net: add rte prefix to ether structures
[dpdk.git] / drivers / net / vmxnet3 / vmxnet3_rxtx.c
index d08f5eb..b691141 100644 (file)
@@ -1,34 +1,5 @@
-/*-
- *   BSD LICENSE
- *
- *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
- *   All rights reserved.
- *
- *   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 Intel Corporation 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.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2015 Intel Corporation
  */
 
 #include <sys/queue.h>
@@ -61,7 +32,7 @@
 #include <rte_malloc.h>
 #include <rte_mbuf.h>
 #include <rte_ether.h>
-#include <rte_ethdev.h>
+#include <rte_ethdev_driver.h>
 #include <rte_prefetch.h>
 #include <rte_ip.h>
 #include <rte_udp.h>
@@ -79,6 +50,8 @@
 
 #define        VMXNET3_TX_OFFLOAD_MASK ( \
                PKT_TX_VLAN_PKT | \
+               PKT_TX_IPV6 |     \
+               PKT_TX_IPV4 |     \
                PKT_TX_L4_MASK |  \
                PKT_TX_TCP_SEG)
 
@@ -203,6 +176,8 @@ vmxnet3_dev_tx_queue_release(void *txq)
                vmxnet3_cmd_ring_release(&tq->cmd_ring);
                /* Release the memzone */
                rte_memzone_free(tq->mz);
+               /* Release the queue */
+               rte_free(tq);
        }
 }
 
@@ -223,6 +198,9 @@ vmxnet3_dev_rx_queue_release(void *rxq)
 
                /* Release the memzone */
                rte_memzone_free(rq->mz);
+
+               /* Release the queue */
+               rte_free(rq);
        }
 }
 
@@ -265,11 +243,9 @@ vmxnet3_dev_rx_queue_reset(void *rxq)
        struct vmxnet3_rx_data_ring *data_ring = &rq->data_ring;
        int size;
 
-       if (rq != NULL) {
-               /* Release both the cmd_rings mbufs */
-               for (i = 0; i < VMXNET3_RX_CMDRING_SIZE; i++)
-                       vmxnet3_rx_cmd_ring_release_mbufs(&rq->cmd_ring[i]);
-       }
+       /* Release both the cmd_rings mbufs */
+       for (i = 0; i < VMXNET3_RX_CMDRING_SIZE; i++)
+               vmxnet3_rx_cmd_ring_release_mbufs(&rq->cmd_ring[i]);
 
        ring0 = &rq->cmd_ring[0];
        ring1 = &rq->cmd_ring[1];
@@ -483,6 +459,14 @@ vmxnet3_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
                    rte_pktmbuf_pkt_len(txm) <= txq->txdata_desc_size) {
                        struct Vmxnet3_TxDataDesc *tdd;
 
+                       /* Skip empty packets */
+                       if (unlikely(rte_pktmbuf_pkt_len(txm) == 0)) {
+                               txq->stats.drop_total++;
+                               rte_pktmbuf_free(txm);
+                               nb_tx++;
+                               continue;
+                       }
+
                        tdd = (struct Vmxnet3_TxDataDesc *)
                                ((uint8 *)txq->data_ring.base +
                                 txq->cmd_ring.next2fill *
@@ -503,6 +487,11 @@ vmxnet3_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
                         * maximum size of mbuf segment size.
                         */
                        gdesc = txq->cmd_ring.base + txq->cmd_ring.next2fill;
+
+                       /* Skip empty segments */
+                       if (unlikely(m_seg->data_len == 0))
+                               continue;
+
                        if (copy_size) {
                                uint64 offset =
                                        (uint64)txq->cmd_ring.next2fill *
@@ -511,7 +500,7 @@ vmxnet3_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
                                        rte_cpu_to_le_64(txq->data_ring.basePA +
                                                         offset);
                        } else {
-                               gdesc->txd.addr = rte_mbuf_data_dma_addr(m_seg);
+                               gdesc->txd.addr = rte_mbuf_data_iova(m_seg);
                        }
 
