* Netronome vNIC DPDK Poll-Mode Driver: Main entry point
*/
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdint.h>
-#include <unistd.h>
-#include <string.h>
-#include <sys/mman.h>
-#include <sys/socket.h>
-#include <sys/io.h>
-#include <assert.h>
-#include <time.h>
#include <math.h>
-#include <inttypes.h>
#include <rte_byteorder.h>
#include <rte_common.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_ethdev.h>
+#include <rte_ethdev_pci.h>
#include <rte_dev.h>
#include <rte_ether.h>
#include <rte_malloc.h>
#include <rte_version.h>
#include <rte_string_fns.h>
#include <rte_alarm.h>
+#include <rte_spinlock.h>
#include "nfp_net_pmd.h"
#include "nfp_net_logs.h"
/* Prototypes */
static void nfp_net_close(struct rte_eth_dev *dev);
static int nfp_net_configure(struct rte_eth_dev *dev);
+static void nfp_net_dev_interrupt_handler(void *param);
+static void nfp_net_dev_interrupt_delayed_handler(void *param);
+static int nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
+static void nfp_net_infos_get(struct rte_eth_dev *dev,
+ struct rte_eth_dev_info *dev_info);
static int nfp_net_init(struct rte_eth_dev *eth_dev);
static int nfp_net_link_update(struct rte_eth_dev *dev, int wait_to_complete);
+static void nfp_net_promisc_enable(struct rte_eth_dev *dev);
+static void nfp_net_promisc_disable(struct rte_eth_dev *dev);
static int nfp_net_rx_fill_freelist(struct nfp_net_rxq *rxq);
static uint32_t nfp_net_rx_queue_count(struct rte_eth_dev *dev,
uint16_t queue_idx);
nn_writeq(rte_cpu_to_le_64(val), hw->ctrl_bar + off);
}
-/* Creating memzone for hardware rings. */
-static const struct rte_memzone *
-ring_dma_zone_reserve(struct rte_eth_dev *dev, const char *ring_name,
- uint16_t queue_id, uint32_t ring_size, int socket_id)
-{
- char z_name[RTE_MEMZONE_NAMESIZE];
- const struct rte_memzone *mz;
-
- snprintf(z_name, sizeof(z_name), "%s_%s_%d_%d",
- dev->driver->pci_drv.name,
- ring_name, dev->data->port_id, queue_id);
-
- mz = rte_memzone_lookup(z_name);
- if (mz)
- return mz;
-
- return rte_memzone_reserve_aligned(z_name, ring_size, socket_id, 0,
- NFP_MEMZONE_ALIGN);
-}
-
/*
* Atomically reads link status information from global structure rte_eth_dev.
*
nfp_net_reset_rx_queue(struct nfp_net_rxq *rxq)
{
nfp_net_rx_queue_release_mbufs(rxq);
- rxq->wr_p = 0;
rxq->rd_p = 0;
rxq->nb_rx_hold = 0;
}
for (i = 0; i < txq->tx_count; i++) {
if (txq->txbufs[i].mbuf) {
- rte_pktmbuf_free_seg(txq->txbufs[i].mbuf);
+ rte_pktmbuf_free(txq->txbufs[i].mbuf);
txq->txbufs[i].mbuf = NULL;
}
}
nfp_net_tx_queue_release_mbufs(txq);
txq->wr_p = 0;
txq->rd_p = 0;
- txq->tail = 0;
}
static int
if (new == 0)
break;
if (new & NFP_NET_CFG_UPDATE_ERR) {
- PMD_INIT_LOG(ERR, "Reconfig error: 0x%08x\n", new);
+ PMD_INIT_LOG(ERR, "Reconfig error: 0x%08x", new);
return -1;
}
if (cnt >= NFP_NET_POLL_TIMEOUT) {
PMD_INIT_LOG(ERR, "Reconfig timeout for 0x%08x after"
- " %dms\n", update, cnt);
+ " %dms", update, cnt);
rte_panic("Exiting\n");
}
nanosleep(&wait, 0); /* waiting for a 1ms */
PMD_DRV_LOG(DEBUG, "nfp_net_reconfig: ctrl=%08x update=%08x\n",
ctrl, update);
+ rte_spinlock_lock(&hw->reconfig_lock);
+
nn_cfg_writel(hw, NFP_NET_CFG_CTRL, ctrl);
nn_cfg_writel(hw, NFP_NET_CFG_UPDATE, update);
err = __nfp_net_reconfig(hw, update);
+ rte_spinlock_unlock(&hw->reconfig_lock);
+
if (!err)
return 0;
* Reconfig errors imply situations where they can be handled.
