#include <netinet/in.h>
#include <linux/ethtool.h>
#include <linux/sockios.h>
-#include <fcntl.h>
#include <rte_ether.h>
#include <rte_ethdev.h>
#include <rte_mempool.h>
#include <rte_prefetch.h>
#include <rte_malloc.h>
-#include <rte_spinlock.h>
-#include <rte_log.h>
-#include <rte_alarm.h>
#include <rte_memory.h>
#include <rte_flow.h>
#include <rte_kvargs.h>
#include <rte_interrupts.h>
#include <rte_branch_prediction.h>
+#include <rte_common.h>
/* Generated configuration header. */
#include "mlx4_autoconf.h"
/* PMD headers. */
#include "mlx4.h"
#include "mlx4_flow.h"
-
-/* Convenience macros for accessing mbuf fields. */
-#define NEXT(m) ((m)->next)
-#define DATA_LEN(m) ((m)->data_len)
-#define PKT_LEN(m) ((m)->pkt_len)
-#define DATA_OFF(m) ((m)->data_off)
-#define SET_DATA_OFF(m, o) ((m)->data_off = (o))
-#define NB_SEGS(m) ((m)->nb_segs)
-#define PORT(m) ((m)->port)
+#include "mlx4_rxtx.h"
+#include "mlx4_utils.h"
/** Configuration structure for device arguments. */
struct mlx4_conf {
NULL,
};
-static int
-mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx);
-
-static int
-mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx);
-
-static int
-priv_rx_intr_vec_enable(struct priv *priv);
-
-static void
-priv_rx_intr_vec_disable(struct priv *priv);
-
-/**
- * Lock private structure to protect it from concurrent access in the
- * control path.
- *
- * @param priv
- * Pointer to private structure.
- */
-void priv_lock(struct priv *priv)
-{
- rte_spinlock_lock(&priv->lock);
-}
-
-/**
- * Unlock private structure.
- *
- * @param priv
- * Pointer to private structure.
- */
-void priv_unlock(struct priv *priv)
-{
- rte_spinlock_unlock(&priv->lock);
-}
-
/* Allocate a buffer on the stack and fill it with a printf format string. */
#define MKSTR(name, ...) \
char name[snprintf(NULL, 0, __VA_ARGS__) + 1]; \
}
/**
- * Set device MTU.
+ * DPDK callback to change the MTU.
*
* @param priv
- * Pointer to private structure.
+ * Pointer to Ethernet device structure.
* @param mtu
* MTU value to set.
*
* 0 on success, negative errno value otherwise and rte_errno is set.
*/
static int
-priv_set_mtu(struct priv *priv, uint16_t mtu)
+mlx4_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
{
+ struct priv *priv = dev->data->dev_private;
uint16_t new_mtu;
int ret = priv_set_sysfs_ulong(priv, "mtu", mtu);
ret = priv_get_mtu(priv, &new_mtu);
if (ret)
return ret;
- if (new_mtu == mtu)
+ if (new_mtu == mtu) {
+ priv->mtu = mtu;
return 0;
+ }
rte_errno = EINVAL;
return -rte_errno;
}
priv_mac_addr_del(struct priv *priv);
/**
- * Ethernet device configuration.
+ * DPDK callback for Ethernet device configuration.
*
* Prepare the driver for a given number of TX and RX queues.
*
* 0 on success, negative errno value otherwise and rte_errno is set.
*/
static int
-dev_configure(struct rte_eth_dev *dev)
+mlx4_dev_configure(struct rte_eth_dev *dev)
{
struct priv *priv = dev->data->dev_private;
unsigned int rxqs_n = dev->data->nb_rx_queues;
return 0;
}
-/**
- * DPDK callback for Ethernet device configuration.
- *
- * @param dev
- * Pointer to Ethernet device structure.
