/*
- * 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.
* Netronome vNIC DPDK Poll-Mode Driver: Main entry point
*/
-#include <math.h>
-
#include <rte_byteorder.h>
#include <rte_common.h>
#include <rte_log.h>
#include <rte_debug.h>
-#include <rte_ethdev.h>
+#include <rte_ethdev_driver.h>
+#include <rte_ethdev_pci.h>
#include <rte_dev.h>
#include <rte_ether.h>
#include <rte_malloc.h>
#include <rte_alarm.h>
#include <rte_spinlock.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"
/* 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(struct rte_intr_handle *handle,
- void *param);
+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,
uint16_t nb_desc, unsigned int socket_id,
const struct rte_eth_txconf *tx_conf);
static int nfp_net_start(struct rte_eth_dev *dev);
-static void nfp_net_stats_get(struct rte_eth_dev *dev,
+static int nfp_net_stats_get(struct rte_eth_dev *dev,
struct rte_eth_stats *stats);
static void nfp_net_stats_reset(struct rte_eth_dev *dev);
static void nfp_net_stop(struct rte_eth_dev *dev);
static uint16_t nfp_net_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts);
-/*
- * 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) (0x80000 + (0x800 * ((_q) & 0xff)))
+static int nfp_net_rss_config_default(struct rte_eth_dev *dev);
+static int nfp_net_rss_hash_update(struct rte_eth_dev *dev,
+ struct rte_eth_rss_conf *rss_conf);
+static int nfp_net_rss_reta_write(struct rte_eth_dev *dev,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ uint16_t reta_size);
+static int nfp_net_rss_hash_write(struct rte_eth_dev *dev,
+ struct rte_eth_rss_conf *rss_conf);
+static int nfp_set_mac_addr(struct rte_eth_dev *dev,
+ struct ether_addr *mac_addr);
+
+/* The offset of the queue controller queues in the PCIe Target */
+#define NFP_PCIE_QUEUE(_q) (0x80000 + (NFP_QCP_QUEUE_ADDR_SZ * ((_q) & 0xff)))
/* Maximum value which can be added to a queue with one transaction */
#define NFP_QCP_MAX_ADD 0x7f
#define RTE_MBUF_DMA_ADDR_DEFAULT(mb) \
- (uint64_t)((mb)->buf_physaddr + RTE_PKTMBUF_HEADROOM)
+ (uint64_t)((mb)->buf_iova + RTE_PKTMBUF_HEADROOM)
/* nfp_qcp_ptr - Read or Write Pointer of a queue */
enum nfp_qcp_ptr {
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.
- *
- * @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)
{
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;
}
nfp_net_tx_queue_release_mbufs(txq);
txq->wr_p = 0;
txq->rd_p = 0;
- txq->tail = 0;
- txq->qcp_rd_p = 0;
}
static int
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)
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++) {
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, "Ack DONE\n");
+ PMD_DRV_LOG(DEBUG, "Ack DONE");
return 0;
}
{
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);
* 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;
}
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);
* 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;
}
/* 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\n");
- return -EINVAL;
- }
- }
-
- if (rxmode->split_hdr_size) {
- PMD_INIT_LOG(INFO, "rxmode does not support split header\n");
- 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\n");
- return -EINVAL;
- }
- }
-
- if (rxmode->hw_vlan_filter) {
- PMD_INIT_LOG(INFO, "VLAN filter not supported\n");
- 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\n");
- return -EINVAL;
- }
- }
-
- if (rxmode->hw_vlan_extend) {
- PMD_INIT_LOG(INFO, "VLAN extended not supported\n");
- return -EINVAL;
- }
-
- /* Supporting VLAN insertion by default */
- if (hw->cap & NFP_NET_CFG_CTRL_TXVLAN)
- new_ctrl |= NFP_NET_CFG_CTRL_TXVLAN;
-
- if (rxmode->jumbo_frame)
- /* this is handled in rte_eth_dev_configure */
-
- if (rxmode->hw_strip_crc) {
- PMD_INIT_LOG(INFO, "strip CRC not supported\n");
- return -EINVAL;
- }
-
- if (rxmode->enable_scatter) {
- PMD_INIT_LOG(INFO, "Scatter not supported\n");
+ 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 (!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;
+ /* Checking RX offloads */
+ if (!(rxmode->offloads & DEV_RX_OFFLOAD_CRC_STRIP))
+ PMD_INIT_LOG(INFO, "HW does strip CRC. No configurable!");
return 0;
}
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;
}
+#define ETH_ADDR_LEN 6
+
+static void
+nfp_eth_copy_mac(uint8_t *dst, const uint8_t *src)
+{
+ int i;
+
+ for (i = 0; i < ETH_ADDR_LEN; i++)
+ dst[i] = src[i];
+}
+
+static int
+nfp_net_pf_read_mac(struct nfp_net_hw *hw, int port)
+{
+ 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((uint8_t *)&hw->mac_addr,
+ (uint8_t *)&nfp_eth_table->ports[port].mac_addr);
+
+ free(nfp_eth_table);
+ return 0;
+}
+
+static void
+nfp_net_vf_read_mac(struct nfp_net_hw *hw)
+{
+ uint32_t tmp;
+
+ tmp = rte_be_to_cpu_32(nn_cfg_readl(hw, NFP_NET_CFG_MACADDR));
+ memcpy(&hw->mac_addr[0], &tmp, 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 void
+nfp_net_write_mac(struct nfp_net_hw *hw, uint8_t *mac)
+{
+ uint32_t mac0 = *(uint32_t *)mac;
+ uint16_t mac1;
+
+ nn_writel(rte_cpu_to_be_32(mac0), hw->ctrl_bar + NFP_NET_CFG_MACADDR);
+
+ mac += 4;
+ mac1 = *(uint16_t *)mac;
+ nn_writew(rte_cpu_to_be_16(mac1),
+ 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)
+{
+ 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);
+ intr_handle->intr_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);
+ intr_handle->intr_vec[i] = i + 1;
+ PMD_INIT_LOG(DEBUG, "intr_vec[%d]= %d", i,
+ intr_handle->intr_vec[i]);
+ }
+ }
+
+ /* Avoiding TX interrupts */
+ hw->ctrl |= NFP_NET_CFG_CTRL_MSIX_TX_OFF;
+ 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;
uint32_t new_ctrl, update = 0;
struct nfp_net_hw *hw;
+ struct rte_eth_conf *dev_conf;
+ struct rte_eth_rxmode *rxmode;
+ uint32_t intr_vector;
int ret;
hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
- PMD_INIT_LOG(DEBUG, "Start\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 (hw->pf_multiport_enabled) {
+ PMD_INIT_LOG(ERR, "PMD rx interrupt is not supported "
+ "with NFP multiport PF");
+ return -EINVAL;
+ }
+ if (intr_handle->type == RTE_INTR_HANDLE_UIO) {
+ /*
+ * Better not to share LSC with RX interrupts.
