diff options
Diffstat (limited to 'drivers/net/nfp/nfp_net.c')
| -rw-r--r-- | drivers/net/nfp/nfp_net.c | 873 |
1 files changed, 497 insertions, 376 deletions
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c index e5bfde62..6e5e305f 100644 --- a/drivers/net/nfp/nfp_net.c +++ b/drivers/net/nfp/nfp_net.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2014, 2015 Netronome Systems, Inc. + * Copyright (c) 2014-2018 Netronome Systems, Inc. * All rights reserved. * * Small portions derived from code Copyright(c) 2010-2015 Intel Corporation. @@ -55,7 +55,13 @@ #include <rte_alarm.h> #include <rte_spinlock.h> -#include "nfp_nfpu.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_net_logs.h" #include "nfp_net_ctrl.h" @@ -103,13 +109,11 @@ static int nfp_net_rss_reta_write(struct rte_eth_dev *dev, 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 ether_addr *mac_addr); -/* - * The offset of the queue controller queues in the PCIe Target. These - * happen to be at the same offset on the NFP6000 and the NFP3200 so - * we use a single macro here. - */ -#define NFP_PCIE_QUEUE(_q) (0x800 * ((_q) & 0xff)) +/* 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 @@ -213,57 +217,6 @@ 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); } -/* - * Atomically reads link status information from global structure rte_eth_dev. - * - * @param dev - * - Pointer to the structure rte_eth_dev to read from. - * - Pointer to the buffer to be saved with the link status. - * - * @return - * - On success, zero. - * - On failure, negative value. - */ -static inline int -nfp_net_dev_atomic_read_link_status(struct rte_eth_dev *dev, - struct rte_eth_link *link) -{ - struct rte_eth_link *dst = link; - struct rte_eth_link *src = &dev->data->dev_link; - - if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst, - *(uint64_t *)src) == 0) - return -1; - - return 0; -} - -/* - * Atomically writes the link status information into global - * structure rte_eth_dev. - * - * @param dev - * - Pointer to the structure rte_eth_dev to read from. - * - Pointer to the buffer to be saved with the link status. - * - * @return - * - On success, zero. - * - On failure, negative value. - */ -static inline int -nfp_net_dev_atomic_write_link_status(struct rte_eth_dev *dev, - struct rte_eth_link *link) -{ - struct rte_eth_link *dst = &dev->data->dev_link; - struct rte_eth_link *src = link; - - if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst, - *(uint64_t *)src) == 0) - return -1; - - return 0; -} - static void nfp_net_rx_queue_release_mbufs(struct nfp_net_rxq *rxq) { @@ -310,7 +263,7 @@ nfp_net_tx_queue_release_mbufs(struct nfp_net_txq *txq) for (i = 0; i < txq->tx_count; i++) { if (txq->txbufs[i].mbuf) { - rte_pktmbuf_free(txq->txbufs[i].mbuf); + rte_pktmbuf_free_seg(txq->txbufs[i].mbuf); txq->txbufs[i].mbuf = NULL; } } @@ -343,7 +296,7 @@ __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update) uint32_t new; struct timespec wait; - PMD_DRV_LOG(DEBUG, "Writing to the configuration queue (%p)...\n", + PMD_DRV_LOG(DEBUG, "Writing to the configuration queue (%p)...", hw->qcp_cfg); if (hw->qcp_cfg == NULL) @@ -354,7 +307,7 @@ __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update) wait.tv_sec = 0; wait.tv_nsec = 1000000; - PMD_DRV_LOG(DEBUG, "Polling for update ack...\n"); + PMD_DRV_LOG(DEBUG, "Polling for update ack..."); /* Poll update field, waiting for NFP to ack the config */ for (cnt = 0; ; cnt++) { @@ -372,7 +325,7 @@ __nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t update) } nanosleep(&wait, 0); /* waiting for a 1ms */ } - PMD_DRV_LOG(DEBUG, "Ack DONE\n"); + PMD_DRV_LOG(DEBUG, "Ack DONE"); return 0; } @@ -390,7 +343,7 @@ nfp_net_reconfig(struct nfp_net_hw *hw, uint32_t ctrl, uint32_t update) { uint32_t err; - PMD_DRV_LOG(DEBUG, "nfp_net_reconfig: ctrl=%08x update=%08x\n", + PMD_DRV_LOG(DEBUG, "nfp_net_reconfig: ctrl=%08x update=%08x", ctrl, update); rte_spinlock_lock(&hw->reconfig_lock); @@ -427,8 +380,6 @@ nfp_net_configure(struct rte_eth_dev *dev) struct rte_eth_conf *dev_conf; struct rte_eth_rxmode *rxmode; struct rte_eth_txmode *txmode; - uint32_t new_ctrl = 0; - uint32_t update = 0; struct nfp_net_hw *hw; hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); @@ -454,96 +405,17 @@ nfp_net_configure(struct rte_eth_dev *dev) } /* Checking RX mode */ - if (rxmode->mq_mode & ETH_MQ_RX_RSS) { - if (hw->cap & NFP_NET_CFG_CTRL_RSS) { - update = NFP_NET_CFG_UPDATE_RSS; - new_ctrl = NFP_NET_CFG_CTRL_RSS; - } else { - PMD_INIT_LOG(INFO, "RSS not supported"); - return -EINVAL; - } - } - - if (rxmode->split_hdr_size) { - PMD_INIT_LOG(INFO, "rxmode does not support split header"); - return -EINVAL; - } - - if (rxmode->hw_ip_checksum) { - if (hw->cap & NFP_NET_CFG_CTRL_RXCSUM) { - new_ctrl |= NFP_NET_CFG_CTRL_RXCSUM; - } else { - PMD_INIT_LOG(INFO, "RXCSUM not supported"); - return -EINVAL; - } - } - - if (rxmode->hw_vlan_filter) { - PMD_INIT_LOG(INFO, "VLAN filter not supported"); - return -EINVAL; - } - - if (rxmode->hw_vlan_strip) { - if (hw->cap & NFP_NET_CFG_CTRL_RXVLAN) { - new_ctrl |= NFP_NET_CFG_CTRL_RXVLAN; - } else { - PMD_INIT_LOG(INFO, "hw vlan strip not supported"); - return -EINVAL; - } - } - - if (rxmode->hw_vlan_extend) { - PMD_INIT_LOG(INFO, "VLAN extended not supported"); - return -EINVAL; - } - - if (rxmode->jumbo_frame) - hw->mtu = rxmode->max_rx_pkt_len; - - if (!rxmode->hw_strip_crc) - PMD_INIT_LOG(INFO, "HW does strip CRC and it is not configurable"); - - if (rxmode->enable_scatter) { - PMD_INIT_LOG(INFO, "Scatter not supported"); + if (rxmode->mq_mode & ETH_MQ_RX_RSS && + !