAA2_GET_FD_OFFSET(fd), DPAA2_GET_FD_ADDR(fd),
                rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size,
                DPAA2_GET_FD_BPID(fd), DPAA2_GET_FD_LEN(fd));
-}
+       if (RTE_MBUF_DIRECT(mbuf)) {
+               if (rte_mbuf_refcnt_read(mbuf) > 1) {
+                       DPAA2_SET_FD_IVP(fd);
+                       rte_mbuf_refcnt_update(mbuf, -1);
+               }
+       } else {
+               struct rte_mbuf *mi;
 
+               mi = rte_mbuf_from_indirect(mbuf);
+               if (rte_mbuf_refcnt_read(mi) > 1)
+                       DPAA2_SET_FD_IVP(fd);
+               else
+                       rte_mbuf_refcnt_update(mi, 1);
+               rte_pktmbuf_free(mbuf);
+       }
+}
 
 static inline int __attribute__((hot))
 eth_copy_mbuf_to_fd(struct rte_mbuf *mbuf,
@@ -200,10 +452,15 @@ eth_copy_mbuf_to_fd(struct rte_mbuf *mbuf,
        struct rte_mbuf *m;
        void *mb = NULL;
 
+       if (unlikely(mbuf->ol_flags & PKT_TX_VLAN_PKT)) {
+               int ret = rte_vlan_insert(&mbuf);
+               if (ret)
+                       return ret;
+       }
+
        if (rte_dpaa2_mbuf_alloc_bulk(
                rte_dpaa2_bpid_info[bpid].bp_list->mp, &mb, 1)) {
                PMD_TX_LOG(WARNING, "Unable to allocated DPAA2 buffer");
-               rte_pktmbuf_free(mbuf);
                return -1;
        }
        m = (struct rte_mbuf *)mb;
@@ -220,11 +477,7 @@ eth_copy_mbuf_to_fd(struct rte_mbuf *mbuf,
        /*Resetting the buffer pool id and offset field*/
        fd->simple.bpid_offset = 0;
 
-       DPAA2_SET_FD_ADDR(fd, DPAA2_MBUF_VADDR_TO_IOVA(m));
-       DPAA2_SET_FD_LEN(fd, mbuf->data_len);
-       DPAA2_SET_FD_BPID(fd, bpid);
-       DPAA2_SET_FD_OFFSET(fd, mbuf->data_off);
-       DPAA2_SET_FD_ASAL(fd, DPAA2_ASAL_VAL);
+       DPAA2_MBUF_TO_CONTIG_FD(m, fd, bpid);
 
        PMD_TX_LOG(DEBUG, " mbuf %p BMAN buf addr %p",
                   (void *)mbuf, mbuf->buf_addr);
@@ -235,24 +488,23 @@ eth_copy_mbuf_to_fd(struct rte_mbuf *mbuf,
                rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size,
                DPAA2_GET_FD_OFFSET(fd),
                DPAA2_GET_FD_LEN(fd));
-       /*free the original packet */
-       rte_pktmbuf_free(mbuf);
 
        return 0;
 }
 
 uint16_t
-dpaa2_dev_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+dpaa2_dev_prefetch_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
 {
-       /* Function is responsible to receive frames for a given device and VQ*/
+       /* Function receive frames for a given device and VQ*/
        struct dpaa2_queue *dpaa2_q = (struct dpaa2_queue *)queue;
-       struct qbman_result *dq_storage;
+       struct qbman_result *dq_storage, *dq_storage1 = 0;
        uint32_t fqid = dpaa2_q->fqid;
-       int ret, num_rx = 0;
-       uint8_t is_last = 0, status;
+       int ret, num_rx = 0, next_pull = 0, num_pulled, num_to_pull;
+       uint8_t pending, is_repeat, status;
        struct qbman_swp *swp;
-       const struct qbman_fd *fd;
+       const struct qbman_fd *fd, *next_fd;
        struct qbman_pull_desc pulldesc;
+       struct queue_storage_info_t *q_storage = dpaa2_q->q_storage;
        struct rte_eth_dev *dev = dpaa2_q->dev;
 
        if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
@@ -263,78 +515,228 @@ dpaa2_dev_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
                }
        }
        swp = DPAA2_PER_LCORE_PORTAL;
-       dq_storage = dpaa2_q->q_storage->dq_storage[0];
 