- *
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-mlx4_dev_configure(struct rte_eth_dev *dev)
-{
- struct priv *priv = dev->data->dev_private;
- int ret;
-
- priv_lock(priv);
- ret = dev_configure(dev);
- priv_unlock(priv);
- return ret;
-}
-
static uint16_t mlx4_tx_burst(void *, struct rte_mbuf **, uint16_t);
static uint16_t removed_rx_burst(void *, struct rte_mbuf **, uint16_t);
txq->elts_head = 0;
txq->elts_tail = 0;
txq->elts_comp = 0;
- /* Request send completion every MLX4_PMD_TX_PER_COMP_REQ packets or
- * at least 4 times per ring. */
+ /*
+ * Request send completion every MLX4_PMD_TX_PER_COMP_REQ packets or
+ * at least 4 times per ring.
+ */
txq->elts_comp_cd_init =
((MLX4_PMD_TX_PER_COMP_REQ < (elts_n / 4)) ?
MLX4_PMD_TX_PER_COMP_REQ : (elts_n / 4));
return 0;
error:
rte_free(elts);
-
DEBUG("%p: failed, freed everything", (void *)txq);
assert(ret > 0);
rte_errno = ret;
rte_free(elts);
}
-
/**
* Clean up a TX queue.
*
claim_zero(ibv_destroy_qp(txq->qp));
if (txq->cq != NULL)
claim_zero(ibv_destroy_cq(txq->cq));
- for (i = 0; (i != elemof(txq->mp2mr)); ++i) {
+ for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) {
if (txq->mp2mr[i].mp == NULL)
break;
assert(txq->mp2mr[i].mr != NULL);
(void)mp;
(void)mem_idx;
-
/* It already failed, skip the next chunks. */
if (data->ret != 0)
return;
rte_mempool_mem_iter(mp, mlx4_check_mempool_cb, &data);
*start = (uintptr_t)data.start;
*end = (uintptr_t)data.end;
-
return data.ret;
}
(void *)mp);
return NULL;
}
-
DEBUG("mempool %p area start=%p end=%p size=%zu",
(void *)mp, (void *)start, (void *)end,
(size_t)(end - start));
unsigned int i;
struct ibv_mr *mr;
- for (i = 0; (i != elemof(txq->mp2mr)); ++i) {
+ for (i = 0; (i != RTE_DIM(txq->mp2mr)); ++i) {
if (unlikely(txq->mp2mr[i].mp == NULL)) {
/* Unknown MP, add a new MR for it. */
break;
(void *)txq);
return (uint32_t)-1;
}
- if (unlikely(i == elemof(txq->mp2mr))) {
+ if (unlikely(i == RTE_DIM(txq->mp2mr))) {
/* Table is full, remove oldest entry. */
DEBUG("%p: MR <-> MP table full, dropping oldest entry.",
(void *)txq);
struct txq_mp2mr_mbuf_check_data *data = arg;
struct rte_mbuf *buf = obj;
- /* Check whether mbuf structure fits element size and whether mempool
- * pointer is valid. */
+ /*
+ * Check whether mbuf structure fits element size and whether mempool
+ * pointer is valid.
+ */
if (sizeof(*buf) > mp->elt_size || buf->pool != mp)
data->ret = -1;
}
struct txq_elt *elt_next = &(*txq->elts)[elts_head_next];
struct txq_elt *elt = &(*txq->elts)[elts_head];
struct ibv_send_wr *wr = &elt->wr;
- unsigned int segs = NB_SEGS(buf);
+ unsigned int segs = buf->nb_segs;
unsigned int sent_size = 0;
uint32_t send_flags = 0;
#endif
/* Faster than rte_pktmbuf_free(). */
do {
- struct rte_mbuf *next = NEXT(tmp);
+ struct rte_mbuf *next = tmp->next;
rte_pktmbuf_free_seg(tmp);
tmp = next;
/* Retrieve buffer information. */
addr = rte_pktmbuf_mtod(buf, uintptr_t);
- length = DATA_LEN(buf);
+ length = buf->data_len;
/* Retrieve Memory Region key for this memory pool. */
lkey = txq_mp2mr(txq, txq_mb2mp(buf));
if (unlikely(lkey == (uint32_t)-1)) {
.max_inline_data = MLX4_PMD_MAX_INLINE,
},
.qp_type = IBV_QPT_RAW_PACKET,
- /* Do *NOT* enable this, completions events are managed per
- * TX burst. */
+ /*
+ * Do *NOT* enable this, completions events are managed per
+ * Tx burst.