* Otherwise, rte_panic is called inside __nfp_net_reconfig
*/
- PMD_INIT_LOG(ERR, "Error nfp_net reconfig for ctrl: %x update: %x\n",
+ PMD_INIT_LOG(ERR, "Error nfp_net reconfig for ctrl: %x update: %x",
ctrl, update);
return -EIO;
}
* called after that internal process
*/
- PMD_INIT_LOG(DEBUG, "Configure\n");
+ PMD_INIT_LOG(DEBUG, "Configure");
dev_conf = &dev->data->dev_conf;
rxmode = &dev_conf->rxmode;
/* Checking TX mode */
if (txmode->mq_mode) {
- PMD_INIT_LOG(INFO, "TX mq_mode DCB and VMDq not supported\n");
+ PMD_INIT_LOG(INFO, "TX mq_mode DCB and VMDq not supported");
return -EINVAL;
}
update = NFP_NET_CFG_UPDATE_RSS;
new_ctrl = NFP_NET_CFG_CTRL_RSS;
} else {
- PMD_INIT_LOG(INFO, "RSS not supported\n");
+ PMD_INIT_LOG(INFO, "RSS not supported");
return -EINVAL;
}
}
if (rxmode->split_hdr_size) {
- PMD_INIT_LOG(INFO, "rxmode does not support split header\n");
+ PMD_INIT_LOG(INFO, "rxmode does not support split header");
return -EINVAL;
}
if (hw->cap & NFP_NET_CFG_CTRL_RXCSUM) {
new_ctrl |= NFP_NET_CFG_CTRL_RXCSUM;
} else {
- PMD_INIT_LOG(INFO, "RXCSUM not supported\n");
+ PMD_INIT_LOG(INFO, "RXCSUM not supported");
return -EINVAL;
}
}
if (rxmode->hw_vlan_filter) {
- PMD_INIT_LOG(INFO, "VLAN filter not supported\n");
+ PMD_INIT_LOG(INFO, "VLAN filter not supported");
return -EINVAL;
}
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\n");
+ PMD_INIT_LOG(INFO, "hw vlan strip not supported");
return -EINVAL;
}
}
if (rxmode->hw_vlan_extend) {
- PMD_INIT_LOG(INFO, "VLAN extended not supported\n");
+ PMD_INIT_LOG(INFO, "VLAN extended not supported");
return -EINVAL;
}
/* this is handled in rte_eth_dev_configure */
if (rxmode->hw_strip_crc) {
- PMD_INIT_LOG(INFO, "strip CRC not supported\n");
+ PMD_INIT_LOG(INFO, "strip CRC not supported");
return -EINVAL;
}
if (rxmode->enable_scatter) {
- PMD_INIT_LOG(INFO, "Scatter not supported\n");
+ PMD_INIT_LOG(INFO, "Scatter not supported");
return -EINVAL;
}
static void
nfp_net_params_setup(struct nfp_net_hw *hw)
{
- uint32_t *mac_address;
-
nn_cfg_writel(hw, NFP_NET_CFG_MTU, hw->mtu);
nn_cfg_writel(hw, NFP_NET_CFG_FLBUFSZ, hw->flbufsz);
-
- /* A MAC address is 8 bytes long */
- mac_address = (uint32_t *)(hw->mac_addr);
-
- nn_cfg_writel(hw, NFP_NET_CFG_MACADDR,
- rte_cpu_to_be_32(*mac_address));
- nn_cfg_writel(hw, NFP_NET_CFG_MACADDR + 4,
- rte_cpu_to_be_32(*(mac_address + 4)));
}
static void
hw->qcp_cfg = hw->tx_bar + NFP_QCP_QUEUE_ADDR_SZ;
}
+static void nfp_net_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));
+
+ tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR + 4));
+ memcpy(&hw->mac_addr[4], &tmp, 2);
+}
+
+static int
+nfp_configure_rx_interrupt(struct rte_eth_dev *dev,
+ struct rte_intr_handle *intr_handle)
+{
+ struct nfp_net_hw *hw;
+ int i;
+
+ if (!intr_handle->intr_vec) {
+ intr_handle->intr_vec =
+ rte_zmalloc("intr_vec",
+ dev->data->nb_rx_queues * sizeof(int), 0);
+ if (!intr_handle->intr_vec) {
+ PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
+ " intr_vec", dev->data->nb_rx_queues);
+ return -ENOMEM;
+ }
+ }
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+ PMD_INIT_LOG(INFO, "VF: enabling RX interrupt with UIO");
+ /* UIO just supports one queue and no LSC*/
+ nn_cfg_writeb(hw, NFP_NET_CFG_RXR_VEC(0), 0);
+ } else {
+ PMD_INIT_LOG(INFO, "VF: enabling RX interrupt with VFIO");
+ for (i = 0; i < dev->data->nb_rx_queues; i++)
+ /*
+ * The first msix vector is reserved for non
+ * efd interrupts
+ */
+ nn_cfg_writeb(hw, NFP_NET_CFG_RXR_VEC(i), i + 1);
+ }
+
+ /* Avoiding TX interrupts */
+ hw->ctrl |= NFP_NET_CFG_CTRL_MSIX_TX_OFF;
+ return 0;
+}
+
static int
nfp_net_start(struct rte_eth_dev *dev)
{
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
uint32_t new_ctrl, update = 0;
struct nfp_net_hw *hw;
+ uint32_t intr_vector;
int ret;
hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
- PMD_INIT_LOG(DEBUG, "Start\n");
+ PMD_INIT_LOG(DEBUG, "Start");
/* Disabling queues just in case... */
nfp_net_disable_queues(dev);
/* Enabling the required queues in the device */
nfp_net_enable_queues(dev);
+ /* check and configure queue intr-vector mapping */
+ if (dev->data->dev_conf.intr_conf.rxq != 0) {
+ if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+ /*
+ * Better not to share LSC with RX interrupts.
+ * Unregistering LSC interrupt handler
+ */
+ rte_intr_callback_unregister(&pci_dev->intr_handle,
+ nfp_net_dev_interrupt_handler, (void *)dev);
+
+ if (dev->data->nb_rx_queues > 1) {
+ PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+ "supports 1 queue with UIO");
+ return -EIO;
+ }
+ }
+ intr_vector = dev->data->nb_rx_queues;
+ if (rte_intr_efd_enable(intr_handle, intr_vector))
+ return -1;
+ }
+
+ if (rte_intr_dp_is_en(intr_handle))
+ nfp_configure_rx_interrupt(dev, intr_handle);
+
+ rte_intr_enable(intr_handle);
+
/* Enable device */
- new_ctrl = hw->ctrl | NFP_NET_CFG_CTRL_ENABLE | NFP_NET_CFG_UPDATE_MSIX;
+ new_ctrl = hw->ctrl | NFP_NET_CFG_CTRL_ENABLE;
update = NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+ /* Just configuring queues interrupts when necessary */
+ if (rte_intr_dp_is_en(intr_handle))
+ update |= NFP_NET_CFG_UPDATE_MSIX;
+
if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
{
int i;
- PMD_INIT_LOG(DEBUG, "Stop\n");
+ PMD_INIT_LOG(DEBUG, "Stop");
nfp_net_disable_queues(dev);
nfp_net_close(struct rte_eth_dev *dev)
{
struct nfp_net_hw *hw;
+ struct rte_pci_device *pci_dev;
- PMD_INIT_LOG(DEBUG, "Close\n");
+ PMD_INIT_LOG(DEBUG, "Close");
hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ pci_dev = RTE_ETH_DEV_TO_PCI(dev);
/*
* We assume that the DPDK application is stopping all the
nfp_net_stop(dev);
+ rte_intr_disable(&pci_dev->intr_handle);
nn_cfg_writeb(hw, NFP_NET_CFG_LSC, 0xff);
+ /* unregister callback func from eal lib */
+ rte_intr_callback_unregister(&pci_dev->intr_handle,
+ nfp_net_dev_interrupt_handler,
+ (void *)dev);
+
/*
* The ixgbe PMD driver disables the pcie master on the
* device. The i40e does not...