                        gdesc->dword[2] = dw2 | m_seg->data_len;
@@ -619,7 +608,7 @@ vmxnet3_renew_desc(vmxnet3_rx_queue_t *rxq, uint8_t ring_id,
         */
        buf_info->m = mbuf;
        buf_info->len = (uint16_t)(mbuf->buf_len - RTE_PKTMBUF_HEADROOM);
-       buf_info->bufPA = rte_mbuf_data_dma_addr_default(mbuf);
+       buf_info->bufPA = rte_mbuf_data_iova_default(mbuf);
 
        /* Load Rx Descriptor with the buffer's GPA */
        rxd->addr = buf_info->bufPA;
@@ -672,35 +661,154 @@ vmxnet3_post_rx_bufs(vmxnet3_rx_queue_t *rxq, uint8_t ring_id)
                return i;
 }
 
-
-/* Receive side checksum and other offloads */
-static void
-vmxnet3_rx_offload(const Vmxnet3_RxCompDesc *rcd, struct rte_mbuf *rxm)
+/* MSS not provided by vmxnet3, guess one with available information */
+static uint16_t
+vmxnet3_guess_mss(struct vmxnet3_hw *hw, const Vmxnet3_RxCompDesc *rcd,
+               struct rte_mbuf *rxm)
 {
-       /* Check for RSS */
-       if (rcd->rssType != VMXNET3_RCD_RSS_TYPE_NONE) {
-               rxm->ol_flags |= PKT_RX_RSS_HASH;
-               rxm->hash.rss = rcd->rssHash;
-       }
+       uint32_t hlen, slen;
+       struct ipv4_hdr *ipv4_hdr;
+       struct ipv6_hdr *ipv6_hdr;
+       struct tcp_hdr *tcp_hdr;
+       char *ptr;
+
+       RTE_ASSERT(rcd->tcp);
+
+       ptr = rte_pktmbuf_mtod(rxm, char *);
+       slen = rte_pktmbuf_data_len(rxm);
+       hlen = sizeof(struct rte_ether_hdr);
 
-       /* Check packet type, checksum errors, etc. Only support IPv4 for now. */
        if (rcd->v4) {
-               struct ether_hdr *eth = rte_pktmbuf_mtod(rxm, struct ether_hdr *);
-               struct ipv4_hdr *ip = (struct ipv4_hdr *)(eth + 1);
+               if (unlikely(slen < hlen + sizeof(struct ipv4_hdr)))
+                       return hw->mtu - sizeof(struct ipv4_hdr)
+                                       - sizeof(struct tcp_hdr);
+
+               ipv4_hdr = (struct ipv4_hdr *)(ptr + hlen);
+               hlen += (ipv4_hdr->version_ihl & IPV4_HDR_IHL_MASK) *
+                               IPV4_IHL_MULTIPLIER;
+       } else if (rcd->v6) {
+               if (unlikely(slen < hlen + sizeof(struct ipv6_hdr)))
+                       return hw->mtu - sizeof(struct ipv6_hdr) -
+                                       sizeof(struct tcp_hdr);
+
+               ipv6_hdr = (struct ipv6_hdr *)(ptr + hlen);
+               hlen += sizeof(struct ipv6_hdr);
+               if (unlikely(ipv6_hdr->proto != IPPROTO_TCP)) {
+                       int frag;
+
+                       rte_net_skip_ip6_ext(ipv6_hdr->proto, rxm,
+                                       &hlen, &frag);
+               }
+       }
 
-               if (((ip->version_ihl & 0xf) << 2) > (int)sizeof(struct ipv4_hdr))
-                       rxm->packet_type = RTE_PTYPE_L3_IPV4_EXT;
-               else
-                       rxm->packet_type = RTE_PTYPE_L3_IPV4;
+       if (unlikely(slen < hlen + sizeof(struct tcp_hdr)))
+               return hw->mtu - hlen - sizeof(struct tcp_hdr) +
+                               sizeof(struct rte_ether_hdr);
 