+ * Unregistering LSC interrupt handler
+ */
+ rte_intr_callback_unregister(&pci_dev->intr_handle,
+ nfp_net_dev_interrupt_handler, (void *)dev);
+
+ if (dev->data->nb_rx_queues > 1) {
+ PMD_INIT_LOG(ERR, "PMD rx interrupt only "
+ "supports 1 queue with UIO");
+ return -EIO;
+ }
+ }
+ intr_vector = dev->data->nb_rx_queues;
+ if (rte_intr_efd_enable(intr_handle, intr_vector))
+ return -1;
+
+ nfp_configure_rx_interrupt(dev, intr_handle);
+ update = NFP_NET_CFG_UPDATE_MSIX;
+ }
+
+ rte_intr_enable(intr_handle);
+
+ new_ctrl = nfp_check_offloads(dev);
+
/* Writing configuration parameters in the device */
nfp_net_params_setup(hw);
- /* Enabling the required queues in the device */
- nfp_net_enable_queues(dev);
+ dev_conf = &dev->data->dev_conf;
+ rxmode = &dev_conf->rxmode;
+
+ if (rxmode->mq_mode & ETH_MQ_RX_RSS) {
+ nfp_net_rss_config_default(dev);
+ update |= NFP_NET_CFG_UPDATE_RSS;
+ new_ctrl |= NFP_NET_CFG_CTRL_RSS;
+ }
/* Enable device */
- new_ctrl = hw->ctrl | NFP_NET_CFG_CTRL_ENABLE | NFP_NET_CFG_UPDATE_MSIX;
- update = NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
+ new_ctrl |= NFP_NET_CFG_CTRL_ENABLE;
+
+ update |= NFP_NET_CFG_UPDATE_GEN | NFP_NET_CFG_UPDATE_RING;
if (hw->cap & NFP_NET_CFG_CTRL_RINGCFG)
new_ctrl |= NFP_NET_CFG_CTRL_RINGCFG;
goto error;
}
+ if (hw->is_pf)
+ /* Configure the physical port up */
+ nfp_eth_set_configured(hw->cpp, hw->pf_port_idx, 1);
+
hw->ctrl = new_ctrl;
return 0;
nfp_net_stop(struct rte_eth_dev *dev)
{
int i;
+ struct nfp_net_hw *hw;
+
+ PMD_INIT_LOG(DEBUG, "Stop");
- PMD_INIT_LOG(DEBUG, "Stop\n");
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
nfp_net_disable_queues(dev);
nfp_net_reset_rx_queue(
(struct nfp_net_rxq *)dev->data->rx_queues[i]);
}
+
+ if (hw->is_pf)
+ /* Configure the physical port down */
+ nfp_eth_set_configured(hw->cpp, hw->pf_port_idx, 0);
}
/* Reset and stop device. The device can not be restarted. */
nfp_net_close(struct rte_eth_dev *dev)
{
struct nfp_net_hw *hw;
+ struct rte_pci_device *pci_dev;
+ int i;
- 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
* threads/queues before calling the device close function.
*/
- nfp_net_stop(dev);
+ nfp_net_disable_queues(dev);
+
+ /* Clear queues */
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ nfp_net_reset_tx_queue(
+ (struct nfp_net_txq *)dev->data->tx_queues[i]);
+ }
+
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ nfp_net_reset_rx_queue(
+ (struct nfp_net_rxq *)dev->data->rx_queues[i]);
+ }
- rte_intr_disable(&dev->pci_dev->intr_handle);
+ 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...