*/
}
+static void
+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");
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ if (!(hw->cap & NFP_NET_CFG_CTRL_PROMISC)) {
+ PMD_INIT_LOG(INFO, "Promiscuous mode not supported");
+ return;
+ }
+
+ if (hw->ctrl & NFP_NET_CFG_CTRL_PROMISC) {
+ PMD_DRV_LOG(INFO, "Promiscuous mode already enabled\n");
+ return;
+ }
+
+ new_ctrl = hw->ctrl | NFP_NET_CFG_CTRL_PROMISC;
+ update = NFP_NET_CFG_UPDATE_GEN;
+
+ /*
+ * DPDK sets promiscuous mode on just after this call assuming
+ * it can not fail ...
+ */
+ if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+ return;
+
+ hw->ctrl = new_ctrl;
+}
+
+static void
+nfp_net_promisc_disable(struct rte_eth_dev *dev)
+{
+ uint32_t new_ctrl, update = 0;
+ struct nfp_net_hw *hw;
+
+ 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");
+ return;
+ }
+
+ new_ctrl = hw->ctrl & ~NFP_NET_CFG_CTRL_PROMISC;
+ update = NFP_NET_CFG_UPDATE_GEN;
+
+ /*
+ * DPDK sets promiscuous mode off just before this call
+ * assuming it can not fail ...
+ */
+ if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+ return;
+
+ hw->ctrl = new_ctrl;
+}
+
/*
* return 0 means link status changed, -1 means not changed
*
struct rte_eth_link link, old;
uint32_t nn_link_status;
+ static const uint32_t ls_to_ethtool[] = {
+ [NFP_NET_CFG_STS_LINK_RATE_UNSUPPORTED] = ETH_SPEED_NUM_NONE,
+ [NFP_NET_CFG_STS_LINK_RATE_UNKNOWN] = ETH_SPEED_NUM_NONE,
+ [NFP_NET_CFG_STS_LINK_RATE_1G] = ETH_SPEED_NUM_1G,
+ [NFP_NET_CFG_STS_LINK_RATE_10G] = ETH_SPEED_NUM_10G,
+ [NFP_NET_CFG_STS_LINK_RATE_25G] = ETH_SPEED_NUM_25G,
+ [NFP_NET_CFG_STS_LINK_RATE_40G] = ETH_SPEED_NUM_40G,
+ [NFP_NET_CFG_STS_LINK_RATE_50G] = ETH_SPEED_NUM_50G,
+ [NFP_NET_CFG_STS_LINK_RATE_100G] = ETH_SPEED_NUM_100G,
+ };
+
PMD_DRV_LOG(DEBUG, "Link update\n");
hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
memset(&link, 0, sizeof(struct rte_eth_link));
if (nn_link_status & NFP_NET_CFG_STS_LINK)
- link.link_status = 1;
+ link.link_status = ETH_LINK_UP;
link.link_duplex = ETH_LINK_FULL_DUPLEX;
- /* Other cards can limit the tx and rx rate per VF */
- link.link_speed = ETH_LINK_SPEED_40G;
+
+ nn_link_status = (nn_link_status >> NFP_NET_CFG_STS_LINK_RATE_SHIFT) &
+ NFP_NET_CFG_STS_LINK_RATE_MASK;
+
+ if ((NFD_CFG_MAJOR_VERSION_of(hw->ver) < 4) ||
+ ((NFD_CFG_MINOR_VERSION_of(hw->ver) == 4) &&
+ (NFD_CFG_MINOR_VERSION_of(hw->ver) == 0)))
+ /* We really do not know the speed wil old firmware */
+ link.link_speed = ETH_SPEED_NUM_NONE;
+ else {
+ if (nn_link_status >= RTE_DIM(ls_to_ethtool))
+ link.link_speed = ETH_SPEED_NUM_NONE;
+ 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);
nfp_dev_stats.obytes -= hw->eth_stats_base.obytes;
- nfp_dev_stats.imcasts =
- nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
- nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
-
/* reading general device stats */
nfp_dev_stats.ierrors =
nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_ERRORS);
nfp_dev_stats.oerrors -= hw->eth_stats_base.oerrors;
- /* Multicast frames received */
- nfp_dev_stats.imcasts =
- nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
- nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
-
/* RX ring mbuf allocation failures */
nfp_dev_stats.rx_nombuf = dev->data->rx_mbuf_alloc_failed;
hw->eth_stats_base.obytes =
nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_OCTETS);
- hw->eth_stats_base.imcasts =
- nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
/* reading general device stats */
hw->eth_stats_base.ierrors =
nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_ERRORS);
hw->eth_stats_base.oerrors =
nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_ERRORS);
- /* Multicast frames received */
- hw->eth_stats_base.imcasts =
- nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
/* RX ring mbuf allocation failures */
dev->data->rx_mbuf_alloc_failed = 0;
nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_DISCARDS);
}
+static void
+nfp_net_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
+{
+ struct nfp_net_hw *hw;
+
+ 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;
+ dev_info->max_rx_pktlen = hw->mtu;
+ /* Next should change when PF support is implemented */
+ dev_info->max_mac_addrs = 1;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_RXVLAN)
+ dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_RXCSUM)
+ dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_IPV4_CKSUM |
+ DEV_RX_OFFLOAD_UDP_CKSUM |
+ DEV_RX_OFFLOAD_TCP_CKSUM;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)
+ dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT;
+
+ if (hw->cap & NFP_NET_CFG_CTRL_TXCSUM)
+ dev_info->tx_offload_capa |= DEV_TX_OFFLOAD_IPV4_CKSUM |
+ DEV_TX_OFFLOAD_UDP_CKSUM |
+ DEV_TX_OFFLOAD_TCP_CKSUM;
+
+ dev_info->default_rxconf = (struct rte_eth_rxconf) {