(hw->cap & NFP_NET_CFG_CTRL_RSS)) { + PMD_INIT_LOG(INFO, "RSS not supported"); return -EINVAL; } - /* If next capabilities are supported, configure them by default */ - - /* VLAN insertion */ - if (hw->cap & NFP_NET_CFG_CTRL_TXVLAN) - new_ctrl |= NFP_NET_CFG_CTRL_TXVLAN; - - /* L2 broadcast */ - if (hw->cap & NFP_NET_CFG_CTRL_L2BC) - new_ctrl |= NFP_NET_CFG_CTRL_L2BC; - - /* L2 multicast */ - if (hw->cap & NFP_NET_CFG_CTRL_L2MC) - new_ctrl |= NFP_NET_CFG_CTRL_L2MC; - - /* TX checksum offload */ - if (hw->cap & NFP_NET_CFG_CTRL_TXCSUM) - new_ctrl |= NFP_NET_CFG_CTRL_TXCSUM; - - /* LSO offload */ - if (hw->cap & NFP_NET_CFG_CTRL_LSO) - new_ctrl |= NFP_NET_CFG_CTRL_LSO; - - /* RX gather */ - if (hw->cap & NFP_NET_CFG_CTRL_GATHER) - new_ctrl |= NFP_NET_CFG_CTRL_GATHER; - - if (!new_ctrl) - return 0; - - update |= NFP_NET_CFG_UPDATE_GEN; - - nn_cfg_writel(hw, NFP_NET_CFG_CTRL, new_ctrl); - if (nfp_net_reconfig(hw, new_ctrl, update) < 0) - return -EIO; - - hw->ctrl = new_ctrl; + /* KEEP_CRC offload flag is not supported by PMD + * can remove the below block when DEV_RX_OFFLOAD_CRC_STRIP removed + */ + if (rte_eth_dev_must_keep_crc(rxmode->offloads)) + PMD_INIT_LOG(INFO, "HW does strip CRC. No configurable!"); return 0; } @@ -625,47 +497,29 @@ nfp_net_cfg_queue_setup(struct nfp_net_hw *hw) #define ETH_ADDR_LEN 6 static void -nfp_eth_copy_mac_reverse(uint8_t *dst, const uint8_t *src) +nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src) { int i; for (i = 0; i < ETH_ADDR_LEN; i++) - dst[ETH_ADDR_LEN - i - 1] = src[i]; + dst[i] = src[i]; } static int nfp_net_pf_read_mac(struct nfp_net_hw *hw, int port) { - union eth_table_entry *entry; - int idx, i; - - idx = port; - entry = hw->eth_table; - - /* Reading NFP ethernet table obtained before */ - for (i = 0; i < NSP_ETH_MAX_COUNT; i++) { - if (!(entry->port & NSP_ETH_PORT_LANES_MASK)) { - /* port not in use */ - entry++; - continue; - } - if (idx == 0) - break; - idx--; - entry++; - } - - if (i == NSP_ETH_MAX_COUNT) - return -EINVAL; + struct nfp_eth_table *nfp_eth_table; + nfp_eth_table = nfp_eth_read_ports(hw->cpp); /* * hw points to port0 private data. We need hw now pointing to * right port. */ hw += port; - nfp_eth_copy_mac_reverse((uint8_t *)&hw->mac_addr, - (uint8_t *)&entry->mac_addr); + nfp_eth_copy_mac((uint8_t *)&hw->mac_addr, + (uint8_t *)&nfp_eth_table->ports[port].mac_addr); + free(nfp_eth_table); return 0; } @@ -675,7 +529,7 @@ 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, sizeof(struct ether_addr)); + 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); @@ -695,6 +549,37 @@ nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac) hw->ctrl_bar + NFP_NET_CFG_MACADDR + 6); } +int +nfp_set_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr) +{ + struct nfp_net_hw *hw; + uint32_t update, ctrl; + + hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); + if ((hw->ctrl & NFP_NET_CFG_CTRL_ENABLE) && + !(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR)) { + PMD_INIT_LOG(INFO, "MAC address unable to change when" + " port enabled"); + return -EBUSY; + } + + if ((hw->ctrl & NFP_NET_CFG_CTRL_ENABLE) && + !(hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR)) + return -EBUSY; + + /* Writing new MAC to the specific port BAR address */ + nfp_net_write_mac(hw, (uint8_t *)mac_addr); + + /* Signal the NIC about the change */ + update = NFP_NET_CFG_UPDATE_MACADDR; + ctrl = hw->ctrl | NFP_NET_CFG_CTRL_LIVE_ADDR; + if (nfp_net_reconfig(hw, ctrl, update) < 0) { + PMD_INIT_LOG(INFO, "MAC address update failed"); + return -EIO; + } + return 0; +} + static int nfp_configure_rx_interrupt(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle) @@ -729,7 +614,7 @@ nfp_configure_rx_interrupt(struct rte_eth_dev *dev, */ nn_cfg_writeb(hw, NFP_NET_CFG_RXR_VEC(i), i + 1); intr_handle->intr_vec[i] = i + 1; - PMD_INIT_LOG(DEBUG, "intr_vec[%d]= %d\n", i, + PMD_INIT_LOG(DEBUG, "intr_vec[%d]= %d", i, intr_handle->intr_vec[i]); } } @@ -739,15 +624,75 @@ nfp_configure_rx_interrupt(struct rte_eth_dev *dev, return 0; } +static uint32_t +nfp_check_offloads(struct rte_eth_dev *dev) +{ + struct nfp_net_hw *hw; + struct rte_eth_conf *dev_conf; + struct rte_eth_rxmode *rxmode; + struct rte_eth_txmode *txmode; + uint32_t ctrl = 0; + + hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); + + dev_conf = &dev->data->dev_conf; + rxmode = &dev_conf->rxmode; + txmode = &dev_conf->txmode; + + if (rxmode->offloads & DEV_RX_OFFLOAD_IPV4_CKSUM) { + if (hw->cap & NFP_NET_CFG_CTRL_RXCSUM) + ctrl |= NFP_NET_CFG_CTRL_RXCSUM; + } + + if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) { + if (hw->cap & NFP_NET_CFG_CTRL_RXVLAN) + ctrl |= NFP_NET_CFG_CTRL_RXVLAN; + } + + if (rxmode->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) + hw->mtu = rxmode->max_rx_pkt_len; + + if (txmode->offloads & DEV_TX_OFFLOAD_VLAN_INSERT) + ctrl |= NFP_NET_CFG_CTRL_TXVLAN; + + /* L2 broadcast */ + if (hw->cap & NFP_NET_CFG_CTRL_L2BC) + ctrl |= NFP_NET_CFG_CTRL_L2BC; + + /* L2 multicast */ + if (hw->cap & NFP_NET_CFG_CTRL_L2MC) + ctrl |= NFP_NET_CFG_CTRL_L2MC; + + /* TX checksum offload */ + if (txmode->offloads & DEV_TX_OFFLOAD_IPV4_CKSUM || + txmode->offloads & DEV_TX_OFFLOAD_UDP_CKSUM || + txmode->offloads & DEV_TX_OFFLOAD_TCP_CKSUM) + ctrl |= NFP_NET_CFG_CTRL_TXCSUM; + + /* LSO offload */ + if (txmode->offloads & DEV_TX_OFFLOAD_TCP_TSO) { + if (hw->cap & NFP_NET_CFG_CTRL_LSO) + ctrl |= NFP_NET_CFG_CTRL_LSO; + else + ctrl |= NFP_NET_CFG_CTRL_LSO2; + } + + /* RX gather */ + if (txmode->offloads & DEV_TX_OFFLOAD_MULTI_SEGS) + ctrl |= NFP_NET_CFG_CTRL_GATHER; + + 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; - struct rte_eth_conf *dev_conf; - struct rte_eth_rxmode *rxmode; 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; @@ -758,9 +703,6 @@ nfp_net_start(struct rte_eth_dev *dev) /* Disabling queues just in case... */ nfp_net_disable_queues(dev); - /* Writing configuration parameters in the device */ - nfp_net_params_setup(hw); - /* Enabling the required queues in the device */ nfp_net_enable_queues(dev); @@ -795,21 +737,22 @@ nfp_net_start(struct rte_eth_dev *dev) 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; - /* Checking RX mode */ if (rxmode->mq_mode & ETH_MQ_RX_RSS) { - if (hw->cap & NFP_NET_CFG_CTRL_RSS) { - if (!nfp_net_rss_config_default(dev)) - update |= NFP_NET_CFG_UPDATE_RSS; - } else { - PMD_INIT_LOG(INFO, "RSS not supported"); - return -EINVAL; - } + nfp_net_rss_config_default(dev); + update |= NFP_NET_CFG_UPDATE_RSS; + new_ctrl |= NFP_NET_CFG_CTRL_RSS; } + /* Enable device */ - new_ctrl = hw->ctrl | NFP_NET_CFG_CTRL_ENABLE; + new_ctrl |= NFP_NET_CFG_CTRL_ENABLE; update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING; @@ -831,7 +774,7 @@ nfp_net_start(struct rte_eth_dev *dev) if (hw->is_pf) /* Configure the physical port up */ - nfp_nsp_eth_config(hw->nspu_desc, hw->pf_port_idx, 1); + nfp_eth_set_configured(hw->cpp, hw->pf_port_idx, 1); hw->ctrl = new_ctrl; @@ -882,7 +825,7 @@ nfp_net_stop(struct rte_eth_dev *dev) if (hw->is_pf) /* Configure the physical port down */ - nfp_nsp_eth_config(hw->nspu_desc, hw->pf_port_idx, 0); + nfp_eth_set_configured(hw->cpp, hw->pf_port_idx, 0); } /* Reset and stop device. The device can not be restarted. */ @@ -936,7 +879,7 @@ nfp_net_promisc_enable(struct rte_eth_dev *dev) uint32_t new_ctrl, update = 0; struct nfp_net_hw *hw; - PMD_DRV_LOG(DEBUG, "Promiscuous mode enable\n"); + PMD_DRV_LOG(DEBUG, "Promiscuous mode enable"); hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); @@ -946,7 +889,7 @@ nfp_net_promisc_enable(struct rte_eth_dev *dev) } if (hw->ctrl & NFP_NET_CFG_CTRL_PROMISC) { - PMD_DRV_LOG(INFO, "Promiscuous mode already enabled\n"); + PMD_DRV_LOG(INFO, "Promiscuous mode already enabled"); return; } @@ -972,7 +915,7 @@ nfp_net_promisc_disable(struct rte_eth_dev *dev) hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); if ((hw->ctrl & NFP_NET_CFG_CTRL_PROMISC) == 0) { - PMD_DRV_LOG(INFO, "Promiscuous mode already disabled\n"); + PMD_DRV_LOG(INFO, "Promiscuous mode already disabled"); return; } @@ -999,8 +942,9 @@ static int nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete) { struct nfp_net_hw *hw; - struct rte_eth_link link, old; + struct rte_eth_link link; uint32_t nn_link_status; + int ret; static const uint32_t ls_to_ethtool[] = { [NFP_NET_CFG_STS_LINK_RATE_UNSUPPORTED] = ETH_SPEED_NUM_NONE, @@ -1013,13 +957,10 @@ nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete) [NFP_NET_CFG_STS_LINK_RATE_100G] = ETH_SPEED_NUM_100G, }; - PMD_DRV_LOG(DEBUG, "Link update\n"); + PMD_DRV_LOG(DEBUG, "Link update"); hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); - memset(&old, 0, sizeof(old)); - nfp_net_dev_atomic_read_link_status(dev, &old); - nn_link_status = nn_cfg_readl(hw, NFP_NET_CFG_STS); memset(&link, 0, sizeof(struct rte_eth_link)); @@ -1037,16 +978,14 @@ nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete) else link.link_speed = ls_to_ethtool[nn_link_status]; - if (old.link_status != link.link_status) { - nfp_net_dev_atomic_write_link_status(dev, &link); + ret = rte_eth_linkstatus_set(dev, &link); + if (ret == 0) { if (link.link_status) - PMD_DRV_LOG(INFO, "NIC Link is Up\n"); + PMD_DRV_LOG(INFO, "NIC Link is Up"); else - PMD_DRV_LOG(INFO, "NIC Link is Down\n"); - return 0; + PMD_DRV_LOG(INFO, "NIC Link is Down"); } - - return -1; + return ret; } static int @@ -1214,7 +1153,6 