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);
if (!(hw->cap & NFP_NET_CFG_CTRL_PROMISC)) {
- PMD_INIT_LOG(INFO, "Promiscuous mode not supported\n");
+ 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");
+ PMD_DRV_LOG(INFO, "Promiscuous mode already enabled");
return;
}
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;
}
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,
+ [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");
+ 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));
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_SPEED_NUM_40G;
- if (old.link_status != link.link_status) {
- nfp_net_dev_atomic_write_link_status(dev, &link);
+ nn_link_status = (nn_link_status >> NFP_NET_CFG_STS_LINK_RATE_SHIFT) &
+ NFP_NET_CFG_STS_LINK_RATE_MASK;
+
+ 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];
+
+ 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 void
+static int
nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
{
int i;
/* RTE_ETHDEV_QUEUE_STAT_CNTRS default value is 16 */
+ memset(&nfp_dev_stats, 0, sizeof(nfp_dev_stats));
+
/* reading per RX ring stats */
for (i = 0; i < dev->data->nb_rx_queues; i++) {
if (i == RTE_ETHDEV_QUEUE_STAT_CNTRS)
nfp_dev_stats.imissed -= hw->eth_stats_base.imissed;
- if (stats)
+ if (stats) {
memcpy(stats, &nfp_dev_stats, sizeof(*stats));
+ return 0;
+ }
+ return -EINVAL;
}
static void
hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
- dev_info->driver_name = dev->driver->pci_drv.name;
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;
+ dev_info->max_rx_pktlen = hw->max_mtu;
/* Next should change when PF support is implemented */
dev_info->max_mac_addrs = 1;
DEV_RX_OFFLOAD_UDP_CKSUM |
DEV_RX_OFFLOAD_TCP_CKSUM;
+ dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_JUMBO_FRAME;
+
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_RX_OFFLOAD_UDP_CKSUM |
- DEV_RX_OFFLOAD_TCP_CKSUM;
+ 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 = {
},
.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 |
+ ETH_RSS_NONFRAG_IPV4_UDP |
+ ETH_RSS_NONFRAG_IPV6_TCP |
+ ETH_RSS_NONFRAG_IPV6_UDP;
+
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_LINK_SPEED_40G | ETH_LINK_SPEED_100G;
+ 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 *
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);
+ rte_eth_linkstatus_get(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");
+ 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",
- (int)(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",
- dev->pci_dev->addr.domain, dev->pci_dev->addr.bus,
- dev->pci_dev->addr.devid, dev->pci_dev->addr.function);
+ 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);
}
/* Interrupt configuration and handling */
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(&dev->pci_dev->intr_handle);
+ rte_intr_enable(&pci_dev->intr_handle);
} else {
/* Make sure all updates are written before un-masking */
rte_wmb();
}
static void
-nfp_net_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
- void *param)
+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");
+ 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);
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);
}
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);
+ _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
nfp_net_dev_link_status_print(dev);
if ((mtu < ETHER_MIN_MTU) || ((uint32_t)mtu > hw->max_mtu))
return -EINVAL;
- /* switch to jumbo mode if needed */
+ /* mtu setting is forbidden if port is started */
+ if (dev->data->dev_started) {
+ PMD_DRV_LOG(ERR, "port %d must be stopped before configuration",
+ dev->data->port_id);
+ return -EBUSY;
+ }
+
+ /* switch to jumbo mode if needed */
if ((uint32_t)mtu > ETHER_MAX_LEN)
dev->data->dev_conf.rxmode.jumbo_frame = 1;
else
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;
}
* 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");
+ PMD_DRV_LOG(ERR, "Error allocatig rx dma");
nfp_net_rx_queue_release(rxq);
return -ENOMEM;
}
/* Saving physical and virtual addresses for the RX ring */
- rxq->dma = (uint64_t)tz->phys_addr;
+ rxq->dma = (uint64_t)tz->iova;
rxq->rxds = (struct nfp_net_rx_desc *)tz->addr;
/* mbuf pointers array for referencing mbufs linked to RX descriptors */
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);
* of descriptors in log2 format
*/
nn_cfg_writeq(hw, NFP_NET_CFG_RXR_ADDR(queue_idx), rxq->dma);
- nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), log2(nb_desc));