-       qbman_pull_desc_clear(&pulldesc);
-       qbman_pull_desc_set_numframes(&pulldesc,
-                                     (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
-                                      DPAA2_DQRR_RING_SIZE : nb_pkts);
-       qbman_pull_desc_set_fq(&pulldesc, fqid);
-       /* todo optimization - we can have dq_storage_phys available*/
-       qbman_pull_desc_set_storage(&pulldesc, dq_storage,
-                       (dma_addr_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
+       /* if the original request for this q was from another portal */
+       if (unlikely(DPAA2_PER_LCORE_DPIO->index !=
+               q_storage->active_dpio_id)) {
+               if (check_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index)) {
+                       while (!qbman_check_command_complete(get_swp_active_dqs
+                               (DPAA2_PER_LCORE_DPIO->index)))
+                               ;
+                       clear_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index);
+               }
+               q_storage->active_dpio_id = DPAA2_PER_LCORE_DPIO->index;
+       }
 
-       /*Issue a volatile dequeue command. */
-       while (1) {
-               if (qbman_swp_pull(swp, &pulldesc)) {
-                       PMD_RX_LOG(ERR, "VDQ command is not issued."
-                                  "QBMAN is busy\n");
-                       /* Portal was busy, try again */
-                       continue;
+       if (unlikely(!q_storage->active_dqs)) {
+               q_storage->toggle = 0;
+               dq_storage = q_storage->dq_storage[q_storage->toggle];
+               q_storage->last_num_pkts = (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
+                                              DPAA2_DQRR_RING_SIZE : nb_pkts;
+               qbman_pull_desc_clear(&pulldesc);
+               qbman_pull_desc_set_numframes(&pulldesc,
+                                             q_storage->last_num_pkts);
+               qbman_pull_desc_set_fq(&pulldesc, fqid);
+               qbman_pull_desc_set_storage(&pulldesc, dq_storage,
+                       (dma_addr_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
+               while (1) {
+                       if (qbman_swp_pull(swp, &pulldesc)) {
+                               PMD_RX_LOG(WARNING,
+                                       "VDQ command not issued.QBMAN busy\n");
+                               /* Portal was busy, try again */
+                               continue;
+                       }
+                       break;
                }
-               break;
-       };
+               q_storage->active_dqs = dq_storage;
+               set_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index, dq_storage);
+       }
 
-       /* Receive the packets till Last Dequeue entry is found with
-        * respect to the above issues PULL command.
+       /* pkt to pull in current pull request */
+       num_to_pull = q_storage->last_num_pkts;
+
+       /* Number of packet requested is more than current pull request */
+       if (nb_pkts > num_to_pull)
+               next_pull = nb_pkts - num_to_pull;
+
+       dq_storage = q_storage->active_dqs;
+       /* Check if the previous issued command is completed.
+        * Also seems like the SWP is shared between the Ethernet Driver
+        * and the SEC driver.
         */
-       while (!is_last) {
-               struct rte_mbuf *mbuf;
-               /*Check if the previous issued command is completed.
-                * Also seems like the SWP is shared between the
-                * Ethernet Driver and the SEC driver.
-                */
-               while (!qbman_check_command_complete(swp, dq_storage))
-                       ;
+       while (!qbman_check_command_complete(dq_storage))
+               ;
+       if (dq_storage == get_swp_active_dqs(q_storage->active_dpio_id))
+               clear_swp_active_dqs(q_storage->active_dpio_id);
+
+repeat:
+       is_repeat = 0;
+
+       /* issue the deq command one more time to get another set of packets */
+       if (next_pull) {
+               q_storage->toggle ^= 1;
+               dq_storage1 = q_storage->dq_storage[q_storage->toggle];
+               qbman_pull_desc_clear(&pulldesc);
+
+               if (next_pull > DPAA2_DQRR_RING_SIZE) {
+                       qbman_pull_desc_set_numframes(&pulldesc,
+                                       DPAA2_DQRR_RING_SIZE);
+                       next_pull = next_pull - DPAA2_DQRR_RING_SIZE;
+                       q_storage->last_num_pkts = DPAA2_DQRR_RING_SIZE;
+               } else {
+                       qbman_pull_desc_set_numframes(&pulldesc, next_pull);
+                       q_storage->last_num_pkts = next_pull;
+                       next_pull = 0;
+               }
+               qbman_pull_desc_set_fq(&pulldesc, fqid);
+               qbman_pull_desc_set_storage(&pulldesc, dq_storage1,
+                       (dma_addr_t)(DPAA2_VADDR_TO_IOVA(dq_storage1)), 1);
+               while (1) {
+                       if (qbman_swp_pull(swp, &pulldesc)) {
+                               PMD_RX_LOG(WARNING,
+                                       "VDQ command not issued.QBMAN busy\n");
+                               /* Portal was busy, try again */
+                               continue;
+                       }
+                       break;
+               }
+               is_repeat = 1;
+               q_storage->active_dqs = dq_storage1;
+               set_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index, dq_storage1);
+       }
+
+       rte_prefetch0((void *)((uint64_t)(dq_storage + 1)));
+
+       num_pulled = 0;
+       pending = 1;
+
+       do {
                /* Loop until the dq_storage is updated with
                 * new token by QBMAN
                 */
-               while (!qbman_result_has_new_result(swp, dq_storage))
+               while (!qbman_check_new_result(dq_storage))
                        ;
+               rte_prefetch0((void *)((uint64_t)(dq_storage + 2)));
                /* Check whether Last Pull command is Expired and
                 * setting Condition for Loop termination
                 */
                if (qbman_result_DQ_is_pull_complete(dq_storage)) {
-                       is_last = 1;
+                       pending = 0;
                        /* Check for valid frame. */
-                       status = (uint8_t)qbman_result_DQ_flags(dq_storage);
+                       status = qbman_result_DQ_flags(dq_storage);
                        if (unlikely((status & QBMAN_DQ_STAT_VALIDFRAME) == 0))
                                continue;
                }
-
                fd = qbman_result_DQ_fd(dq_storage);
-               mbuf = (struct rte_mbuf *)DPAA2_IOVA_TO_VADDR(
-                  DPAA2_GET_FD_ADDR(fd)
-                  - rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size);
-               /* Prefeth mbuf */
-               rte_prefetch0(mbuf);
+
+               next_fd = qbman_result_DQ_fd(dq_storage + 1);
                /* Prefetch Annotation address for the parse results */
-               rte_prefetch0((void *)((uint64_t)DPAA2_GET_FD_ADDR(fd)
-                                               + DPAA2_FD_PTA_SIZE + 16));
+               rte_prefetch0((void *)(DPAA2_GET_FD_ADDR(next_fd)
+                               + DPAA2_FD_PTA_SIZE + 16));
 