-               if (!rcd->cnc) {
-                       if (!rcd->ipc)
-                               rxm->ol_flags |= PKT_RX_IP_CKSUM_BAD;
+       tcp_hdr = (struct tcp_hdr *)(ptr + hlen);
+       hlen += (tcp_hdr->data_off & 0xf0) >> 2;
+
+       if (rxm->udata64 > 1)
+               return (rte_pktmbuf_pkt_len(rxm) - hlen +
+                               rxm->udata64 - 1) / rxm->udata64;
+       else
+               return hw->mtu - hlen + sizeof(struct rte_ether_hdr);
+}
 
-                       if ((rcd->tcp || rcd->udp) && !rcd->tuc)
-                               rxm->ol_flags |= PKT_RX_L4_CKSUM_BAD;
+/* Receive side checksum and other offloads */
+static inline void
+vmxnet3_rx_offload(struct vmxnet3_hw *hw, const Vmxnet3_RxCompDesc *rcd,
+               struct rte_mbuf *rxm, const uint8_t sop)
+{
+       uint64_t ol_flags = rxm->ol_flags;
+       uint32_t packet_type = rxm->packet_type;
+
+       /* Offloads set in sop */
+       if (sop) {
+               /* Set packet type */
+               packet_type |= RTE_PTYPE_L2_ETHER;
+
+               /* Check large packet receive */
+               if (VMXNET3_VERSION_GE_2(hw) &&
+                   rcd->type == VMXNET3_CDTYPE_RXCOMP_LRO) {
+                       const Vmxnet3_RxCompDescExt *rcde =
+                                       (const Vmxnet3_RxCompDescExt *)rcd;
+
+                       rxm->tso_segsz = rcde->mss;
+                       rxm->udata64 = rcde->segCnt;
+                       ol_flags |= PKT_RX_LRO;
+               }
+       } else { /* Offloads set in eop */
+               /* Check for RSS */
+               if (rcd->rssType != VMXNET3_RCD_RSS_TYPE_NONE) {
+                       ol_flags |= PKT_RX_RSS_HASH;
+                       rxm->hash.rss = rcd->rssHash;
+               }
+
+               /* Check for hardware stripped VLAN tag */
+               if (rcd->ts) {
+                       ol_flags |= (PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED);
+                       rxm->vlan_tci = rte_le_to_cpu_16((uint16_t)rcd->tci);
+               }
+
+               /* Check packet type, checksum errors, etc. */
+               if (rcd->cnc) {
+                       ol_flags |= PKT_RX_L4_CKSUM_UNKNOWN;
+               } else {
+                       if (rcd->v4) {
+                               packet_type |= RTE_PTYPE_L3_IPV4_EXT_UNKNOWN;
+
+                               if (rcd->ipc)
+                                       ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+                               else
+                                       ol_flags |= PKT_RX_IP_CKSUM_BAD;
+
+                               if (rcd->tuc) {
+                                       ol_flags |= PKT_RX_L4_CKSUM_GOOD;
+                                       if (rcd->tcp)
+                                               packet_type |= RTE_PTYPE_L4_TCP;
+                                       else
+                                               packet_type |= RTE_PTYPE_L4_UDP;
+                               } else {
+                                       if (rcd->tcp) {
+                                               packet_type |= RTE_PTYPE_L4_TCP;
+                                               ol_flags |= PKT_RX_L4_CKSUM_BAD;
+                                       } else if (rcd->udp) {
+                                               packet_type |= RTE_PTYPE_L4_UDP;
+                                               ol_flags |= PKT_RX_L4_CKSUM_BAD;
+                                       }
+                               }
+                       } else if (rcd->v6) {
+                               packet_type |= RTE_PTYPE_L3_IPV6_EXT_UNKNOWN;
+
+                               if (rcd->tuc) {
+                                       ol_flags |= PKT_RX_L4_CKSUM_GOOD;
+                                       if (rcd->tcp)
+                                               packet_type |= RTE_PTYPE_L4_TCP;
+                                       else
+                                               packet_type |= RTE_PTYPE_L4_UDP;
+                               } else {
+                                       if (rcd->tcp) {
+                                               packet_type |= RTE_PTYPE_L4_TCP;
+                                               ol_flags |= PKT_RX_L4_CKSUM_BAD;
+                                       } else if (rcd->udp) {
+                                               packet_type |= RTE_PTYPE_L4_UDP;
+                                               ol_flags |= PKT_RX_L4_CKSUM_BAD;
+                                       }
+                               }
+                       } else {
+                               packet_type |= RTE_PTYPE_UNKNOWN;
+                       }
+
+                       /* Old variants of vmxnet3 do not provide MSS */
+                       if ((ol_flags & PKT_RX_LRO) && rxm->tso_segsz == 0)
+                               rxm->tso_segsz = vmxnet3_guess_mss(hw,
+                                               rcd, rxm);
                }
        }
+
+       rxm->ol_flags = ol_flags;
+       rxm->packet_type = packet_type;
 }
 