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); - dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev); dev_info->max_rx_queues = (uint16_t)hw->max_rx_queues; dev_info->max_tx_queues = (uint16_t)hw->max_tx_queues; dev_info->min_rx_bufsize = ETHER_MIN_MTU; @@ -1230,6 +1168,9 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) DEV_RX_OFFLOAD_UDP_CKSUM | DEV_RX_OFFLOAD_TCP_CKSUM; + dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_JUMBO_FRAME | + DEV_RX_OFFLOAD_KEEP_CRC; + if (hw->cap & NFP_NET_CFG_CTRL_TXVLAN) dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT; @@ -1238,6 +1179,12 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) DEV_TX_OFFLOAD_UDP_CKSUM | DEV_TX_OFFLOAD_TCP_CKSUM; + if (hw->cap & NFP_NET_CFG_CTRL_LSO_ANY) + dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_TCP_TSO; + + if (hw->cap & NFP_NET_CFG_CTRL_GATHER) + dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_MULTI_SEGS; + dev_info->default_rxconf = (struct rte_eth_rxconf) { .rx_thresh = { .pthresh = DEFAULT_RX_PTHRESH, @@ -1256,8 +1203,6 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) }, .tx_free_thresh = DEFAULT_TX_FREE_THRESH, .tx_rs_thresh = DEFAULT_TX_RSBIT_THRESH, - .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS | - ETH_TXQ_FLAGS_NOOFFLOADS, }; dev_info->flow_type_rss_offloads = ETH_RSS_NONFRAG_IPV4_TCP | @@ -1268,12 +1213,9 @@ nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) dev_info->reta_size = NFP_NET_CFG_RSS_ITBL_SZ; dev_info->hash_key_size = NFP_NET_CFG_RSS_KEY_SZ; - dev_info->speed_capa = ETH_SPEED_NUM_1G | ETH_LINK_SPEED_10G | - ETH_SPEED_NUM_25G | ETH_SPEED_NUM_40G | - ETH_SPEED_NUM_50G | ETH_LINK_SPEED_100G; - - if (hw->cap & NFP_NET_CFG_CTRL_LSO) - dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_TCP_TSO; + dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G | + ETH_LINK_SPEED_25G | ETH_LINK_SPEED_40G | + ETH_LINK_SPEED_50G | ETH_LINK_SPEED_100G; } static const uint32_t * @@ -1376,18 +1318,17 @@ nfp_net_dev_link_status_print(struct rte_eth_dev *dev) struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev); struct rte_eth_link link; - memset(&link, 0, sizeof(link)); - nfp_net_dev_atomic_read_link_status(dev, &link); + rte_eth_linkstatus_get(dev, &link); if (link.link_status) - RTE_LOG(INFO, PMD, "Port %d: Link Up - speed %u Mbps - %s\n", - dev->data->port_id, link.link_speed, - link.link_duplex == ETH_LINK_FULL_DUPLEX - ? "full-duplex" : "half-duplex"); + PMD_DRV_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s", + dev->data->port_id, link.link_speed, + link.link_duplex == ETH_LINK_FULL_DUPLEX + ? "full-duplex" : "half-duplex"); else - RTE_LOG(INFO, PMD, " Port %d: Link Down\n", - dev->data->port_id); + PMD_DRV_LOG(INFO, " Port %d: Link Down", + dev->data->port_id); - RTE_LOG(INFO, PMD, "PCI Address: %04d:%02d:%02d:%d\n", + PMD_DRV_LOG(INFO, "PCI Address: %04d:%02d:%02d:%d", pci_dev->addr.domain, pci_dev->addr.bus, pci_dev->addr.devid, pci_dev->addr.function); } @@ -1428,11 +1369,9 @@ nfp_net_dev_interrupt_handler(void *param) struct rte_eth_link link; struct rte_eth_dev *dev = (struct rte_eth_dev *)param; - PMD_DRV_LOG(DEBUG, "We got a LSC interrupt!!!\n"); + PMD_DRV_LOG(DEBUG, "We got a LSC interrupt!!!"); - /* get the link status */ - memset(&link, 0, sizeof(link)); - nfp_net_dev_atomic_read_link_status(dev, &link); + rte_eth_linkstatus_get(dev, &link); nfp_net_link_update(dev, 0); @@ -1449,7 +1388,7 @@ nfp_net_dev_interrupt_handler(void *param) if (rte_eal_alarm_set(timeout * 1000, nfp_net_dev_interrupt_delayed_handler, (void *)dev) < 0) { - RTE_LOG(ERR, PMD, "Error setting alarm"); + PMD_INIT_LOG(ERR, "Error setting alarm"); /* Unmasking */ nfp_net_irq_unmask(dev); } @@ -1500,9 +1439,9 @@ nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu) /* switch to jumbo mode if needed */ if ((uint32_t)mtu > ETHER_MAX_LEN) - dev->data->dev_conf.rxmode.jumbo_frame = 1; + dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_JUMBO_FRAME; else - dev->data->dev_conf.rxmode.jumbo_frame = 0; + dev->data->dev_conf.rxmode.offloads &= ~DEV_RX_OFFLOAD_JUMBO_FRAME; /* update max frame size */ dev->data->dev_conf.rxmode.max_rx_pkt_len = (uint32_t)mtu; @@ -1534,7 +1473,7 @@ nfp_net_rx_queue_setup(struct rte_eth_dev *dev, if (((nb_desc * sizeof(struct nfp_net_rx_desc)) % 128) != 0 || (nb_desc > NFP_NET_MAX_RX_DESC) || (nb_desc < NFP_NET_MIN_RX_DESC)) { - RTE_LOG(ERR, PMD, "Wrong nb_desc value\n"); + PMD_DRV_LOG(ERR, "Wrong nb_desc value"); return -EINVAL; } @@ -1572,8 +1511,6 @@ nfp_net_rx_queue_setup(struct rte_eth_dev *dev, rxq->rx_count = nb_desc; rxq->port_id = dev->data->port_id; rxq->rx_free_thresh = rx_conf->rx_free_thresh; - rxq->crc_len = (uint8_t) ((dev->data->dev_conf.rxmode.hw_strip_crc) ? 