-               bufs[num_rx] = eth_fd_to_mbuf(fd);
+               if (unlikely(DPAA2_FD_GET_FORMAT(fd) == qbman_fd_sg))
+                       bufs[num_rx] = eth_sg_fd_to_mbuf(fd);
+               else
+                       bufs[num_rx] = eth_fd_to_mbuf(fd);
                bufs[num_rx]->port = dev->data->port_id;
 
-               num_rx++;
+               if (dev->data->dev_conf.rxmode.hw_vlan_strip)
+                       rte_vlan_strip(bufs[num_rx]);
+
                dq_storage++;
-       } /* End of Packet Rx loop */
+               num_rx++;
+               num_pulled++;
+       } while (pending);
+
+       /* Another VDQ request pending and this request returned full */
+       if (is_repeat) {
+               /* all packets pulled from this pull request */
+               if (num_pulled == num_to_pull)  {
+                       /* pkt to pull in current pull request */
+                       num_to_pull = q_storage->last_num_pkts;
+
+                       dq_storage = dq_storage1;
+
+                       while (!qbman_check_command_complete(dq_storage))
+                               ;
+                       goto repeat;
+               } else {
+                       /* if this request did not returned all pkts */
+                       goto next_time;
+               }
+       }
+
+       q_storage->toggle ^= 1;
+       dq_storage = q_storage->dq_storage[q_storage->toggle];
+       q_storage->last_num_pkts = (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
+                                      DPAA2_DQRR_RING_SIZE : nb_pkts;
+       qbman_pull_desc_clear(&pulldesc);
+       qbman_pull_desc_set_numframes(&pulldesc, q_storage->last_num_pkts);
+       qbman_pull_desc_set_fq(&pulldesc, fqid);
+       qbman_pull_desc_set_storage(&pulldesc, dq_storage,
+                       (dma_addr_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
+       /* issue a volatile dequeue command for next pull */
+       while (1) {
+               if (qbman_swp_pull(swp, &pulldesc)) {
+                       PMD_RX_LOG(WARNING, "VDQ command is not issued."
+                                  "QBMAN is busy\n");
+                       continue;
+               }
+               break;
+       }
+       q_storage->active_dqs = dq_storage;
+       set_swp_active_dqs(DPAA2_PER_LCORE_DPIO->index, dq_storage);
 
+next_time:
        dpaa2_q->rx_pkts += num_rx;
 
-       /*Return the total number of packets received to DPAA2 app*/
        return num_rx;
 }
 