+ nn_cfg_writeb(hw, NFP_NET_CFG_RXR_SZ(queue_idx), rte_log2_u32(nb_desc));
return 0;
}
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++) {
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;
}
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);
-
- rxq->wr_p++;
+ 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);
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;
}
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,
- (int)dev->data->port_id, (int)queue_idx);
+ "queue=%d)", (unsigned int)tx_free_thresh,
+ dev->data->port_id, (int)queue_idx);
return -(EINVAL);
}
* 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;
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;
}
* 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");
+ PMD_DRV_LOG(ERR, "Error allocating tx dma");
nfp_net_tx_queue_release(txq);
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;
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->phys_addr;
+ txq->dma = (uint64_t)tz->iova;
txq->txds = (struct nfp_net_tx_desc *)tz->addr;
/* mbuf pointers array for referencing mbufs linked to TX descriptors */
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);
* of descriptors in log2 format
*/
nn_cfg_writeq(hw, NFP_NET_CFG_TXR_ADDR(queue_idx), txq->dma);
- nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), log2(nb_desc));
+ nn_cfg_writeb(hw, NFP_NET_CFG_TXR_SZ(queue_idx), rte_log2_u32(nb_desc));
return 0;
}
+/* nfp_net_tx_tso - Set TX descriptor for TSO */
+static inline void
+nfp_net_tx_tso(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
+ struct rte_mbuf *mb)
+{
+ uint64_t ol_flags;
+ struct nfp_net_hw *hw = txq->hw;
+
+ if (!(hw->cap & NFP_NET_CFG_CTRL_LSO_ANY))
+ goto clean_txd;
+
+ ol_flags = mb->ol_flags;
+
+ if (!(ol_flags & PKT_TX_TCP_SEG))
+ goto clean_txd;
+
+ txd->l3_offset = mb->l2_len;
+ txd->l4_offset = mb->l2_len + mb->l3_len;
+ txd->lso_hdrlen = mb->l2_len + mb->l3_len + mb->l4_len;
+ txd->mss = rte_cpu_to_le_16(mb->tso_segsz);
+ txd->flags = PCIE_DESC_TX_LSO;
+ return;
+
+clean_txd:
+ txd->flags = 0;
+ txd->l3_offset = 0;
+ txd->l4_offset = 0;
+ txd->lso_hdrlen = 0;
+ txd->mss = 0;
+}
+
/* nfp_net_tx_cksum - Set TX CSUM offload flags in TX descriptor */
static inline void
nfp_net_tx_cksum(struct nfp_net_txq *txq, struct nfp_net_tx_desc *txd,
#define NFP_HASH_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 4)
#define NFP_HASH_TYPE_OFFSET ((uint8_t *)mbuf->buf_addr + mbuf->data_off - 8)
+#define NFP_DESC_META_LEN(d) (d->rxd.meta_len_dd & PCIE_DESC_RX_META_LEN_MASK)
+
/*
* nfp_net_set_hash - Set mbuf hash data
*
nfp_net_set_hash(struct nfp_net_rxq *rxq, struct nfp_net_rx_desc *rxd,
struct rte_mbuf *mbuf)
{
- uint32_t hash;
- uint32_t hash_type;
struct nfp_net_hw *hw = rxq->hw;
+ uint8_t *meta_offset;
+ uint32_t meta_info;
+ uint32_t hash = 0;
+ uint32_t hash_type = 0;
if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
return;
- if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
- return;
+ /* this is true for new firmwares */
+ if (likely(((hw->cap & NFP_NET_CFG_CTRL_RSS2) ||
+ (NFD_CFG_MAJOR_VERSION_of(hw->ver) == 4)) &&
+ NFP_DESC_META_LEN(rxd))) {
+ /*
+ * new metadata api:
+ * <---- 32 bit ----->
+ * m field type word
+ * e data field #2
+ * t data field #1
+ * a data field #0
+ * ====================
+ * packet data
+ *
+ * Field type word contains up to 8 4bit field types
+ * A 4bit field type refers to a data field word
+ * A data field word can have several 4bit field types
+ */
+ meta_offset = rte_pktmbuf_mtod(mbuf, uint8_t *);
+ meta_offset -= NFP_DESC_META_LEN(rxd);
+ meta_info = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+ meta_offset += 4;
+ /* NFP PMD just supports metadata for hashing */
+ switch (meta_info & NFP_NET_META_FIELD_MASK) {
+ case NFP_NET_META_HASH:
+ /* next field type is about the hash type */
+ meta_info >>= NFP_NET_META_FIELD_SIZE;
+ /* hash value is in the data field */
+ hash = rte_be_to_cpu_32(*(uint32_t *)meta_offset);
+ hash_type = meta_info & NFP_NET_META_FIELD_MASK;
+ break;
+ default:
+ /* Unsupported metadata can be a performance issue */
+ return;
+ }
+ } else {
+ if (!(rxd->rxd.flags & PCIE_DESC_RX_RSS))
+ return;
- hash = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_OFFSET);
- hash_type = rte_be_to_cpu_32(*(uint32_t *)NFP_HASH_TYPE_OFFSET);
+ 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;
}
+ 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[idx];
- 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
*/