+ */
.sq_sig_all = 0,
};
tmpl.qp = ibv_create_qp(priv->pd, &attr.init);
struct txq *txq = (*priv->txqs)[idx];
int ret;
- priv_lock(priv);
DEBUG("%p: configuring queue %u for %u descriptors",
(void *)dev, idx, desc);
if (idx >= priv->txqs_n) {
rte_errno = EOVERFLOW;
ERROR("%p: queue index out of range (%u >= %u)",
(void *)dev, idx, priv->txqs_n);
- priv_unlock(priv);
return -rte_errno;
}
if (txq != NULL) {
(void *)dev, idx, (void *)txq);
if (priv->started) {
rte_errno = EEXIST;
- priv_unlock(priv);
return -rte_errno;
}
(*priv->txqs)[idx] = NULL;
rte_errno = ENOMEM;
ERROR("%p: unable to allocate queue index %u",
(void *)dev, idx);
- priv_unlock(priv);
return -rte_errno;
}
}
/* Update send callback. */
dev->tx_pkt_burst = mlx4_tx_burst;
}
- priv_unlock(priv);
return ret;
}
if (txq == NULL)
return;
priv = txq->priv;
- priv_lock(priv);
for (i = 0; (i != priv->txqs_n); ++i)
if ((*priv->txqs)[i] == txq) {
DEBUG("%p: removing TX queue %p from list",
}
txq_cleanup(txq);
rte_free(txq);
- priv_unlock(priv);
}
/* RX queues handling. */
wr->sg_list = sge;
wr->num_sge = 1;
/* Headroom is reserved by rte_pktmbuf_alloc(). */
- assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM);
+ assert(buf->data_off == RTE_PKTMBUF_HEADROOM);
/* Buffer is supposed to be empty. */
assert(rte_pktmbuf_data_len(buf) == 0);
assert(rte_pktmbuf_pkt_len(buf) == 0);
return 0;
error:
if (elts != NULL) {
- for (i = 0; (i != elemof(*elts)); ++i)
+ for (i = 0; (i != RTE_DIM(*elts)); ++i)
rte_pktmbuf_free_seg((*elts)[i].buf);
rte_free(elts);
}
rxq->elts = NULL;
if (elts == NULL)
return;
- for (i = 0; (i != elemof(*elts)); ++i)
+ for (i = 0; (i != RTE_DIM(*elts)); ++i)
rte_pktmbuf_free_seg((*elts)[i].buf);
rte_free(elts);
}
++rxq->priv->dev->data->rx_mbuf_alloc_failed;
goto repost;
}
-
/* Reconfigure sge to use rep instead of seg. */
elt->sge.addr = (uintptr_t)rep->buf_addr + RTE_PKTMBUF_HEADROOM;
assert(elt->sge.lkey == rxq->mr->lkey);
elt->buf = rep;
-
/* Update seg information. */
- SET_DATA_OFF(seg, RTE_PKTMBUF_HEADROOM);
- NB_SEGS(seg) = 1;
- PORT(seg) = rxq->port_id;
- NEXT(seg) = NULL;
- PKT_LEN(seg) = len;
- DATA_LEN(seg) = len;
+ seg->data_off = RTE_PKTMBUF_HEADROOM;
+ seg->nb_segs = 1;
+ seg->port = rxq->port_id;
+ seg->next = NULL;
+ seg->pkt_len = len;
+ seg->data_len = len;
seg->packet_type = 0;
seg->ol_flags = 0;
-
/* Return packet. */
*(pkts++) = seg;
++pkts_ret;
(void *)dev, strerror(rte_errno));
goto error;
}
+ if (mlx4_fd_set_non_blocking(tmpl.channel->fd) < 0) {
+ ERROR("%p: unable to make Rx interrupt completion"
+ " channel non-blocking: %s",
+ (void *)dev, strerror(rte_errno));
+ goto error;
+ }
}
tmpl.cq = ibv_create_cq(priv->ctx, desc, NULL, tmpl.channel, 0);
if (tmpl.cq == NULL) {
struct rxq *rxq = (*priv->rxqs)[idx];
int ret;
- priv_lock(priv);
DEBUG("%p: configuring queue %u for %u descriptors",
(void *)dev, idx, desc);