0 - : ETHER_CRC_LEN); rxq->drop_en = rx_conf->rx_drop_en; /* @@ -1587,7 +1524,7 @@ nfp_net_rx_queue_setup(struct rte_eth_dev *dev, socket_id); if (tz == NULL) { - RTE_LOG(ERR, PMD, "Error allocatig rx dma\n"); + PMD_DRV_LOG(ERR, "Error allocatig rx dma"); nfp_net_rx_queue_release(rxq); return -ENOMEM; } @@ -1605,7 +1542,7 @@ nfp_net_rx_queue_setup(struct rte_eth_dev *dev, return -ENOMEM; } - PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64 "\n", + 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); @@ -1630,7 +1567,7 @@ nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq) uint64_t dma_addr; unsigned i; - PMD_RX_LOG(DEBUG, "nfp_net_rx_fill_freelist for %u descriptors\n", + PMD_RX_LOG(DEBUG, "nfp_net_rx_fill_freelist for %u descriptors", rxq->rx_count); for (i = 0; i < rxq->rx_count; i++) { @@ -1638,7 +1575,7 @@ nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq) struct rte_mbuf *mbuf = rte_pktmbuf_alloc(rxq->mem_pool); if (mbuf == NULL) { - RTE_LOG(ERR, PMD, "RX mbuf alloc failed queue_id=%u\n", + PMD_DRV_LOG(ERR, "RX mbuf alloc failed queue_id=%u", (unsigned)rxq->qidx); return -ENOMEM; } @@ -1650,14 +1587,14 @@ nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq) 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 "\n", i, dma_addr); + 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\n", + 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); @@ -1683,7 +1620,7 @@ nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, if (((nb_desc * sizeof(struct nfp_net_tx_desc)) % 128) != 0 || (nb_desc > NFP_NET_MAX_TX_DESC) || (nb_desc < NFP_NET_MIN_TX_DESC)) { - RTE_LOG(ERR, PMD, "Wrong nb_desc value\n"); + PMD_DRV_LOG(ERR, "Wrong nb_desc value"); return -EINVAL; } @@ -1692,10 +1629,10 @@ nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, DEFAULT_TX_FREE_THRESH); if (tx_free_thresh > (nb_desc)) { - RTE_LOG(ERR, PMD, + PMD_DRV_LOG(ERR, "tx_free_thresh must be less than the number of TX " "descriptors. (tx_free_thresh=%u port=%d " - "queue=%d)\n", (unsigned int)tx_free_thresh, + "queue=%d)", (unsigned int)tx_free_thresh, dev->data->port_id, (int)queue_idx); return -(EINVAL); } @@ -1705,7 +1642,7 @@ nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, * calling nfp_net_stop */ if (dev->data->tx_queues[queue_idx]) { - PMD_TX_LOG(DEBUG, "Freeing memory prior to re-allocation %d\n", + 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; @@ -1715,7 +1652,7 @@ nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct nfp_net_txq), RTE_CACHE_LINE_SIZE, socket_id); if (txq == NULL) { - RTE_LOG(ERR, PMD, "Error allocating tx dma\n"); + PMD_DRV_LOG(ERR, "Error allocating tx dma"); return -ENOMEM; } @@ -1729,7 +1666,7 @@ nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN, socket_id); if (tz == NULL) { - RTE_LOG(ERR, PMD, "Error allocating tx dma\n"); + PMD_DRV_LOG(ERR, "Error allocating tx dma"); nfp_net_tx_queue_release(txq); return -ENOMEM; } @@ -1746,7 +1683,6 @@ nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, txq->qcp_q = hw->tx_bar + NFP_QCP_QUEUE_OFF(txq->tx_qcidx); txq->port_id = dev->data->port_id; - txq->txq_flags = tx_conf->txq_flags; /* Saving physical and virtual addresses for the TX ring */ txq->dma = (uint64_t)tz->iova; @@ -1760,7 +1696,7 @@ nfp_net_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, nfp_net_tx_queue_release(txq); return -ENOMEM; } - PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64 "\n", + 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); @@ -1786,7 +1722,7 @@ nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd, uint64_t ol_flags; struct nfp_net_hw *hw = txq->hw; - if (!(hw->cap & NFP_NET_CFG_CTRL_LSO)) + if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY)) goto clean_txd; ol_flags = mb->ol_flags; @@ -1794,15 +1730,19 @@ nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd, if (!(ol_flags & PKT_TX_TCP_SEG)) goto clean_txd; - txd->l4_offset = mb->l2_len + mb->l3_len + mb->l4_len; - txd->lso = rte_cpu_to_le_16(mb->tso_segsz); + 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 = 0; + txd->lso_hdrlen = 0; + txd->mss = 0; } /* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */ @@ -1888,14 +1828,10 @@ nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd, if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS)) return; - if (NFD_CFG_MAJOR_VERSION_of(hw->ver) <= 3) { - 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); - - } else if (NFP_DESC_META_LEN(rxd)) { + /* 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 -----> @@ -1928,7 +1864,11 @@ nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd, return; } } else { - return; + 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; @@ -2019,16 +1959,16 @@ nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) 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(); - rxds = &rxq->rxds[rxq->rd_p]; - if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0) - break; - /* * We got a packet. Let's alloc a new mbuff for refilling the * free descriptor ring as soon as possible @@ -2051,7 +1991,7 @@ nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) mb = rxb->mbuf; rxb->mbuf = new_mb; - PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u\n", + PMD_RX_LOG(DEBUG, "Packet len: %u, mbuf_size: %u", rxds->rxd.data_len, rxq->mbuf_size); /* Size of this segment */ @@ -2089,6 +2029,8 @@ nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) mb->nb_segs = 1; mb->next = NULL; + mb->port = rxq->port_id; + /* Checking the RSS flag */ nfp_net_set_hash(rxq, rxds, mb); @@ -2120,7 +2062,7 @@ nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) if (nb_hold == 0) return nb_hold; - PMD_RX_LOG(DEBUG, "RX port_id=%u queue_id=%u, %d packets received\n", + 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; @@ -2131,7 +2073,7 @@ nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) */ rte_wmb(); if (nb_hold > rxq->rx_free_thresh) { - PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u\n", + 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); @@ -2155,14 +2097,14 @@ nfp_net_tx_free_bufs(struct nfp_net_txq *txq) int todo; PMD_TX_LOG(DEBUG, "queue %u. Check for descriptor with a complete" - " status\n", txq->qidx); + " 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)\n", txq->qidx, + "packets (%u, %u)", txq->qidx, qcp_rd_p, txq->rd_p); return 0; } @@ -2172,7 +2114,7 @@ nfp_net_tx_free_bufs(struct nfp_net_txq *txq) 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\n", + 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) @@ -2226,7 +2168,7 @@ nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) hw = txq->hw; txds = &txq->txds[txq->wr_p]; - PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets\n", + 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))) @@ -2240,7 +2182,7 @@ nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) i = 0; issued_descs = 0; - PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets\n", + PMD_TX_LOG(DEBUG, "queue: %u. Sending %u packets", txq->qidx, nb_pkts); /* Sending packets */ while ((i < nb_pkts) && free_descs) { @@ -2299,7 +2241,7 @@ nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) dma_size = pkt->data_len; dma_addr = rte_mbuf_data_iova(pkt); PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:" - "%" PRIx64 "\n", dma_addr); + "%" PRIx64 "", dma_addr); /* Filling descriptors fields */ txds->dma_len = dma_size; @@ -2314,11 +2256,15 @@ nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) txq->wr_p = 0; pkt_size -= dma_size; - if (!pkt_size) - /* End of packet */ - txds->offset_eop |= PCIE_DESC_TX_EOP; + + /* + * 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 &= PCIE_DESC_TX_OFFSET_MASK; + txds->offset_eop = 0; pkt = pkt->next; /* Referencing next free TX descriptor */ @@ -2349,7 +2295,7 @@ nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask) if ((mask & ETH_VLAN_FILTER_OFFLOAD) || (mask & ETH_VLAN_EXTEND_OFFLOAD)) - RTE_LOG(INFO, PMD, "No support for ETH_VLAN_FILTER_OFFLOAD or" + PMD_DRV_LOG(INFO, "No support for ETH_VLAN_FILTER_OFFLOAD or" " ETH_VLAN_EXTEND_OFFLOAD"); /* Enable vlan strip if it is not configured yet */ @@ -2386,9 +2332,9 @@ nfp_net_rss_reta_write(struct rte_eth_dev *dev, NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private); if (reta_size != NFP_NET_CFG_RSS_ITBL_SZ) { - RTE_LOG(ERR, PMD, "The size of hash lookup table configured " + PMD_DRV_LOG(ERR, "The size of hash lookup table configured " "(%d) doesn't match the number hardware can supported " - "(%d)\n", reta_size, NFP_NET_CFG_RSS_ITBL_SZ); + "(%d)", reta_size, NFP_NET_CFG_RSS_ITBL_SZ); return -EINVAL; } @@ -2467,9 +2413,9 @@ nfp_net_reta_query(struct rte_eth_dev *dev, return -EINVAL; if (reta_size != NFP_NET_CFG_RSS_ITBL_SZ) { - RTE_LOG(ERR, PMD, "The size of hash lookup table configured " + PMD_DRV_LOG(ERR, "The size of hash lookup table configured " "(%d) doesn't match the number hardware can supported " - "(%d)\n", reta_size, NFP_NET_CFG_RSS_ITBL_SZ); + "(%d)", reta_size, NFP_NET_CFG_RSS_ITBL_SZ); return -EINVAL; } @@ -2555,14 +2501,14 @@ nfp_net_rss_hash_update(struct rte_eth_dev *dev, /* Checking if RSS is enabled */ if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS)) { if (rss_hf != 0) { /* Enable RSS? */ - RTE_LOG(ERR, PMD, "RSS unsupported\n"); + PMD_DRV_LOG(ERR, "RSS unsupported"); return -EINVAL; } return 0; /* Nothing to do */ } if (rss_conf->rss_key_len > NFP_NET_CFG_RSS_KEY_SZ) { - RTE_LOG(ERR, PMD, "hash key too long\n"); + PMD_DRV_LOG(ERR, "hash key too long"); return -EINVAL; } @@ -2634,7 +2580,7 @@ nfp_net_rss_config_default(struct rte_eth_dev *dev) uint16_t queue; int i, j, ret; - RTE_LOG(INFO, PMD, "setting default RSS conf for %u queues\n", + PMD_DRV_LOG(INFO, "setting default RSS conf for %u queues", rx_queues); nfp_reta_conf[0].mask = ~0x0; @@ -2654,7 +2600,7 @@ nfp_net_rss_config_default(struct rte_eth_dev *dev) dev_conf = &dev->data->dev_conf; if (!dev_conf) { - RTE_LOG(INFO, PMD, "wrong rss conf"); + PMD_DRV_LOG(INFO, "wrong rss conf"); return -EINVAL; } rss_conf = dev_conf->rx_adv_conf.rss_conf; @@ -2679,6 +2625,7 @@ static const struct eth_dev_ops nfp_net_eth_dev_ops = { .