+void __attribute__((hot))
+dpaa2_dev_process_parallel_event(struct qbman_swp *swp,
+                                const struct qbman_fd *fd,
+                                const struct qbman_result *dq,
+                                struct dpaa2_queue *rxq,
+                                struct rte_event *ev)
+{
+       ev->mbuf = eth_fd_to_mbuf(fd);
+
+       ev->flow_id = rxq->ev.flow_id;
+       ev->sub_event_type = rxq->ev.sub_event_type;
+       ev->event_type = RTE_EVENT_TYPE_ETHDEV;
+       ev->op = RTE_EVENT_OP_NEW;
+       ev->sched_type = rxq->ev.sched_type;
+       ev->queue_id = rxq->ev.queue_id;
+       ev->priority = rxq->ev.priority;
+
+       qbman_swp_dqrr_consume(swp, dq);
+}
+
+void dpaa2_dev_process_atomic_event(struct qbman_swp *swp __attribute__((unused)),
+                                   const struct qbman_fd *fd,
+                                   const struct qbman_result *dq,
+                                   struct dpaa2_queue *rxq,
+                                   struct rte_event *ev)
+{
+       uint8_t dqrr_index = qbman_get_dqrr_idx(dq);
+
+       ev->mbuf = eth_fd_to_mbuf(fd);
+
+       ev->flow_id = rxq->ev.flow_id;
+       ev->sub_event_type = rxq->ev.sub_event_type;
+       ev->event_type = RTE_EVENT_TYPE_ETHDEV;
+       ev->op = RTE_EVENT_OP_NEW;
+       ev->sched_type = rxq->ev.sched_type;
+       ev->queue_id = rxq->ev.queue_id;
+       ev->priority = rxq->ev.priority;
+
+       ev->mbuf->seqn = dqrr_index + 1;
+       DPAA2_PER_LCORE_DQRR_SIZE++;
+       DPAA2_PER_LCORE_DQRR_HELD |= 1 << dqrr_index;
+       DPAA2_PER_LCORE_DQRR_MBUF(dqrr_index) = ev->mbuf;
+}
+
 /*
  * Callback to handle sending packets through WRIOP based interface
  */
@@ -342,9 +744,10 @@ uint16_t
 dpaa2_dev_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
 {
        /* Function to transmit the frames to given device and VQ*/
-       uint32_t loop;
+       uint32_t loop, retry_count;
        int32_t ret;
        struct qbman_fd fd_arr[MAX_TX_RING_SLOTS];
+       struct rte_mbuf *mi;
        uint32_t frames_to_send;
        struct rte_mempool *mp;
        struct qbman_eq_desc eqdesc;
@@ -354,6 +757,7 @@ dpaa2_dev_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
        uint16_t bpid;
        struct rte_eth_dev *dev = dpaa2_q->dev;
        struct dpaa2_dev_priv *priv = dev->data->dev_private;
+       uint32_t flags[MAX_TX_RING_SLOTS] = {0};
 
        if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
                ret = dpaa2_affine_qbman_swp();
@@ -372,51 +776,146 @@ dpaa2_dev_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
        qbman_eq_desc_set_response(&eqdesc, 0, 0);
        qbman_eq_desc_set_qd(&eqdesc, priv->qdid,
                             dpaa2_q->flow_id, dpaa2_q->tc_index);
-
        /*Clear the unused FD fields before sending*/
        while (nb_pkts) {
+               /*Check if the queue is congested*/
+               retry_count = 0;
+               while (qbman_result_SCN_state(dpaa2_q->cscn)) {
+                       retry_count++;
+                       /* Retry for some time before giving up */
+                       if (retry_count > CONG_RETRY_COUNT)
+                               goto skip_tx;
+               }
+
                frames_to_send = (nb_pkts >> 3) ? MAX_TX_RING_SLOTS : nb_pkts;
 