if (idx >= priv->rxqs_n) {
rte_errno = EOVERFLOW;
ERROR("%p: queue index out of range (%u >= %u)",
(void *)dev, idx, priv->rxqs_n);
- priv_unlock(priv);
return -rte_errno;
}
if (rxq != NULL) {
(void *)dev, idx, (void *)rxq);
if (priv->started) {
rte_errno = EEXIST;
- priv_unlock(priv);
return -rte_errno;
}
(*priv->rxqs)[idx] = NULL;
rte_errno = ENOMEM;
ERROR("%p: unable to allocate queue index %u",
(void *)dev, idx);
- priv_unlock(priv);
return -rte_errno;
}
}
/* Update receive callback. */
dev->rx_pkt_burst = mlx4_rx_burst;
}
- priv_unlock(priv);
return ret;
}
if (rxq == NULL)
return;
priv = rxq->priv;
- priv_lock(priv);
for (i = 0; (i != priv->rxqs_n); ++i)
if ((*priv->rxqs)[i] == rxq) {
DEBUG("%p: removing RX queue %p from list",
}
rxq_cleanup(rxq);
rte_free(rxq);
- priv_unlock(priv);
}
-static int
-priv_dev_interrupt_handler_install(struct priv *, struct rte_eth_dev *);
-
-static int
-priv_dev_removal_interrupt_handler_install(struct priv *, struct rte_eth_dev *);
-
-static int
-priv_dev_link_interrupt_handler_install(struct priv *, struct rte_eth_dev *);
-
/**
* DPDK callback to start the device.
*
struct priv *priv = dev->data->dev_private;
int ret;
- priv_lock(priv);
- if (priv->started) {
- priv_unlock(priv);
+ if (priv->started)
return 0;
- }
DEBUG("%p: attaching configured flows to all RX queues", (void *)dev);
priv->started = 1;
ret = priv_mac_addr_add(priv);
if (ret)
goto err;
- ret = priv_dev_link_interrupt_handler_install(priv, dev);
- if (ret) {
- ERROR("%p: LSC handler install failed",
- (void *)dev);
- goto err;
- }
- ret = priv_dev_removal_interrupt_handler_install(priv, dev);
+ ret = mlx4_intr_install(priv);
if (ret) {
- ERROR("%p: RMV handler install failed",
+ ERROR("%p: interrupt handler installation failed",
(void *)dev);
goto err;
}
- ret = priv_rx_intr_vec_enable(priv);
- if (ret) {
- ERROR("%p: Rx interrupt vector creation failed",
- (void *)dev);
- goto err;
- }
ret = mlx4_priv_flow_start(priv);
if (ret) {
ERROR("%p: flow start failed: %s",
(void *)dev, strerror(ret));
goto err;
}
- priv_unlock(priv);
return 0;
err:
/* Rollback. */
priv_mac_addr_del(priv);
priv->started = 0;
- priv_unlock(priv);
return ret;
}
{
struct priv *priv = dev->data->dev_private;
- priv_lock(priv);
- if (!priv->started) {
- priv_unlock(priv);
+ if (!priv->started)
return;
- }
DEBUG("%p: detaching flows from all RX queues", (void *)dev);
priv->started = 0;
mlx4_priv_flow_stop(priv);
+ mlx4_intr_uninstall(priv);
priv_mac_addr_del(priv);
- priv_unlock(priv);
}
/**
return 0;
}
-static int
-priv_dev_interrupt_handler_uninstall(struct priv *, struct rte_eth_dev *);
-
-static int
-priv_dev_removal_interrupt_handler_uninstall(struct priv *,
- struct rte_eth_dev *);
-
-static int
-priv_dev_link_interrupt_handler_uninstall(struct priv *, struct rte_eth_dev *);
-
/**
* DPDK callback to close the device.