 /*
@@ -800,6 +908,7 @@ vmxnet3_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
                rxm->data_off = RTE_PKTMBUF_HEADROOM;
                rxm->ol_flags = 0;
                rxm->vlan_tci = 0;
+               rxm->packet_type = 0;
 
                /*
                 * If this is the first buffer of the received packet,
@@ -831,28 +940,28 @@ vmxnet3_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
                        }
 
                        rxq->start_seg = rxm;
-                       vmxnet3_rx_offload(rcd, rxm);
+                       rxq->last_seg = rxm;
+                       vmxnet3_rx_offload(hw, rcd, rxm, 1);
                } else {
                        struct rte_mbuf *start = rxq->start_seg;
 
                        RTE_ASSERT(rxd->btype == VMXNET3_RXD_BTYPE_BODY);
 
-                       start->pkt_len += rxm->data_len;
-                       start->nb_segs++;
+                       if (rxm->data_len) {
+                               start->pkt_len += rxm->data_len;
+                               start->nb_segs++;
 
-                       rxq->last_seg->next = rxm;
+                               rxq->last_seg->next = rxm;
+                               rxq->last_seg = rxm;
+                       } else {
+                               rte_pktmbuf_free_seg(rxm);
+                       }
                }
-               rxq->last_seg = rxm;
 
                if (rcd->eop) {
                        struct rte_mbuf *start = rxq->start_seg;
 
-                       /* Check for hardware stripped VLAN tag */
-                       if (rcd->ts) {
-                               start->ol_flags |= (PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED);
-                               start->vlan_tci = rte_le_to_cpu_16((uint16_t)rcd->tci);
-                       }
-
+                       vmxnet3_rx_offload(hw, rcd, start, 0);
                        rx_pkts[nb_rx++] = start;
                        rxq->start_seg = NULL;
                }
@@ -906,7 +1015,7 @@ vmxnet3_dev_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_eth_txconf *tx_conf __rte_unused)
 {
        struct vmxnet3_hw *hw = dev->data->dev_private;
        const struct rte_memzone *mz;
@@ -918,12 +1027,6 @@ vmxnet3_dev_tx_queue_setup(struct rte_eth_dev *dev,
 
        PMD_INIT_FUNC_TRACE();
 
-       if ((tx_conf->txq_flags & ETH_TXQ_FLAGS_NOXSUMSCTP) !=
-           ETH_TXQ_FLAGS_NOXSUMSCTP) {
-               PMD_INIT_LOG(ERR, "SCTP checksum offload not supported");
-               return -EINVAL;
-       }
-
        txq = rte_zmalloc("ethdev_tx_queue", sizeof(struct vmxnet3_tx_queue),
                          RTE_CACHE_LINE_SIZE);
        if (txq == NULL) {
@@ -933,7 +1036,7 @@ vmxnet3_dev_tx_queue_setup(struct rte_eth_dev *dev,
 
        txq->queue_id = queue_idx;
        txq->port_id = dev->data->port_id;
-       txq->shared = &hw->tqd_start[queue_idx];
+       txq->shared = NULL; /* set in vmxnet3_setup_driver_shared() */
        txq->hw = hw;
        txq->qid = queue_idx;
        txq->stopped = TRUE;
@@ -980,7 +1083,7 @@ vmxnet3_dev_tx_queue_setup(struct rte_eth_dev *dev,
 