new_mb = rte_pktmbuf_alloc(rxq->mem_pool);
if (unlikely(new_mb == NULL)) {
- RTE_LOG(DEBUG, PMD, "RX mbuf alloc failed port_id=%u "
- "queue_id=%u\n", (unsigned)rxq->port_id,
- (unsigned)rxq->qidx);
+ RTE_LOG_DP(DEBUG, PMD,
+ "RX mbuf alloc failed port_id=%u queue_id=%u\n",
+ rxq->port_id, (unsigned int)rxq->qidx);
nfp_net_mbuf_alloc_failed(rxq);
break;
}
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 */
* 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"
mb->nb_segs = 1;
mb->next = NULL;
+ mb->port = rxq->port_id;
+
/* Checking the RSS flag */
nfp_net_set_hash(rxq, rxds, mb);
/* Checking the checksum flag */
nfp_net_rx_cksum(rxq, rxds, mb);
- /* 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_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)
return nb_hold;
- PMD_RX_LOG(DEBUG, "RX port_id=%u queue_id=%u, %d packets received\n",
- (unsigned)rxq->port_id, (unsigned)rxq->qidx, nb_hold);
+ PMD_RX_LOG(DEBUG, "RX port_id=%u queue_id=%u, %d packets received",
+ rxq->port_id, (unsigned int)rxq->qidx, nb_hold);
nb_hold += rxq->nb_rx_hold;
*/
rte_wmb();
if (nb_hold > rxq->rx_free_thresh) {
- PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u\n",
- (unsigned)rxq->port_id, (unsigned)rxq->qidx,
+ PMD_RX_LOG(DEBUG, "port=%u queue=%u nb_hold=%u avail=%u",
+ rxq->port_id, (unsigned int)rxq->qidx,
(unsigned)nb_hold, (unsigned)avail);
nfp_qcp_ptr_add(rxq->qcp_fl, NFP_QCP_WRITE_PTR, nb_hold);
nb_hold = 0;
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->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);
+ "packets (%u, %u)", txq->qidx,
+ 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",
+ 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);
+ 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)))
+ 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;
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) {
/* 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 */
+ while (pkt) {
+ /* Copying TSO, VLAN and cksum info */
+ *txds = txd;
+
+ /* Releasing mbuf used by this descriptor previously*/
if (*lmbuf)
rte_pktmbuf_free_seg(*lmbuf);
+ /*
+ * Linking mbuf with descriptor for being released
+ * next time descriptor is used
+ */
+ *lmbuf = pkt;
+
dma_size = pkt->data_len;
- dma_addr = rte_mbuf_data_dma_addr(pkt);
+ 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;
- 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) {
+ if (!pkt_size)
/* End of packet */
txds->offset_eop |= PCIE_DESC_TX_EOP;
- } else {
+ else
txds->offset_eop &= PCIE_DESC_TX_OFFSET_MASK;
- pkt = pkt->next;
- }
+
+ 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
+static int
nfp_net_vlan_offload_set(struct rte_eth_dev *dev, int mask)
{
uint32_t new_ctrl, update;
struct nfp_net_hw *hw;
+ int ret;
hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
new_ctrl = 0;
if ((mask & ETH_VLAN_FILTER_OFFLOAD) ||
- (mask & ETH_VLAN_FILTER_OFFLOAD))
- RTE_LOG(INFO, PMD, "Not support for ETH_VLAN_FILTER_OFFLOAD or"
- " ETH_VLAN_FILTER_EXTEND");
+ (mask & ETH_VLAN_EXTEND_OFFLOAD))
+ 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 */
if ((mask & ETH_VLAN_STRIP_OFFLOAD) &&
new_ctrl = hw->ctrl & ~NFP_NET_CFG_CTRL_RXVLAN;
if (new_ctrl == 0)
- return;
+ return 0;
update = NFP_NET_CFG_UPDATE_GEN;
- if (nfp_net_reconfig(hw, new_ctrl, update) < 0)
- return;
+ ret = nfp_net_reconfig(hw, new_ctrl, update);
+ if (!ret)
+ hw->ctrl = new_ctrl;
- hw->ctrl = new_ctrl;
+ return ret;
}
-/* Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device */
static int
-nfp_net_reta_update(struct rte_eth_dev *dev,
+nfp_net_rss_reta_write(struct rte_eth_dev *dev,
struct rte_eth_rss_reta_entry64 *reta_conf,
uint16_t reta_size)
{
uint32_t reta, mask;
int i, j;
int idx, shift;
- uint32_t update;
struct nfp_net_hw *hw =
NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
- if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
- 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;
}
reta &= ~(0xFF << (8 * j));
reta |= reta_conf[idx].reta[shift + j] << (8 * j);
}
- nn_cfg_writel(hw, NFP_NET_CFG_RSS_ITBL + shift, reta);
+ nn_cfg_writel(hw, NFP_NET_CFG_RSS_ITBL + (idx * 64) + shift,
+ reta);
}
+ return 0;
+}
+
+/* Update Redirection Table(RETA) of Receive Side Scaling of Ethernet device */
+static int
+nfp_net_reta_update(struct rte_eth_dev *dev,
+ struct rte_eth_rss_reta_entry64 *reta_conf,
+ uint16_t reta_size)
+{
+ struct nfp_net_hw *hw =
+ NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ uint32_t update;
+ int ret;
+