*
if (priv == NULL)
return;
- priv_lock(priv);
DEBUG("%p: closing device \"%s\"",
(void *)dev,
((priv->ctx != NULL) ? priv->ctx->device->name : ""));
priv_mac_addr_del(priv);
- /* Prevent crashes when queues are still in use. This is unfortunately
+ /*
+ * Prevent crashes when queues are still in use. This is unfortunately
* still required for DPDK 1.3 because some programs (such as testpmd)
- * never release them before closing the device. */
+ * never release them before closing the device.
+ */
dev->rx_pkt_burst = removed_rx_burst;
dev->tx_pkt_burst = removed_tx_burst;
if (priv->rxqs != NULL) {
claim_zero(ibv_close_device(priv->ctx));
} else
assert(priv->ctx == NULL);
- priv_dev_removal_interrupt_handler_uninstall(priv, dev);
- priv_dev_link_interrupt_handler_uninstall(priv, dev);
- priv_rx_intr_vec_disable(priv);
- priv_unlock(priv);
+ mlx4_intr_uninstall(priv);
memset(priv, 0, sizeof(*priv));
}
mlx4_set_link_down(struct rte_eth_dev *dev)
{
struct priv *priv = dev->data->dev_private;
- int err;
- priv_lock(priv);
- err = priv_set_link(priv, 0);
- priv_unlock(priv);
- return err;
+ return priv_set_link(priv, 0);
}
/**
mlx4_set_link_up(struct rte_eth_dev *dev)
{
struct priv *priv = dev->data->dev_private;
- int err;
- priv_lock(priv);
- err = priv_set_link(priv, 1);
- priv_unlock(priv);
- return err;
+ return priv_set_link(priv, 1);
}
+
/**
* DPDK callback to get information about the device.
*
char ifname[IF_NAMESIZE];
info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-
if (priv == NULL)
return;
- priv_lock(priv);
/* FIXME: we should ask the device for these values. */
info->min_rx_bufsize = 32;
info->max_rx_pktlen = 65536;
ETH_LINK_SPEED_20G |
ETH_LINK_SPEED_40G |
ETH_LINK_SPEED_56G;
- priv_unlock(priv);
}
/**
if (priv == NULL)
return;
- priv_lock(priv);
/* Add software counters. */
for (i = 0; (i != priv->rxqs_n); ++i) {
struct rxq *rxq = (*priv->rxqs)[i];
tmp.oerrors += txq->stats.odropped;
}
*stats = tmp;
- priv_unlock(priv);
}
/**
if (priv == NULL)
return;
- priv_lock(priv);
for (i = 0; (i != priv->rxqs_n); ++i) {
if ((*priv->rxqs)[i] == NULL)
continue;
(*priv->txqs)[i]->stats =
(struct mlx4_txq_stats){ .idx = idx };
}
- priv_unlock(priv);
}
/**
* @return
* 0 on success, negative errno value otherwise and rte_errno is set.
*/
-static int
+int
mlx4_link_update(struct rte_eth_dev *dev, int wait_to_complete)
{
const struct priv *priv = dev->data->dev_private;
struct rte_eth_link dev_link;
int link_speed = 0;
- /* priv_lock() is not taken to allow concurrent calls. */
-
if (priv == NULL) {
rte_errno = EINVAL;
return -rte_errno;
return 0;
}
-/**
- * DPDK callback to change the MTU.