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, @@ -2734,10 +2681,8 @@ nfp_net_init(struct rte_eth_dev *eth_dev) uint64_t tx_bar_off = 0, rx_bar_off = 0; uint32_t start_q; int stride = 4; - - nspu_desc_t *nspu_desc = NULL; - uint64_t bar_offset; int port = 0; + int err; PMD_INIT_FUNC_TRACE(); @@ -2747,18 +2692,17 @@ nfp_net_init(struct rte_eth_dev *eth_dev) (pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC)) { port = get_pf_port_number(eth_dev->data->name); if (port < 0 || port > 7) { - RTE_LOG(ERR, PMD, "Port value is wrong\n"); + PMD_DRV_LOG(ERR, "Port value is wrong"); return -ENODEV; } - PMD_INIT_LOG(DEBUG, "Working with PF port value %d\n", port); + PMD_INIT_LOG(DEBUG, "Working with PF port value %d", port); /* This points to port 0 private data */ hwport0 = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private); /* This points to the specific port private data */ hw = &hwport0[port]; - hw->pf_port_idx = port; } else { hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private); hwport0 = 0; @@ -2786,26 +2730,21 @@ nfp_net_init(struct rte_eth_dev *eth_dev) hw->ctrl_bar = (uint8_t *)pci_dev->mem_resource[0].addr; if (hw->ctrl_bar == NULL) { - RTE_LOG(ERR, PMD, - "hw->ctrl_bar is NULL. BAR0 not configured\n"); + PMD_DRV_LOG(ERR, + "hw->ctrl_bar is NULL. BAR0 not configured"); return -ENODEV; } if (hw->is_pf && port == 0) { - nspu_desc = hw->nspu_desc; - - if (nfp_nsp_map_ctrl_bar(nspu_desc, &bar_offset) != 0) { - /* - * A firmware should be there after PF probe so this - * should not happen. - */ - RTE_LOG(ERR, PMD, "PF BAR symbol resolution failed\n"); - return -ENODEV; + hw->ctrl_bar = nfp_rtsym_map(hw->sym_tbl, "_pf0_net_bar0", + hw->total_ports * 32768, + &hw->ctrl_area); + if (!hw->ctrl_bar) { + printf("nfp_rtsym_map fails for _pf0_net_ctrl_bar"); + return -EIO; } - /* vNIC PF control BAR is a subset of PF PCI device BAR */ - hw->ctrl_bar += bar_offset; - PMD_INIT_LOG(DEBUG, "ctrl bar: %p\n", hw->ctrl_bar); + PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar); } if (port > 0) { @@ -2817,7 +2756,7 @@ nfp_net_init(struct rte_eth_dev *eth_dev) (port * NFP_PF_CSR_SLICE_SIZE); } - PMD_INIT_LOG(DEBUG, "ctrl bar: %p\n", hw->ctrl_bar); + 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); @@ -2828,31 +2767,34 @@ nfp_net_init(struct rte_eth_dev *eth_dev) 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 = NFP_PCIE_QUEUE(start_q); + tx_bar_off = start_q * NFP_QCP_QUEUE_ADDR_SZ; start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_RXQ); - rx_bar_off = NFP_PCIE_QUEUE(start_q); + rx_bar_off = start_q * NFP_QCP_QUEUE_ADDR_SZ; break; default: - RTE_LOG(ERR, PMD, "nfp_net: no device ID matching\n"); - return -ENODEV; + 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 "\n", tx_bar_off); - PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%" PRIx64 "\n", rx_bar_off); + 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_pf && port == 0) { /* configure access to tx/rx vNIC BARs */ - nfp_nsp_map_queues_bar(nspu_desc, &bar_offset); - PMD_INIT_LOG(DEBUG, "tx/rx bar_offset: %" PRIx64 "\n", - bar_offset); - hwport0->hw_queues = (uint8_t *)pci_dev->mem_resource[0].addr; - - /* vNIC PF tx/rx BARs are a subset of PF PCI device */ - hwport0->hw_queues += bar_offset; + hwport0->hw_queues = nfp_cpp_map_area(hw->cpp, 0, 0, + NFP_PCIE_QUEUE(0), + NFP_QCP_QUEUE_AREA_SZ, + &hw->hwqueues_area); + + if (!hwport0->hw_queues) { + printf("nfp_rtsym_map fails for net.qc"); + err = -EIO; + goto dev_err_ctrl_map; + } - /* Lets seize the chance to read eth table from hw */ - if (nfp_nsp_eth_read_table(nspu_desc, &hw->eth_table)) - return -ENODEV; + PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p", + hwport0->hw_queues); } if (hw->is_pf) { @@ -2877,14 +2819,20 @@ nfp_net_init(struct rte_eth_dev *eth_dev) hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU); hw->mtu = 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: %#x, Maximum supported MTU: %d", - hw->ver, hw->max_mtu); - PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s", hw->cap, + 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 " : "", @@ -2894,8 +2842,11 @@ nfp_net_init(struct rte_eth_dev *eth_dev) 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_RSS ? "RSS " : ""); + 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; @@ -2912,7 +2863,8 @@ nfp_net_init(struct rte_eth_dev *eth_dev) eth_dev->data->mac_addrs = rte_zmalloc("mac_addr", ETHER_ADDR_LEN, 0); if (eth_dev->data->mac_addrs == NULL) { PMD_INIT_LOG(ERR, "Failed to space for MAC address"); - return -ENOMEM; + err = -ENOMEM; + goto dev_err_queues_map; } if (hw->is_pf) { @@ -2923,6 +2875,8 @@ nfp_net_init(struct rte_eth_dev *eth_dev) } if (!is_valid_assigned_ether_addr((struct ether_addr *)&hw->mac_addr)) { + PMD_INIT_LOG(INFO, "Using random mac address for port %d", + port); /* Using random mac addresses for VFs */ eth_random_addr(&hw->mac_addr[0]); nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr); @@ -2951,11 +2905,19 @@ nfp_net_init(struct rte_eth_dev *eth_dev) 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 int nfp_pf_create_dev(struct rte_pci_device *dev, int port, int ports, - nfpu_desc_t *nfpu_desc, void **priv) + struct nfp_cpp *cpp, struct nfp_hwinfo *hwinfo, + int phys_port, struct nfp_rtsym_table *sym_tbl, void **priv) { struct rte_eth_dev *eth_dev; struct nfp_net_hw *hw; @@ -2993,12 +2955,16 @@ nfp_pf_create_dev(struct rte_pci_device *dev, int port, int ports, * Then dev_private is adjusted per port. */ hw = (struct nfp_net_hw *)(eth_dev->data->dev_private) + port; - hw->nspu_desc = nfpu_desc->nspu; - hw->nfpu_desc = nfpu_desc; + hw->cpp = cpp; + hw->hwinfo = hwinfo; + hw->sym_tbl = sym_tbl; + hw->pf_port_idx = phys_port; hw->is_pf = 1; if (ports > 1) hw->pf_multiport_enabled = 1; + hw->total_ports = ports; + eth_dev->device = &dev->device; rte_eth_copy_pci_info(eth_dev, dev); @@ -3006,82 +2972,228 @@ nfp_pf_create_dev(struct rte_pci_device *dev, int port, int ports, if (ret) rte_eth_dev_release_port(eth_dev); + else + rte_eth_dev_probing_finish(eth_dev); rte_free(port_name); return ret; } +#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 */ + sprintf(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); + + sprintf(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 */ + sprintf(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 */ + sprintf(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); + + sprintf(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_pf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, struct rte_pci_device *dev) { - nfpu_desc_t *nfpu_desc; - nspu_desc_t *nspu_desc; - uint64_t offset_symbol; - uint8_t *bar_offset; - int major, minor; + struct nfp_cpp *cpp; + struct nfp_hwinfo *hwinfo; + struct nfp_rtsym_table *sym_tbl; + struct nfp_eth_table *nfp_eth_table = NULL; int total_ports; void *priv = 0; int ret = -ENODEV; + int err; int i; if (!dev) return ret; - nfpu_desc = rte_malloc("nfp nfpu", sizeof(nfpu_desc_t), 0); - if (!nfpu_desc) - return -ENOMEM; + /* + * 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 (dev->kdrv == RTE_KDRV_VFIO) + cpp = nfp_cpp_from_device_name(dev, 0); + else + cpp = nfp_cpp_from_device_name(dev, 1); - if (nfpu_open(dev, nfpu_desc, 0) < 0) { - RTE_LOG(ERR, PMD, - "nfpu_open failed\n"); - goto nfpu_error; + if (!cpp) { + PMD_DRV_LOG(ERR, "A CPP handle can not be obtained"); + ret = -EIO; + goto error; } - nspu_desc = nfpu_desc->nspu; + hwinfo = nfp_hwinfo_read(cpp); + if (!hwinfo) { + PMD_DRV_LOG(ERR, "Error reading hwinfo table"); + return -EIO; + } + nfp_eth_table = nfp_eth_read_ports(cpp); + if (!nfp_eth_table) { + PMD_DRV_LOG(ERR, "Error reading NFP ethernet table"); + return -EIO; + } - /* Check NSP ABI version */ - if (nfp_nsp_get_abi_version(nspu_desc, &major, &minor) < 0) { - RTE_LOG(INFO, PMD, "NFP NSP not present\n"); + if (nfp_fw_setup(dev, cpp, nfp_eth_table, hwinfo)) { + PMD_DRV_LOG(INFO, "Error when uploading firmware"); + ret = -EIO; goto error; } - PMD_INIT_LOG(INFO, "nspu ABI version: %d.%d\n", major, minor); - if ((major == 0) && (minor < 20)) { - RTE_LOG(INFO, PMD, "NFP NSP ABI version too old. Required 0.20 or higher\n"); + /* Now the symbol table should be there */ + sym_tbl = nfp_rtsym_table_read(cpp); + if (!sym_tbl) { + PMD_DRV_LOG(ERR, "Something is wrong with the firmware" + " symbol table"); + ret = -EIO; goto error; } - ret = nfp_nsp_fw_setup(nspu_desc, "nfd_cfg_pf0_num_ports", - &offset_symbol); - if (ret) + 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 error; - - bar_offset = (uint8_t *)dev->mem_resource[0].addr; - bar_offset += offset_symbol; - total_ports = (uint32_t)*bar_offset; - PMD_INIT_LOG(INFO, "Total pf ports: %d\n", total_ports); + } + PMD_INIT_LOG(INFO, "Total pf ports: %d", total_ports); if (total_ports <= 0 || total_ports > 8) { - RTE_LOG(ERR, PMD, "nfd_cfg_pf0_num_ports symbol with wrong value"); + PMD_DRV_LOG(ERR, "nfd_cfg_pf0_num_ports symbol with wrong value"); ret = -ENODEV; goto error; } for (i = 0; i < total_ports; i++) { - ret = nfp_pf_create_dev(dev, i, total_ports, nfpu_desc, &priv); + ret = nfp_pf_create_dev(dev, i, total_ports, cpp, hwinfo, + nfp_eth_table->ports[i].index, + sym_tbl, &priv); if (ret) - goto error; + break; } - return 0; - error: - nfpu_close(nfpu_desc); -nfpu_error: - rte_free(nfpu_desc); - + free(nfp_eth_table); return ret; } @@ -3129,8 +3241,19 @@ static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev) if ((pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC) || (pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC)) { port = get_pf_port_number(eth_dev->data->name); + /* + * hotplug is not possible with multiport PF although freeing + * data structures can be done for first port. + */ + if (port != 0) + return -ENOTSUP; hwport0 = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private); hw = &hwport0[port]; + nfp_cpp_area_free(hw->ctrl_area); + nfp_cpp_area_free(hw->hwqueues_area); + free(hw->hwinfo); + free(hw->sym_tbl); + nfp_cpp_free(hw->cpp); } else { hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private); } @@ -3161,9 +3284,7 @@ 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_INIT(nfp_init_log); -static void -nfp_init_log(void) +RTE_INIT(nfp_init_log) { nfp_logtype_init = rte_log_register("pmd.net.nfp.init"); if (nfp_logtype_init >= 0) |