+ if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS))
+ return -EINVAL;
+
+ ret = nfp_net_rss_reta_write(dev, reta_conf, reta_size);
+ if (ret != 0)
+ return ret;
update = NFP_NET_CFG_UPDATE_RSS;
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;
}
if (!mask)
continue;
- reta = nn_cfg_readl(hw, NFP_NET_CFG_RSS_ITBL + shift);
+ reta = nn_cfg_readl(hw, NFP_NET_CFG_RSS_ITBL + (idx * 64) +
+ shift);
for (j = 0; j < 4; j++) {
if (!(mask & (0x1 << j)))
continue;
}
static int
-nfp_net_rss_hash_update(struct rte_eth_dev *dev,
+nfp_net_rss_hash_write(struct rte_eth_dev *dev,
struct rte_eth_rss_conf *rss_conf)
{
- uint32_t update;
+ struct nfp_net_hw *hw;
+ uint64_t rss_hf;
uint32_t cfg_rss_ctrl = 0;
uint8_t key;
- uint64_t rss_hf;
int i;
- struct nfp_net_hw *hw;
hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
- rss_hf = rss_conf->rss_hf;
-
- /* 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");
- return -EINVAL;
- }
- return 0; /* Nothing to do */
+ /* Writing the key byte a byte */
+ for (i = 0; i < rss_conf->rss_key_len; i++) {
+ memcpy(&key, &rss_conf->rss_key[i], 1);
+ nn_cfg_writeb(hw, NFP_NET_CFG_RSS_KEY + i, key);
}
- if (rss_conf->rss_key_len > NFP_NET_CFG_RSS_KEY_SZ) {
- RTE_LOG(ERR, PMD, "hash key too long\n");
- return -EINVAL;
- }
+ rss_hf = rss_conf->rss_hf;
if (rss_hf & ETH_RSS_IPV4)
cfg_rss_ctrl |= NFP_NET_CFG_RSS_IPV4 |
NFP_NET_CFG_RSS_IPV6_TCP |
NFP_NET_CFG_RSS_IPV6_UDP;
+ cfg_rss_ctrl |= NFP_NET_CFG_RSS_MASK;
+ cfg_rss_ctrl |= NFP_NET_CFG_RSS_TOEPLITZ;
+
/* configuring where to apply the RSS hash */
nn_cfg_writel(hw, NFP_NET_CFG_RSS_CTRL, cfg_rss_ctrl);
- /* Writing the key byte a byte */
- for (i = 0; i < rss_conf->rss_key_len; i++) {
- memcpy(&key, &rss_conf->rss_key[i], 1);
- nn_cfg_writeb(hw, NFP_NET_CFG_RSS_KEY + i, key);
- }
-
/* Writing the key size */
nn_cfg_writeb(hw, NFP_NET_CFG_RSS_KEY_SZ, rss_conf->rss_key_len);
+ return 0;
+}
+
+static int
+nfp_net_rss_hash_update(struct rte_eth_dev *dev,
+ struct rte_eth_rss_conf *rss_conf)
+{
+ uint32_t update;
+ uint64_t rss_hf;
+ struct nfp_net_hw *hw;
+
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ rss_hf = rss_conf->rss_hf;
+
+ /* Checking if RSS is enabled */
+ if (!(hw->ctrl & NFP_NET_CFG_CTRL_RSS)) {
+ if (rss_hf != 0) { /* Enable RSS? */
+ 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) {
+ PMD_DRV_LOG(ERR, "hash key too long");
+ return -EINVAL;
+ }
+
+ nfp_net_rss_hash_write(dev, rss_conf);
+
update = NFP_NET_CFG_UPDATE_RSS;
if (nfp_net_reconfig(hw, hw->ctrl, update) < 0)
return 0;
}
+static int
+nfp_net_rss_config_default(struct rte_eth_dev *dev)
+{
+ struct rte_eth_conf *dev_conf;
+ struct rte_eth_rss_conf rss_conf;
+ struct rte_eth_rss_reta_entry64 nfp_reta_conf[2];
+ uint16_t rx_queues = dev->data->nb_rx_queues;
+ uint16_t queue;
+ int i, j, ret;
+
+ PMD_DRV_LOG(INFO, "setting default RSS conf for %u queues",
+ rx_queues);
+
+ nfp_reta_conf[0].mask = ~0x0;
+ nfp_reta_conf[1].mask = ~0x0;
+
+ queue = 0;
+ for (i = 0; i < 0x40; i += 8) {
+ for (j = i; j < (i + 8); j++) {
+ nfp_reta_conf[0].reta[j] = queue;
+ nfp_reta_conf[1].reta[j] = queue++;
+ queue %= rx_queues;
+ }
+ }
+ ret = nfp_net_rss_reta_write(dev, nfp_reta_conf, 0x80);
+ if (ret != 0)
+ return ret;
+
+ dev_conf = &dev->data->dev_conf;
+ if (!dev_conf) {
+ PMD_DRV_LOG(INFO, "wrong rss conf");
+ return -EINVAL;
+ }
+ rss_conf = dev_conf->rx_adv_conf.rss_conf;
+
+ ret = nfp_net_rss_hash_write(dev, &rss_conf);
+
+ return ret;
+}
+
+
/* Initialise and register driver with DPDK Application */
static const struct eth_dev_ops nfp_net_eth_dev_ops = {
.dev_configure = nfp_net_configure,
.dev_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,
.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,
};
+/*
+ * All eth_dev created got its private data, but before nfp_net_init, that
+ * private data is referencing private data for all the PF ports. This is due
+ * to how the vNIC bars are mapped based on first port, so all ports need info
+ * about port 0 private data. Inside nfp_net_init the private data pointer is
+ * changed to the right address for each port once the bars have been mapped.
+ *
+ * This functions helps to find out which port and therefore which offset
+ * inside the private data array to use.
+ */
+static int
+get_pf_port_number(char *name)
+{
+ char *pf_str = name;
+ int size = 0;
+
+ while ((*pf_str != '_') && (*pf_str != '\0') && (size++ < 30))
+ pf_str++;
+
+ if (size == 30)
+ /*
+ * This should not happen at all and it would mean major
+ * implementation fault.