- *
- * @param dev
- * Pointer to Ethernet device structure.
- * @param in_mtu
- * New MTU.
- *
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-mlx4_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
-{
- struct priv *priv = dev->data->dev_private;
- int ret = 0;
-
- priv_lock(priv);
- /* Set kernel interface MTU first. */
- if (priv_set_mtu(priv, mtu)) {
- ret = rte_errno;
- WARN("cannot set port %u MTU to %u: %s", priv->port, mtu,
- strerror(rte_errno));
- goto out;
- } else
- DEBUG("adapter port %u MTU set to %u", priv->port, mtu);
- priv->mtu = mtu;
-out:
- priv_unlock(priv);
- assert(ret >= 0);
- return -ret;
-}
-
/**
* DPDK callback to get flow control status.
*
int ret;
ifr.ifr_data = (void *)ðpause;
- priv_lock(priv);
if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
ret = rte_errno;
WARN("ioctl(SIOCETHTOOL, ETHTOOL_GPAUSEPARAM)"
strerror(rte_errno));
goto out;
}
-
fc_conf->autoneg = ethpause.autoneg;
if (ethpause.rx_pause && ethpause.tx_pause)
fc_conf->mode = RTE_FC_FULL;
else
fc_conf->mode = RTE_FC_NONE;
ret = 0;
-
out:
- priv_unlock(priv);
assert(ret >= 0);
return -ret;
}
ethpause.rx_pause = 1;
else
ethpause.rx_pause = 0;
-
if (((fc_conf->mode & RTE_FC_FULL) == RTE_FC_FULL) ||
(fc_conf->mode & RTE_FC_TX_PAUSE))
ethpause.tx_pause = 1;
else
ethpause.tx_pause = 0;
-
- priv_lock(priv);
if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
ret = rte_errno;
WARN("ioctl(SIOCETHTOOL, ETHTOOL_SPAUSEPARAM)"
goto out;
}
ret = 0;
-
out:
- priv_unlock(priv);
assert(ret >= 0);
return -ret;
}
return 0;
}
-static void
-mlx4_dev_link_status_handler(void *);
-static void
-mlx4_dev_interrupt_handler(void *);
-
-/**
- * Link/device status handler.
- *
- * @param priv
- * Pointer to private structure.
- * @param dev
- * Pointer to the rte_eth_dev structure.
- * @param events
- * Pointer to event flags holder.
- *
- * @return
- * Number of events
- */
-static int
-priv_dev_status_handler(struct priv *priv, struct rte_eth_dev *dev,
- uint32_t *events)
-{
- struct ibv_async_event event;
- int port_change = 0;
- struct rte_eth_link *link = &dev->data->dev_link;
- int ret = 0;
-
- *events = 0;
- /* Read all message and acknowledge them. */
- for (;;) {
- if (ibv_get_async_event(priv->ctx, &event))
- break;
- if ((event.event_type == IBV_EVENT_PORT_ACTIVE ||
- event.event_type == IBV_EVENT_PORT_ERR) &&
- (priv->intr_conf.lsc == 1)) {
- port_change = 1;
- ret++;
- } else if (event.event_type == IBV_EVENT_DEVICE_FATAL &&
- priv->intr_conf.rmv == 1) {
- *events |= (1 << RTE_ETH_EVENT_INTR_RMV);
- ret++;
- } else
- DEBUG("event type %d on port %d not handled",
- event.event_type, event.element.port_num);
- ibv_ack_async_event(&event);
- }
- if (!port_change)
- return ret;
- mlx4_link_update(dev, 0);
- if (((link->link_speed == 0) && link->link_status) ||
- ((link->link_speed != 0) && !link->link_status)) {
- if (!priv->pending_alarm) {
- /* Inconsistent status, check again later. */
- priv->pending_alarm = 1;
- rte_eal_alarm_set(MLX4_ALARM_TIMEOUT_US,
- mlx4_dev_link_status_handler,
- dev);
- }
- } else {
- *events |= (1 << RTE_ETH_EVENT_INTR_LSC);
- }
- return ret;
-}
-
-/**
- * Handle delayed link status event.