+ .rx_thresh = {
+ .pthresh = DEFAULT_RX_PTHRESH,
+ .hthresh = DEFAULT_RX_HTHRESH,
+ .wthresh = DEFAULT_RX_WTHRESH,
+ },
+ .rx_free_thresh = DEFAULT_RX_FREE_THRESH,
+ .rx_drop_en = 0,
+ };
+
+ dev_info->default_txconf = (struct rte_eth_txconf) {
+ .tx_thresh = {
+ .pthresh = DEFAULT_TX_PTHRESH,
+ .hthresh = DEFAULT_TX_HTHRESH,
+ .wthresh = DEFAULT_TX_WTHRESH,
+ },
+ .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->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;
+}
+
+static const uint32_t *
+nfp_net_supported_ptypes_get(struct rte_eth_dev *dev)
+{
+ static const uint32_t ptypes[] = {
+ /* refers to nfp_net_set_hash() */
+ RTE_PTYPE_INNER_L3_IPV4,
+ RTE_PTYPE_INNER_L3_IPV6,
+ RTE_PTYPE_INNER_L3_IPV6_EXT,
+ RTE_PTYPE_INNER_L4_MASK,
+ RTE_PTYPE_UNKNOWN
+ };
+
+ if (dev->rx_pkt_burst == nfp_net_recv_pkts)
+ return ptypes;
+ return NULL;
+}
+
static uint32_t
nfp_net_rx_queue_count(struct rte_eth_dev *dev, uint16_t queue_idx)
{
rxq = (struct nfp_net_rxq *)dev->data->rx_queues[queue_idx];
- if (rxq == NULL) {
- PMD_INIT_LOG(ERR, "Bad queue: %u\n", queue_idx);
- return 0;
- }
-
- idx = rxq->rd_p % rxq->rx_count;
- rxds = &rxq->rxds[idx];
+ idx = rxq->rd_p;
count = 0;
* Other PMDs are just checking the DD bit in intervals of 4
* descriptors and counting all four if the first has the DD
* bit on. Of course, this is not accurate but can be good for
- * perfomance. But ideally that should be done in descriptors
+ * performance. But ideally that should be done in descriptors
* chunks belonging to the same cache line
*/
return count;
}
+static int
+nfp_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
+{
+ struct rte_pci_device *pci_dev;
+ struct nfp_net_hw *hw;
+ int base = 0;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+ if (pci_dev->intr_handle.type != RTE_INTR_HANDLE_UIO)
+ base = 1;
+
+ /* Make sure all updates are written before un-masking */
+ rte_wmb();
+ nn_cfg_writeb(hw, NFP_NET_CFG_ICR(base + queue_id),
+ NFP_NET_CFG_ICR_UNMASKED);
+ return 0;
+}
+
+static int
+nfp_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
+{
+ struct rte_pci_device *pci_dev;
+ struct nfp_net_hw *hw;
+ int base = 0;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+ if (pci_dev->intr_handle.type != RTE_INTR_HANDLE_UIO)
+ base = 1;
+
+ /* Make sure all updates are written before un-masking */
+ rte_wmb();
+ nn_cfg_writeb(hw, NFP_NET_CFG_ICR(base + queue_id), 0x1);
+ return 0;
+}
+
+static void
+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);
+ if (link.link_status)
+ RTE_LOG(INFO, PMD, "Port %d: Link Up - speed %u Mbps - %s\n",
+ (int)(dev->data->port_id), (unsigned)link.link_speed,
+ link.link_duplex == ETH_LINK_FULL_DUPLEX
+ ? "full-duplex" : "half-duplex");
+ else
+ RTE_LOG(INFO, PMD, " Port %d: Link Down\n",
+ (int)(dev->data->port_id));
+
+ RTE_LOG(INFO, PMD, "PCI Address: %04d:%02d:%02d:%d\n",
+ pci_dev->addr.domain, pci_dev->addr.bus,
+ pci_dev->addr.devid, pci_dev->addr.function);
+}
+
+/* Interrupt configuration and handling */
+
+/*
+ * nfp_net_irq_unmask - Unmask an interrupt
+ *
+ * If MSI-X auto-masking is enabled clear the mask bit, otherwise
+ * clear the ICR for the entry.
+ */
+static void
+nfp_net_irq_unmask(struct rte_eth_dev *dev)
+{
+ struct nfp_net_hw *hw;
+ struct rte_pci_device *pci_dev;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
+ if (hw->ctrl & NFP_NET_CFG_CTRL_MSIXAUTO) {
+ /* If MSI-X auto-masking is used, clear the entry */
+ rte_wmb();
+ rte_intr_enable(&pci_dev->intr_handle);
+ } else {
+ /* Make sure all updates are written before un-masking */
+ rte_wmb();
+ nn_cfg_writeb(hw, NFP_NET_CFG_ICR(NFP_NET_IRQ_LSC_IDX),
+ NFP_NET_CFG_ICR_UNMASKED);
+ }
+}
+
+static void
+nfp_net_dev_interrupt_handler(void *param)
+{
+ int64_t timeout;
+ 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");
+
+ /* get the link status */
+ memset(&link, 0, sizeof(link));
+ nfp_net_dev_atomic_read_link_status(dev, &link);
+
+ nfp_net_link_update(dev, 0);
+
+ /* likely to up */
+ if (!link.link_status) {
+ /* handle it 1 sec later, wait it being stable */
+ timeout = NFP_NET_LINK_UP_CHECK_TIMEOUT;
+ /* likely to down */
+ } else {
+ /* handle it 4 sec later, wait it being stable */
+ timeout = NFP_NET_LINK_DOWN_CHECK_TIMEOUT;
+ }
+
+ if (rte_eal_alarm_set(timeout * 1000,
+ nfp_net_dev_interrupt_delayed_handler,
+ (void *)dev) < 0) {
+ RTE_LOG(ERR, PMD, "Error setting alarm");
+ /* Unmasking */
+ nfp_net_irq_unmask(dev);
+ }
+}
+
+/*
+ * Interrupt handler which shall be registered for alarm callback for delayed
+ * handling specific interrupt to wait for the stable nic state. As the NIC
+ * interrupt state is not stable for nfp after link is just down, it needs
+ * to wait 4 seconds to get the stable status.
+ *
+ * @param handle Pointer to interrupt handle.