+ */
+ rte_panic("nfp_net: problem with pf device name\n");
+
+ /* Expecting _portX with X within [0,7] */
+ pf_str += 5;
+
+ return (int)strtol(pf_str, NULL, 10);
+}
+
static int
nfp_net_init(struct rte_eth_dev *eth_dev)
{
struct rte_pci_device *pci_dev;
- struct nfp_net_hw *hw;
+ struct nfp_net_hw *hw, *hwport0;
- uint32_t tx_bar_off, rx_bar_off;
+ uint64_t tx_bar_off = 0, rx_bar_off = 0;
uint32_t start_q;
int stride = 4;
+ int port = 0;
+ int err;
PMD_INIT_FUNC_TRACE();
- hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+ pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
+
+ if ((pci_dev->id.device_id == PCI_DEVICE_ID_NFP4000_PF_NIC) ||
+ (pci_dev->id.device_id == PCI_DEVICE_ID_NFP6000_PF_NIC)) {
+ port = get_pf_port_number(eth_dev->data->name);
+ if (port < 0 || port > 7) {
+ PMD_DRV_LOG(ERR, "Port value is wrong");
+ return -ENODEV;
+ }
+
+ 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];
+ } else {
+ hw = NFP_NET_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
+ hwport0 = 0;
+ }
eth_dev->dev_ops = &nfp_net_eth_dev_ops;
eth_dev->rx_pkt_burst = &nfp_net_recv_pkts;
if (rte_eal_process_type() != RTE_PROC_PRIMARY)
return 0;
- pci_dev = eth_dev->pci_dev;
rte_eth_copy_pci_info(eth_dev, pci_dev);
hw->device_id = pci_dev->id.device_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);
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) {
+ 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;
+ }
+
+ PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+ }
+
+ if (port > 0) {
+ if (!hwport0->ctrl_bar)
+ return -ENODEV;
+
+ /* address based on port0 offset */
+ hw->ctrl_bar = hwport0->ctrl_bar +
+ (port * NFP_PF_CSR_SLICE_SIZE);
+ }
+
+ PMD_INIT_LOG(DEBUG, "ctrl bar: %p", hw->ctrl_bar);
+
hw->max_rx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_RXRINGS);
hw->max_tx_queues = nn_cfg_readl(hw, NFP_NET_CFG_MAX_TXRINGS);
/* Work out where in the BAR the queues start. */
switch (pci_dev->id.device_id) {
+ case PCI_DEVICE_ID_NFP4000_PF_NIC:
+ case PCI_DEVICE_ID_NFP6000_PF_NIC:
case PCI_DEVICE_ID_NFP6000_VF_NIC:
start_q = nn_cfg_readl(hw, NFP_NET_CFG_START_TXQ);
- tx_bar_off = 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%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%" 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 */
+ 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;
+ }
+
+ PMD_INIT_LOG(DEBUG, "tx/rx bar address: 0x%p",
+ hwport0->hw_queues);
+ }
- 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;
+ if (hw->is_pf) {
+ hw->tx_bar = hwport0->hw_queues + tx_bar_off;
+ hw->rx_bar = hwport0->hw_queues + rx_bar_off;
+ eth_dev->data->dev_private = hw;
+ } else {
+ hw->tx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+ tx_bar_off;
+ hw->rx_bar = (uint8_t *)pci_dev->mem_resource[2].addr +
+ rx_bar_off;
+ }
- PMD_INIT_LOG(DEBUG, "ctrl_bar: %p, tx_bar: %p, rx_bar: %p\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);
hw->ver = nn_cfg_readl(hw, NFP_NET_CFG_VERSION);
hw->cap = nn_cfg_readl(hw, NFP_NET_CFG_CAP);
hw->max_mtu = nn_cfg_readl(hw, NFP_NET_CFG_MAX_MTU);
- hw->mtu = hw->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\n",
- 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, "VER: %u.%u, Maximum supported MTU: %d",
+ NFD_CFG_MAJOR_VERSION_of(hw->ver),
+ NFD_CFG_MINOR_VERSION_of(hw->ver), hw->max_mtu);
+
+ PMD_INIT_LOG(INFO, "CAP: %#x, %s%s%s%s%s%s%s%s%s%s%s%s%s%s", hw->cap,
hw->cap & NFP_NET_CFG_CTRL_PROMISC ? "PROMISC " : "",
+ hw->cap & NFP_NET_CFG_CTRL_L2BC ? "L2BCFILT " : "",
+ hw->cap & NFP_NET_CFG_CTRL_L2MC ? "L2MCFILT " : "",
hw->cap & NFP_NET_CFG_CTRL_RXCSUM ? "RXCSUM " : "",
hw->cap & NFP_NET_CFG_CTRL_TXCSUM ? "TXCSUM " : "",
hw->cap & NFP_NET_CFG_CTRL_RXVLAN ? "RXVLAN " : "",
hw->cap & NFP_NET_CFG_CTRL_TXVLAN ? "TXVLAN " : "",
hw->cap & NFP_NET_CFG_CTRL_SCATTER ? "SCATTER " : "",
hw->cap & NFP_NET_CFG_CTRL_GATHER ? "GATHER " : "",
+ hw->cap & NFP_NET_CFG_CTRL_LIVE_ADDR ? "LIVE_ADDR " : "",
hw->cap & NFP_NET_CFG_CTRL_LSO ? "TSO " : "",
- hw->cap & NFP_NET_CFG_CTRL_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 " : "");
- 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 */
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;
}
- /* Using random mac addresses for VFs */
- eth_random_addr(&hw->mac_addr[0]);
+ if (hw->is_pf) {
+ nfp_net_pf_read_mac(hwport0, port);
+ nfp_net_write_mac(hw, (uint8_t *)&hw->mac_addr);
+ } else {
+ nfp_net_vf_read_mac(hw);
+ }
+
+ 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);
+ }
/* 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",
nfp_net_dev_interrupt_handler,
(void *)eth_dev);
- /* enable uio intr after callback register */
- rte_intr_enable(&pci_dev->intr_handle);
-
/* Telling the firmware about the LSC interrupt entry */
nn_cfg_writeb(hw, NFP_NET_CFG_LSC, NFP_NET_IRQ_LSC_IDX);
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,
+ 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;
+ char *port_name;
+ int ret;
+
+ port_name = rte_zmalloc("nfp_pf_port_name", 100, 0);
+ if (!port_name)
+ return -ENOMEM;
+
+ if (ports > 1)
+ sprintf(port_name, "%s_port%d", dev->device.name, port);
+ else
+ sprintf(port_name, "%s", dev->device.name);
+
+ eth_dev = rte_eth_dev_allocate(port_name);
+ if (!eth_dev)
+ return -ENOMEM;
+
+ if (port == 0) {
+ *priv = rte_zmalloc(port_name,
+ sizeof(struct nfp_net_adapter) * ports,
+ RTE_CACHE_LINE_SIZE);
+ if (!*priv) {
+ rte_eth_dev_release_port(eth_dev);
+ return -ENOMEM;
+ }
+ }
+
+ eth_dev->data->dev_private = *priv;
+
+ /*
+ * dev_private pointing to port0 dev_private because we need
+ * to configure vNIC bars based on port0 at nfp_net_init.