                for (loop = 0; loop < frames_to_send; loop++) {
+                       if ((*bufs)->seqn) {
+                               uint8_t dqrr_index = (*bufs)->seqn - 1;
+
+                               flags[loop] = QBMAN_ENQUEUE_FLAG_DCA |
+                                               dqrr_index;
+                               DPAA2_PER_LCORE_DQRR_SIZE--;
+                               DPAA2_PER_LCORE_DQRR_HELD &= ~(1 << dqrr_index);
+                               (*bufs)->seqn = DPAA2_INVALID_MBUF_SEQN;
+                       }
+
                        fd_arr[loop].simple.frc = 0;
                        DPAA2_RESET_FD_CTRL((&fd_arr[loop]));
                        DPAA2_SET_FD_FLC((&fd_arr[loop]), NULL);
-                       mp = (*bufs)->pool;
+                       if (likely(RTE_MBUF_DIRECT(*bufs))) {
+                               mp = (*bufs)->pool;
+                               /* Check the basic scenario and set
+                                * the FD appropriately here itself.
+                                */
+                               if (likely(mp && mp->ops_index ==
+                                   priv->bp_list->dpaa2_ops_index &&
+                                   (*bufs)->nb_segs == 1 &&
+                                   rte_mbuf_refcnt_read((*bufs)) == 1)) {
+                                       if (unlikely((*bufs)->ol_flags
+                                               & PKT_TX_VLAN_PKT)) {
+                                               ret = rte_vlan_insert(bufs);
+                                               if (ret)
+                                                       goto send_n_return;
+                                       }
+                                       DPAA2_MBUF_TO_CONTIG_FD((*bufs),
+                                       &fd_arr[loop], mempool_to_bpid(mp));
+                                       bufs++;
+                                       continue;
+                               }
+                       } else {
+                               mi = rte_mbuf_from_indirect(*bufs);
+                               mp = mi->pool;
+                       }
                        /* Not a hw_pkt pool allocated frame */
+                       if (unlikely(!mp || !priv->bp_list)) {
+                               PMD_TX_LOG(ERR, "err: no bpool attached");
+                               goto send_n_return;
+                       }
+
                        if (mp->ops_index != priv->bp_list->dpaa2_ops_index) {
                                PMD_TX_LOG(ERR, "non hw offload bufffer ");
                                /* alloc should be from the default buffer pool
                                 * attached to this interface
                                 */
-                               if (priv->bp_list) {
-                                       bpid = priv->bp_list->buf_pool.bpid;
-                               } else {
-                                       PMD_TX_LOG(ERR, "errr: why no bpool"
-                                                  " attached");
-                                       num_tx = 0;
-                                       goto skip_tx;
+                               bpid = priv->bp_list->buf_pool.bpid;
+
+                               if (unlikely((*bufs)->nb_segs > 1)) {
+                                       PMD_TX_LOG(ERR, "S/G support not added"
+                                               " for non hw offload buffer");
+                                       goto send_n_return;
                                }
                                if (eth_copy_mbuf_to_fd(*bufs,
                                                        &fd_arr[loop], bpid)) {
-                                       bufs++;
-                                       continue;
+                                       goto send_n_return;
                                }
+                               /* free the original packet */
+                               rte_pktmbuf_free(*bufs);
                        } else {
                                bpid = mempool_to_bpid(mp);
-                               eth_mbuf_to_fd(*bufs, &fd_arr[loop], bpid);
+                               if (unlikely((*bufs)->nb_segs > 1)) {
+                                       if (eth_mbuf_to_sg_fd(*bufs,
+                                                       &fd_arr[loop], bpid))
+                                               goto send_n_return;
+                               } else {
+                                       eth_mbuf_to_fd(*bufs,
+                                                      &fd_arr[loop], bpid);
+                               }
                        }
                        bufs++;
                }
                loop = 0;
                while (loop < frames_to_send) {
-                       loop += qbman_swp_send_multiple(swp, &eqdesc,
-                                       &fd_arr[loop], frames_to_send - loop);
+                       loop += qbman_swp_enqueue_multiple(swp, &eqdesc,
+                                       &fd_arr[loop], &flags[loop],
+                                       frames_to_send - loop);
                }
 
                num_tx += frames_to_send;
-               dpaa2_q->tx_pkts += frames_to_send;
                nb_pkts -= frames_to_send;
        }
+       dpaa2_q->tx_pkts += num_tx;
+       return num_tx;
+
+send_n_return:
+       /* send any already prepared fd */
+       if (loop) {
+               unsigned int i = 0;
+
+               while (i < loop) {
+                       i += qbman_swp_enqueue_multiple(swp, &eqdesc,
+                                                       &fd_arr[i],
+                                                       &flags[loop],
+                                                       loop - i);
+               }
+               num_tx += loop;
+       }
 skip_tx:
+       dpaa2_q->tx_pkts += num_tx;
        return num_tx;
 }
+
+/**
+ * Dummy DPDK callback for TX.
+ *
+ * This function is used to temporarily replace the real callback during
+ * unsafe control operations on the queue, or in case of error.
+ *
+ * @param dpdk_txq
+ *   Generic pointer to TX queue structure.
+ * @param[in] pkts
+ *   Packets to transmit.
+ * @param pkts_n
+ *   Number of packets in array.
+ *
+ * @return
+ *   Number of packets successfully transmitted (<= pkts_n).
+ */
+uint16_t
+dummy_dev_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+{
+       (void)queue;
+       (void)bufs;
+       (void)nb_pkts;
+       return 0;
+}