        /* cmd_ring initialization */
        ring->base = mz->addr;
-       ring->basePA = mz->phys_addr;
+       ring->basePA = mz->iova;
 
        /* comp_ring initialization */
        comp_ring->base = ring->base + ring->size;
@@ -1036,7 +1139,7 @@ vmxnet3_dev_rx_queue_setup(struct rte_eth_dev *dev,
        rxq->mp = mp;
        rxq->queue_id = queue_idx;
        rxq->port_id = dev->data->port_id;
-       rxq->shared = &hw->rqd_start[queue_idx];
+       rxq->shared = NULL; /* set in vmxnet3_setup_driver_shared() */
        rxq->hw = hw;
        rxq->qid1 = queue_idx;
        rxq->qid2 = queue_idx + hw->num_rx_queues;
@@ -1091,7 +1194,7 @@ vmxnet3_dev_rx_queue_setup(struct rte_eth_dev *dev,
 
        /* cmd_ring0 initialization */
        ring0->base = mz->addr;
-       ring0->basePA = mz->phys_addr;
+       ring0->basePA = mz->iova;
 
        /* cmd_ring1 initialization */
        ring1->base = ring0->base + ring0->size;
@@ -1188,6 +1291,46 @@ static uint8_t rss_intel_key[40] = {
        0x6A, 0x42, 0xB7, 0x3B, 0xBE, 0xAC, 0x01, 0xFA,
 };
 
+/*
+ * Additional RSS configurations based on vmxnet v4+ APIs
+ */
+int
+vmxnet3_v4_rss_configure(struct rte_eth_dev *dev)
+{
+       struct vmxnet3_hw *hw = dev->data->dev_private;
+       Vmxnet3_DriverShared *shared = hw->shared;
+       Vmxnet3_CmdInfo *cmdInfo = &shared->cu.cmdInfo;
+       struct rte_eth_rss_conf *port_rss_conf;
+       uint64_t rss_hf;
+       uint32_t ret;
+
+       PMD_INIT_FUNC_TRACE();
+
+       cmdInfo->setRSSFields = 0;
+       port_rss_conf = &dev->data->dev_conf.rx_adv_conf.rss_conf;
+       rss_hf = port_rss_conf->rss_hf &
+               (VMXNET3_V4_RSS_MASK | VMXNET3_RSS_OFFLOAD_ALL);
+
+       if (rss_hf & ETH_RSS_NONFRAG_IPV4_TCP)
+               cmdInfo->setRSSFields |= VMXNET3_RSS_FIELDS_TCPIP4;
+       if (rss_hf & ETH_RSS_NONFRAG_IPV6_TCP)
+               cmdInfo->setRSSFields |= VMXNET3_RSS_FIELDS_TCPIP6;
+       if (rss_hf & ETH_RSS_NONFRAG_IPV4_UDP)
+               cmdInfo->setRSSFields |= VMXNET3_RSS_FIELDS_UDPIP4;
+       if (rss_hf & ETH_RSS_NONFRAG_IPV6_UDP)
+               cmdInfo->setRSSFields |= VMXNET3_RSS_FIELDS_UDPIP6;
+
+       VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
+                              VMXNET3_CMD_SET_RSS_FIELDS);
+       ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
+
+       if (ret != VMXNET3_SUCCESS) {
+               PMD_DRV_LOG(ERR, "Set RSS fields (v4) failed: %d", ret);
+       }
+
+       return ret;
+}
+
 /*
  * Configure RSS feature
  */