+ * @param param The address of parameter (struct rte_eth_dev *)
+ *
+ * @return void
+ */
+static void
+nfp_net_dev_interrupt_delayed_handler(void *param)
+{
+ struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
+
+ nfp_net_link_update(dev, 0);
+ _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
+
+ nfp_net_dev_link_status_print(dev);
+
+ /* Unmasking */
+ nfp_net_irq_unmask(dev);
+}
+
+static int
+nfp_net_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
+{
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ /* check that mtu is within the allowed range */
+ if ((mtu < ETHER_MIN_MTU) || ((uint32_t)mtu > hw->max_mtu))
+ return -EINVAL;
+
+ /* switch to jumbo mode if needed */
+ if ((uint32_t)mtu > ETHER_MAX_LEN)
+ dev->data->dev_conf.rxmode.jumbo_frame = 1;
+ else
+ dev->data->dev_conf.rxmode.jumbo_frame = 0;
+
+ /* update max frame size */
+ dev->data->dev_conf.rxmode.max_rx_pkt_len = (uint32_t)mtu;
+
+ /* writing to configuration space */
+ nn_cfg_writel(hw, NFP_NET_CFG_MTU, (uint32_t)mtu);
+
+ hw->mtu = mtu;
+
+ return 0;
+}
+
static int
nfp_net_rx_queue_setup(struct rte_eth_dev *dev,
uint16_t queue_idx, uint16_t nb_desc,
(nb_desc > NFP_NET_MAX_RX_DESC) ||
(nb_desc < NFP_NET_MIN_RX_DESC)) {
RTE_LOG(ERR, PMD, "Wrong nb_desc value\n");
- return (-EINVAL);
+ return -EINVAL;
}
/*
rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct nfp_net_rxq),
RTE_CACHE_LINE_SIZE, socket_id);
if (rxq == NULL)
- return (-ENOMEM);
+ return -ENOMEM;
/* Hw queues mapping based on firmware confifguration */
rxq->qidx = queue_idx;
* handle the maximum ring size is allocated in order to allow for
* resizing in later calls to the queue setup function.
*/
- tz = ring_dma_zone_reserve(dev, "rx_ring", queue_idx,
+ tz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx,
sizeof(struct nfp_net_rx_desc) *
- NFP_NET_MAX_RX_DESC, socket_id);
+ NFP_NET_MAX_RX_DESC, NFP_MEMZONE_ALIGN,
+ socket_id);
if (tz == NULL) {
RTE_LOG(ERR, PMD, "Error allocatig rx dma\n");
nfp_net_rx_queue_release(rxq);
- return (-ENOMEM);
+ return -ENOMEM;
}
/* Saving physical and virtual addresses for the RX ring */
RTE_CACHE_LINE_SIZE, socket_id);
if (rxq->rxbufs == NULL) {
nfp_net_rx_queue_release(rxq);
- return (-ENOMEM);
+ return -ENOMEM;
}
PMD_RX_LOG(DEBUG, "rxbufs=%p hw_ring=%p dma_addr=0x%" PRIx64 "\n",
if (mbuf == NULL) {
RTE_LOG(ERR, PMD, "RX mbuf alloc failed queue_id=%u\n",
(unsigned)rxq->qidx);
- return (-ENOMEM);
+ return -ENOMEM;
}
dma_addr = rte_cpu_to_le_64(RTE_MBUF_DMA_ADDR_DEFAULT(mbuf));
rxd->fld.dma_addr_lo = dma_addr & 0xffffffff;
rxe[i].mbuf = mbuf;
PMD_RX_LOG(DEBUG, "[%d]: %" PRIx64 "\n", i, dma_addr);
-
- rxq->wr_p++;
}
/* Make sure all writes are flushed before telling the hardware */
RTE_CACHE_LINE_SIZE, socket_id);
if (txq == NULL) {
RTE_LOG(ERR, PMD, "Error allocating tx dma\n");
- return (-ENOMEM);
+ return -ENOMEM;
}
/*
* handle the maximum ring size is allocated in order to allow for
* resizing in later calls to the queue setup function.
*/
- tz = ring_dma_zone_reserve(dev, "tx_ring", queue_idx,
+ tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx,
sizeof(struct nfp_net_tx_desc) *
- NFP_NET_MAX_TX_DESC, socket_id);
+ NFP_NET_MAX_TX_DESC, NFP_MEMZONE_ALIGN,
+ socket_id);
if (tz == NULL) {
RTE_LOG(ERR, PMD, "Error allocating tx dma\n");
nfp_net_tx_queue_release(txq);
- return (-ENOMEM);
+ return -ENOMEM;
}
txq->tx_count = nb_desc;
- txq->tail = 0;
txq->tx_free_thresh = tx_free_thresh;
txq->tx_pthresh = tx_conf->tx_thresh.pthresh;
txq->tx_hthresh = tx_conf->tx_thresh.hthresh;
RTE_CACHE_LINE_SIZE, socket_id);
if (txq->txbufs == NULL) {
nfp_net_tx_queue_release(txq);
- return (-ENOMEM);
+ return -ENOMEM;
}
PMD_TX_LOG(DEBUG, "txbufs=%p hw_ring=%p dma_addr=0x%" PRIx64 "\n",
txq->txbufs, txq->txds, (unsigned long int)txq->dma);
return 0;
}
+/* nfp_net_tx_tso - Set TX descriptor for TSO */
+static inline void
+nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+ struct rte_mbuf *mb)
+{
+ uint64_t ol_flags;
+ struct nfp_net_hw *hw = txq->hw;
+
+ if (!(hw->cap & NFP_NET_CFG_CTRL_LSO))
+ goto clean_txd;
+
+ ol_flags = mb->ol_flags;
+
+ 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->flags = PCIE_DESC_TX_LSO;
+ return;
+
+clean_txd:
+ txd->flags = 0;
+ txd->l4_offset = 0;
+ txd->lso = 0;
+}
+
/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
static inline void
nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
struct rte_mbuf *mb)
{
- uint16_t ol_flags;
+ uint64_t ol_flags;
struct nfp_net_hw *hw = txq->hw;
if (!(hw->cap & NFP_NET_CFG_CTRL_TXCSUM))
break;
}
- txd->flags |= PCIE_DESC_TX_CSUM;
+ if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
+ txd->flags |= PCIE_DESC_TX_CSUM;
}
/* nfp_net_rx_cksum - set mbuf checksum flags based on RX descriptor flags */
hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
- /*
- * hash type is sharing the same word with input port info
- * 31-8: input port
- * 7:0: hash type
- */
- hash_type &= 0xff;
mbuf->hash.rss = hash;
mbuf->ol_flags |= PKT_RX_RSS_HASH;
}
}
-/* nfp_net_check_port - Set mbuf in_port field */
-static void
-nfp_net_check_port(struct nfp_net_rx_desc *rxd, struct rte_mbuf *mbuf)
-{
- uint32_t port;
-
- if (!(rxd->rxd.flags & PCIE_DESC_RX_INGRESS_PORT)) {