+ * Then dev_private is adjusted per port.
+ */
+ hw = (struct nfp_net_hw *)(eth_dev->data->dev_private) + port;
+ 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);
+
+ ret = nfp_net_init(eth_dev);
+
+ if (ret)
+ rte_eth_dev_release_port(eth_dev);
+
+ rte_free(port_name);
+
+ return ret;
}
-static struct rte_pci_id pci_id_nfp_net_map[] = {
+#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[100];
+ char serial[100];
+ 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)
+{
+ 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;
+
+ cpp = nfp_cpp_from_device_name(dev->device.name);
+ if (!cpp) {
+ PMD_DRV_LOG(ERR, "A CPP handle can not be obtained");
+ ret = -EIO;
+ goto error;
+ }
+
+ 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;
+ }
+
+ if (nfp_fw_setup(dev, cpp, nfp_eth_table, hwinfo)) {
+ PMD_DRV_LOG(INFO, "Error when uploading firmware");
+ ret = -EIO;
+ goto error;
+ }
+
+ /* 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;
+ }
+
+ 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;
+ }
+ PMD_INIT_LOG(INFO, "Total pf ports: %d", total_ports);
+
+ if (total_ports <= 0 || total_ports > 8) {
+ 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, cpp, hwinfo,
+ nfp_eth_table->ports[i].index,
+ sym_tbl, &priv);
+ if (ret)
+ break;
+ }
+
+error:
+ free(nfp_eth_table);
+ return ret;
+}
+
+int nfp_logtype_init;
+int nfp_logtype_driver;
+
+static const struct rte_pci_id pci_id_nfp_pf_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_NFP4000_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_PF_NIC)
},
{
.vendor_id = 0,
},
};
-static struct eth_driver rte_nfp_net_pmd = {
+static const struct rte_pci_id pci_id_nfp_vf_net_map[] = {
+ {
+ RTE_PCI_DEVICE(PCI_VENDOR_ID_NETRONOME,
+ PCI_DEVICE_ID_NFP6000_VF_NIC)
+ },
{
- .name = "rte_nfp_net_pmd",
- .id_table = pci_id_nfp_net_map,
- .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+ .vendor_id = 0,
},
- .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)
+{
+ struct rte_eth_dev *eth_dev;
+ struct nfp_net_hw *hw, *hwport0;
+ int port = 0;
+
+ eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
+ 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);
+ }
+ /* hotplug is not possible with multiport PF */
+ if (hw->pf_multiport_enabled)
+ return -ENOTSUP;
+ 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_pf_pmd = {
+ .id_table = pci_id_nfp_pf_net_map,
+ .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+ .probe = nfp_pf_pci_probe,
+ .remove = eth_nfp_pci_remove,
};
-PMD_REGISTER_DRIVER(rte_nfp_net_driver);
+static struct rte_pci_driver rte_nfp_net_vf_pmd = {
+ .id_table = pci_id_nfp_vf_net_map,
+ .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+ .probe = eth_nfp_pci_probe,
+ .remove = eth_nfp_pci_remove,
+};
+
+RTE_PMD_REGISTER_PCI(net_nfp_pf, rte_nfp_net_pf_pmd);
+RTE_PMD_REGISTER_PCI(net_nfp_vf, rte_nfp_net_vf_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_pf, pci_id_nfp_pf_net_map);
+RTE_PMD_REGISTER_PCI_TABLE(net_nfp_vf, pci_id_nfp_vf_net_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_pf, "* igb_uio | uio_pci_generic | vfio");
+RTE_PMD_REGISTER_KMOD_DEP(net_nfp_vf, "* igb_uio | uio_pci_generic | vfio");
+RTE_INIT(nfp_init_log);
+static void
+nfp_init_log(void)
+{
+ nfp_logtype_init = rte_log_register("pmd.net.nfp.init");
+ if (nfp_logtype_init >= 0)
+ rte_log_set_level(nfp_logtype_init, RTE_LOG_NOTICE);
+ nfp_logtype_driver = rte_log_register("pmd.net.nfp.driver");
+ if (nfp_logtype_driver >= 0)
+ rte_log_set_level(nfp_logtype_driver, RTE_LOG_NOTICE);
+}
/*
* Local variables:
* c-file-style: "Linux"