- *
- * @param arg
- * Registered argument.
- */
-static void
-mlx4_dev_link_status_handler(void *arg)
-{
- struct rte_eth_dev *dev = arg;
- struct priv *priv = dev->data->dev_private;
- uint32_t events;
- int ret;
-
- priv_lock(priv);
- assert(priv->pending_alarm == 1);
- priv->pending_alarm = 0;
- ret = priv_dev_status_handler(priv, dev, &events);
- priv_unlock(priv);
- if (ret > 0 && events & (1 << RTE_ETH_EVENT_INTR_LSC))
- _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL,
- NULL);
-}
-
-/**
- * Handle interrupts from the NIC.
- *
- * @param[in] intr_handle
- * Interrupt handler.
- * @param cb_arg
- * Callback argument.
- */
-static void
-mlx4_dev_interrupt_handler(void *cb_arg)
-{
- struct rte_eth_dev *dev = cb_arg;
- struct priv *priv = dev->data->dev_private;
- int ret;
- uint32_t ev;
- int i;
-
- priv_lock(priv);
- ret = priv_dev_status_handler(priv, dev, &ev);
- priv_unlock(priv);
- if (ret > 0) {
- for (i = RTE_ETH_EVENT_UNKNOWN;
- i < RTE_ETH_EVENT_MAX;
- i++) {
- if (ev & (1 << i)) {
- ev &= ~(1 << i);
- _rte_eth_dev_callback_process(dev, i, NULL,
- NULL);
- ret--;
- }
- }
- if (ret)
- WARN("%d event%s not processed", ret,
- (ret > 1 ? "s were" : " was"));
- }
-}
-
-/**
- * Uninstall interrupt handler.
- *
- * @param priv
- * Pointer to private structure.
- * @param dev
- * Pointer to the rte_eth_dev structure.
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev)
-{
- int ret;
-
- if (priv->intr_conf.lsc ||
- priv->intr_conf.rmv)
- return 0;
- ret = rte_intr_callback_unregister(&priv->intr_handle,
- mlx4_dev_interrupt_handler,
- dev);
- if (ret < 0) {
- rte_errno = ret;
- ERROR("rte_intr_callback_unregister failed with %d %s",
- ret, strerror(rte_errno));
- }
- priv->intr_handle.fd = 0;
- priv->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN;
- return ret;
-}
-
-/**
- * Install interrupt handler.
- *
- * @param priv
- * Pointer to private structure.
- * @param dev
- * Pointer to the rte_eth_dev structure.
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-priv_dev_interrupt_handler_install(struct priv *priv,
- struct rte_eth_dev *dev)
-{
- int flags;
- int rc;
-
- /* Check whether the interrupt handler has already been installed
- * for either type of interrupt
- */
- if (priv->intr_conf.lsc &&
- priv->intr_conf.rmv &&
- priv->intr_handle.fd)
- return 0;
- assert(priv->ctx->async_fd > 0);
- flags = fcntl(priv->ctx->async_fd, F_GETFL);
- rc = fcntl(priv->ctx->async_fd, F_SETFL, flags | O_NONBLOCK);
- if (rc < 0) {
- rte_errno = errno ? errno : EINVAL;
- INFO("failed to change file descriptor async event queue");
- dev->data->dev_conf.intr_conf.lsc = 0;
- dev->data->dev_conf.intr_conf.rmv = 0;
- return -rte_errno;
- } else {
- priv->intr_handle.fd = priv->ctx->async_fd;
- priv->intr_handle.type = RTE_INTR_HANDLE_EXT;
- rc = rte_intr_callback_register(&priv->intr_handle,
- mlx4_dev_interrupt_handler,
- dev);
- if (rc) {
- rte_errno = -rc;
- ERROR("rte_intr_callback_register failed "
- " (rte_errno: %s)", strerror(rte_errno));
- return -rte_errno;
- }
- }
- return 0;
-}
-
-/**
- * Uninstall interrupt handler.