- mbuf->port = 0;
- return;
- }
-
- port = rte_be_to_cpu_32(*(uint32_t *)((uint8_t *)mbuf->buf_addr +
- mbuf->data_off - 8));
-
- /*
- * hash type is sharing the same word with input port info
- * 31-8: input port
- * 7:0: hash type
- */
- port = (uint8_t)(port >> 8);
- mbuf->port = port;
-}
-
static inline void
nfp_net_mbuf_alloc_failed(struct nfp_net_rxq *rxq)
{
struct nfp_net_hw *hw;
struct rte_mbuf *mb;
struct rte_mbuf *new_mb;
- int idx;
uint16_t nb_hold;
uint64_t dma_addr;
int avail;
* DPDK just checks the queue is lower than max queues
* enabled. But the queue needs to be configured
*/
- RTE_LOG(ERR, PMD, "RX Bad queue\n");
+ RTE_LOG_DP(ERR, PMD, "RX Bad queue\n");
return -EINVAL;
}
nb_hold = 0;
while (avail < nb_pkts) {
- idx = rxq->rd_p % rxq->rx_count;
-
- rxb = &rxq->rxbufs[idx];
+ rxb = &rxq->rxbufs[rxq->rd_p];
if (unlikely(rxb == NULL)) {
- RTE_LOG(ERR, PMD, "rxb does not exist!\n");
+ RTE_LOG_DP(ERR, PMD, "rxb does not exist!\n");
break;
}
*/
rte_rmb();
- rxds = &rxq->rxds[idx];
+ rxds = &rxq->rxds[rxq->rd_p];
if ((rxds->rxd.meta_len_dd & PCIE_DESC_RX_DD) == 0)
break;
*/
new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
if (unlikely(new_mb == NULL)) {
- RTE_LOG(DEBUG, PMD, "RX mbuf alloc failed port_id=%u "
+ RTE_LOG_DP(DEBUG, PMD, "RX mbuf alloc failed port_id=%u "
"queue_id=%u\n", (unsigned)rxq->port_id,
(unsigned)rxq->qidx);
nfp_net_mbuf_alloc_failed(rxq);
* responsibility of avoiding it. But we have
* to give some info about the error
*/
- RTE_LOG(ERR, PMD,
+ RTE_LOG_DP(ERR, PMD,
"mbuf overflow likely due to the RX offset.\n"
"\t\tYour mbuf size should have extra space for"
" RX offset=%u bytes.\n"
/* Checking the checksum flag */
nfp_net_rx_cksum(rxq, rxds, mb);
- /* Checking the port flag */
- nfp_net_check_port(rxds, mb);
-
if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
(hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
- mb->ol_flags |= PKT_RX_VLAN_PKT;
+ mb->ol_flags |= PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED;
}
/* Adding the mbuff to the mbuff array passed by the app */
rxds->fld.dma_addr_lo = dma_addr & 0xffffffff;
rxq->rd_p++;
+ if (unlikely(rxq->rd_p == rxq->rx_count)) /* wrapping?*/
+ rxq->rd_p = 0;
}
if (nb_hold == 0)
/* 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->qcp_rd_p) {
+ 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,
- qcp_rd_p, txq->qcp_rd_p);
+ qcp_rd_p, txq->rd_p);
return 0;
}
- if (qcp_rd_p > txq->qcp_rd_p)
- todo = qcp_rd_p - txq->qcp_rd_p;
+ if (qcp_rd_p > txq->rd_p)
+ todo = qcp_rd_p - txq->rd_p;
else
- todo = qcp_rd_p + txq->tx_count - txq->qcp_rd_p;
+ todo = qcp_rd_p + txq->tx_count - txq->rd_p;
- PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->qcp_rd_p: %u, qcp->rd_p: %u\n",
- qcp_rd_p, txq->qcp_rd_p, txq->rd_p);
+ PMD_TX_LOG(DEBUG, "qcp_rd_p %u, txq->rd_p: %u, qcp->rd_p: %u\n",
+ qcp_rd_p, txq->rd_p, txq->rd_p);
if (todo == 0)
return todo;
- txq->qcp_rd_p += todo;
- txq->qcp_rd_p %= txq->tx_count;
txq->rd_p += todo;
+ if (unlikely(txq->rd_p >= txq->tx_count))
+ txq->rd_p -= txq->tx_count;
return todo;
}
/* Leaving always free descriptors for avoiding wrapping confusion */
-#define NFP_FREE_TX_DESC(t) (t->tx_count - (t->wr_p - t->rd_p) - 8)
+static inline
+uint32_t nfp_free_tx_desc(struct nfp_net_txq *txq)
+{
+ if (txq->wr_p >= txq->rd_p)
+ return txq->tx_count - (txq->wr_p - txq->rd_p) - 8;
+ else
+ return txq->rd_p - txq->wr_p - 8;
+}
/*
* nfp_net_txq_full - Check if the TX queue free descriptors
* This function uses the host copy* of read/write pointers
*/
static inline
-int nfp_net_txq_full(struct nfp_net_txq *txq)
+uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
{
- return NFP_FREE_TX_DESC(txq) < txq->tx_free_thresh;
+ return (nfp_free_tx_desc(txq) < txq->tx_free_thresh);
}
static uint16_t
{
struct nfp_net_txq *txq;
struct nfp_net_hw *hw;
- struct nfp_net_tx_desc *txds;
+ struct nfp_net_tx_desc *txds, txd;
struct rte_mbuf *pkt;
uint64_t dma_addr;
int pkt_size, dma_size;
txq = tx_queue;
hw = txq->hw;
- txds = &txq->txds[txq->tail];
+ txds = &txq->txds[txq->wr_p];
PMD_TX_LOG(DEBUG, "working for queue %u at pos %d and %u packets\n",
- txq->qidx, txq->tail, nb_pkts);
+ txq->qidx, txq->wr_p, nb_pkts);
- if ((NFP_FREE_TX_DESC(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
+ if ((nfp_free_tx_desc(txq) < nb_pkts) || (nfp_net_txq_full(txq)))
nfp_net_tx_free_bufs(txq);
- free_descs = (uint16_t)NFP_FREE_TX_DESC(txq);
+ free_descs = (uint16_t)nfp_free_tx_desc(txq);
if (unlikely(free_descs == 0))
return 0;
/* Sending packets */
while ((i < nb_pkts) && free_descs) {
/* Grabbing the mbuf linked to the current descriptor */
- lmbuf = &txq->txbufs[txq->tail].mbuf;
+ lmbuf = &txq->txbufs[txq->wr_p].mbuf;
/* Warming the cache for releasing the mbuf later on */
RTE_MBUF_PREFETCH_TO_FREE(*lmbuf);
if (unlikely((pkt->nb_segs > 1) &&
!(hw->cap & NFP_NET_CFG_CTRL_GATHER))) {
- PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set\n");
+ PMD_INIT_LOG(INFO, "NFP_NET_CFG_CTRL_GATHER not set");
rte_panic("Multisegment packet unsupported\n");
}
/*
* Checksum and VLAN flags just in the first descriptor for a
- * multisegment packet
+ * multisegment packet, but TSO info needs to be in all of them.
*/
- nfp_net_tx_cksum(txq, txds, pkt);
+ txd.data_len = pkt->pkt_len;
+ nfp_net_tx_tso(txq, &txd, pkt);
+ nfp_net_tx_cksum(txq, &txd, pkt);
if ((pkt->ol_flags & PKT_TX_VLAN_PKT) &&
(hw->cap & NFP_NET_CFG_CTRL_TXVLAN)) {
- txds->flags |= PCIE_DESC_TX_VLAN;
- txds->vlan = pkt->vlan_tci;
+ txd.flags |= PCIE_DESC_TX_VLAN;
+ txd.vlan = pkt->vlan_tci;
}
- if (pkt->ol_flags & PKT_TX_TCP_SEG)
- rte_panic("TSO is not supported\n");
-
/*
* mbuf data_len is the data in one segment and pkt_len data
* in the whole packet. When the packet is just one segment,
pkt_size = pkt->pkt_len;
while (pkt_size) {
- /* Releasing mbuf which was prefetched above */
+ /* Copying TSO, VLAN and cksum info */
+ *txds = txd;
+
+ /* Releasing mbuf used by this descriptor previously*/
if (*lmbuf)
rte_pktmbuf_free_seg(*lmbuf);
+ /*
+ * Linking mbuf with descriptor for being released
+ * next time descriptor is used
+ */
+ *lmbuf = pkt;
+
dma_size = pkt->data_len;
- dma_addr = RTE_MBUF_DATA_DMA_ADDR(pkt);
+ dma_addr = rte_mbuf_data_dma_addr(pkt);
PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
"%" PRIx64 "\n", dma_addr);
/* Filling descriptors fields */
txds->dma_len = dma_size;
- txds->data_len = pkt->pkt_len;
+ txds->data_len = txd.data_len;
txds->dma_addr_hi = (dma_addr >> 32) & 0xff;
txds->dma_addr_lo = (dma_addr & 0xffffffff);
ASSERT(free_descs > 0);
free_descs--;
- /*
- * Linking mbuf with descriptor for being released
- * next time descriptor is used
- */
- *lmbuf = pkt;
-
txq->wr_p++;
- txq->tail++;
- if (unlikely(txq->tail == txq->tx_count)) /* wrapping?*/
- txq->tail = 0;
+ if (unlikely(txq->wr_p == txq->tx_count)) /* wrapping?*/
+ txq->wr_p = 0;
pkt_size -= dma_size;
if (!pkt_size) {
pkt = pkt->next;
}
/* Referencing next free TX descriptor */
- txds = &txq->txds[txq->tail];
+ txds = &txq->txds[txq->wr_p];
+ lmbuf = &txq->txbufs[txq->wr_p].mbuf;
issued_descs++;
}
i++;
return i;
}
+static void
+nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
+{
+ uint32_t new_ctrl, update;
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ new_ctrl = 0;
+
+ if ((mask & ETH_VLAN_FILTER_OFFLOAD) ||
+ (mask & ETH_VLAN_EXTEND_OFFLOAD))
+ RTE_LOG(INFO, PMD, "No support for ETH_VLAN_FILTER_OFFLOAD or"
+ " ETH_VLAN_EXTEND_OFFLOAD");
+
+ /* Enable vlan strip if it is not configured yet */
+ if ((mask & ETH_VLAN_STRIP_OFFLOAD) &&
+ !(hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN))
+ new_ctrl = hw->ctrl | NFP_NET_CFG_CTRL_RXVLAN;
+
+ /* Disable vlan strip just if it is configured */
+ if (!(mask & ETH_VLAN_STRIP_OFFLOAD) &&
+ (hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN))
+ new_ctrl = hw->ctrl & ~NFP_NET_CFG_CTRL_RXVLAN;
+
+ if (new_ctrl == 0)
+ return;
+
+ update = NFP_NET_CFG_UPDATE_GEN;
+
+ if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
+ return;
+
+ hw->ctrl = new_ctrl;
+}
+
/* Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device */
static int
nfp_net_reta_update(struct rte_eth_dev *dev,
}
/* Initialise and register driver with DPDK Application */
-static struct eth_dev_ops nfp_net_eth_dev_ops = {
+static const struct eth_dev_ops nfp_net_eth_dev_ops = {
.dev_configure = nfp_net_configure,
.dev_start = nfp_net_start,
.dev_stop = nfp_net_stop,
.dev_close = nfp_net_close,
+ .promiscuous_enable = nfp_net_promisc_enable,
+ .promiscuous_disable = nfp_net_promisc_disable,
.link_update = nfp_net_link_update,
.stats_get = nfp_net_stats_get,
.stats_reset = nfp_net_stats_reset,
+ .dev_infos_get = nfp_net_infos_get,
+ .dev_supported_ptypes_get = nfp_net_supported_ptypes_get,
+ .mtu_set = nfp_net_dev_mtu_set,
+ .vlan_offload_set = nfp_net_vlan_offload_set,
.reta_update = nfp_net_reta_update,
.reta_query = nfp_net_reta_query,
.rss_hash_update = nfp_net_rss_hash_update,
.rx_queue_count = nfp_net_rx_queue_count,
.tx_queue_setup = nfp_net_tx_queue_setup,
.tx_queue_release = nfp_net_tx_queue_release,
+ .rx_queue_intr_enable = nfp_rx_queue_intr_enable,
+ .rx_queue_intr_disable = nfp_rx_queue_intr_disable,
};
static int
if (rte_eal_process_type() != RTE_PROC_PRIMARY)
return 0;
- pci_dev = eth_dev->pci_dev;
+ pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+ rte_eth_copy_pci_info(eth_dev, pci_dev);
+ eth_dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE;
+
hw->device_id = pci_dev->id.device_id;
hw->vendor_id = pci_dev->id.vendor_id;
hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
- PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u\n",
+ PMD_INIT_LOG(DEBUG, "nfp_net: device (%u:%u) %u:%u:%u:%u",
pci_dev->id.vendor_id, pci_dev->id.device_id,
pci_dev->addr.domain, pci_dev->addr.bus,
pci_dev->addr.devid, pci_dev->addr.function);
return -ENODEV;
}
- PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%08x\n", tx_bar_off);
- PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%08x\n", rx_bar_off);
+ PMD_INIT_LOG(DEBUG, "tx_bar_off: 0x%08x", tx_bar_off);
+ PMD_INIT_LOG(DEBUG, "rx_bar_off: 0x%08x", rx_bar_off);
hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr + tx_bar_off;
hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr + rx_bar_off;
- PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p\n",
+ PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p",
hw->ctrl_bar, hw->tx_bar, hw->rx_bar);
nfp_net_cfg_queue_setup(hw);
else
hw->rx_offset = nn_cfg_readl(hw, NFP_NET_CFG_RX_OFFSET_ADDR);
- PMD_INIT_LOG(INFO, "VER: %#x, Maximum supported MTU: %d\n",
+ 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\n", hw->cap,
+ PMD_INIT_LOG(INFO, "CAP: %#x, %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_RXCSUM ? "RXCSUM " : "",
hw->cap & NFP_NET_CFG_CTRL_TXCSUM ? "TXCSUM " : "",
hw->cap & NFP_NET_CFG_CTRL_LSO ? "TSO " : "",
hw->cap & NFP_NET_CFG_CTRL_RSS ? "RSS " : "");
- pci_dev = eth_dev->pci_dev;
hw->ctrl = 0;
hw->stride_rx = stride;
hw->stride_tx = stride;
- PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u\n",
+ PMD_INIT_LOG(INFO, "max_rx_queues: %u, max_tx_queues: %u",
hw->max_rx_queues, hw->max_tx_queues);
+ /* Initializing spinlock for reconfigs */
+ rte_spinlock_init(&hw->reconfig_lock);
+
/* Allocating memory for mac addr */
eth_dev->data->mac_addrs = rte_zmalloc("mac_addr", ETHER_ADDR_LEN, 0);
if (eth_dev->data->mac_addrs == NULL) {
return -ENOMEM;
}
- /* Using random mac addresses for VFs */
- eth_random_addr(&hw->mac_addr[0]);
+ nfp_net_read_mac(hw);
+
+ if (!is_valid_assigned_ether_addr((struct ether_addr *)&hw->mac_addr))
+ /* Using random mac addresses for VFs */
+ eth_random_addr(&hw->mac_addr[0]);
/* Copying mac address to DPDK eth_dev struct */
- ether_addr_copy(ð_dev->data->mac_addrs[0],
- (struct ether_addr *)hw->mac_addr);
+ ether_addr_copy((struct ether_addr *)hw->mac_addr,
+ ð_dev->data->mac_addrs[0]);
PMD_INIT_LOG(INFO, "port %d VendorID=0x%x DeviceID=0x%x "
"mac=%02x:%02x:%02x:%02x:%02x:%02x",
hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
+ /* Registering LSC interrupt handler */
+ rte_intr_callback_register(&pci_dev->intr_handle,
+ nfp_net_dev_interrupt_handler,
+ (void *)eth_dev);
+
+ /* Telling the firmware about the LSC interrupt entry */
+ nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
+
/* Recording current stats counters values */
nfp_net_stats_reset(eth_dev);
return 0;
}
-static struct rte_pci_id pci_id_nfp_net_map[] = {
+static const struct rte_pci_id pci_id_nfp_net_map[] = {
{
- .vendor_id = PCI_VENDOR_ID_NETRONOME,
- .device_id = PCI_DEVICE_ID_NFP6000_PF_NIC,
- .subsystem_vendor_id = PCI_ANY_ID,
- .subsystem_device_id = PCI_ANY_ID,
+ RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+ PCI_DEVICE_ID_NFP6000_PF_NIC)
},
{
- .vendor_id = PCI_VENDOR_ID_NETRONOME,
- .device_id = PCI_DEVICE_ID_NFP6000_VF_NIC,
- .subsystem_vendor_id = PCI_ANY_ID,
- .subsystem_device_id = PCI_ANY_ID,
+ RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+ PCI_DEVICE_ID_NFP6000_VF_NIC)
},
{
.vendor_id = 0,
},
};
-static struct eth_driver rte_nfp_net_pmd = {
- {
- .name = "rte_nfp_net_pmd",
- .id_table = pci_id_nfp_net_map,
- .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
- },
- .eth_dev_init = nfp_net_init,
- .dev_private_size = sizeof(struct nfp_net_adapter),
-};
-
-static int
-nfp_net_pmd_init(const char *name __rte_unused,
- const char *params __rte_unused)
+static int eth_nfp_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+ struct rte_pci_device *pci_dev)
{
- PMD_INIT_FUNC_TRACE();
- PMD_INIT_LOG(INFO, "librte_pmd_nfp_net version %s\n",
- NFP_NET_PMD_VERSION);
+ return rte_eth_dev_pci_generic_probe(pci_dev,
+ sizeof(struct nfp_net_adapter), nfp_net_init);
+}
- rte_eth_driver_register(&rte_nfp_net_pmd);
- return 0;
+static int eth_nfp_pci_remove(struct rte_pci_device *pci_dev)
+{
+ return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
}
-static struct rte_driver rte_nfp_net_driver = {
- .type = PMD_PDEV,
- .init = nfp_net_pmd_init,
+static struct rte_pci_driver rte_nfp_net_pmd = {
+ .id_table = pci_id_nfp_net_map,
+ .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+ .probe = eth_nfp_pci_probe,
+ .remove = eth_nfp_pci_remove,
};
-PMD_REGISTER_DRIVER(rte_nfp_net_driver);
+RTE_PMD_REGISTER_PCI(net_nfp, rte_nfp_net_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp, pci_id_nfp_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp, "* igb_uio | uio_pci_generic | vfio-pci");
/*
* Local variables:
git.droids-corp.org / dpdk.git / blobdiff