- *
- * @param priv
- * Pointer to private structure.
- * @param dev
- * Pointer to the rte_eth_dev structure.
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-priv_dev_removal_interrupt_handler_uninstall(struct priv *priv,
- struct rte_eth_dev *dev)
-{
- if (dev->data->dev_conf.intr_conf.rmv) {
- priv->intr_conf.rmv = 0;
- return priv_dev_interrupt_handler_uninstall(priv, dev);
- }
- return 0;
-}
-
-/**
- * Uninstall interrupt handler.
- *
- * @param priv
- * Pointer to private structure.
- * @param dev
- * Pointer to the rte_eth_dev structure.
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-priv_dev_link_interrupt_handler_uninstall(struct priv *priv,
- struct rte_eth_dev *dev)
-{
- int ret = 0;
-
- if (dev->data->dev_conf.intr_conf.lsc) {
- priv->intr_conf.lsc = 0;
- ret = priv_dev_interrupt_handler_uninstall(priv, dev);
- if (ret)
- return ret;
- }
- if (priv->pending_alarm)
- if (rte_eal_alarm_cancel(mlx4_dev_link_status_handler,
- dev)) {
- ERROR("rte_eal_alarm_cancel failed "
- " (rte_errno: %s)", strerror(rte_errno));
- return -rte_errno;
- }
- priv->pending_alarm = 0;
- return 0;
-}
-
-/**
- * Install link interrupt handler.
- *
- * @param priv
- * Pointer to private structure.
- * @param dev
- * Pointer to the rte_eth_dev structure.
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-priv_dev_link_interrupt_handler_install(struct priv *priv,
- struct rte_eth_dev *dev)
-{
- int ret;
-
- if (dev->data->dev_conf.intr_conf.lsc) {
- ret = priv_dev_interrupt_handler_install(priv, dev);
- if (ret)
- return ret;
- priv->intr_conf.lsc = 1;
- }
- return 0;
-}
-
-/**
- * Install removal interrupt handler.
- *
- * @param priv
- * Pointer to private structure.
- * @param dev
- * Pointer to the rte_eth_dev structure.
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-priv_dev_removal_interrupt_handler_install(struct priv *priv,
- struct rte_eth_dev *dev)
-{
- int ret;
-
- if (dev->data->dev_conf.intr_conf.rmv) {
- ret = priv_dev_interrupt_handler_install(priv, dev);
- if (ret)
- return ret;
- priv->intr_conf.rmv = 1;
- }
- return 0;
-}
-
-/**
- * Allocate queue vector and fill epoll fd list for Rx interrupts.
- *
- * @param priv
- * Pointer to private structure.
- *
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-priv_rx_intr_vec_enable(struct priv *priv)
-{
- unsigned int i;
- unsigned int rxqs_n = priv->rxqs_n;
- unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
- unsigned int count = 0;
- struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
-
- if (!priv->dev->data->dev_conf.intr_conf.rxq)
- return 0;
- priv_rx_intr_vec_disable(priv);
- intr_handle->intr_vec = malloc(sizeof(intr_handle->intr_vec[rxqs_n]));
- if (intr_handle->intr_vec == NULL) {
- rte_errno = ENOMEM;
- ERROR("failed to allocate memory for interrupt vector,"
- " Rx interrupts will not be supported");
- return -rte_errno;
- }
- intr_handle->type = RTE_INTR_HANDLE_EXT;
- for (i = 0; i != n; ++i) {
- struct rxq *rxq = (*priv->rxqs)[i];
- int fd;
- int flags;
- int rc;
-
- /* Skip queues that cannot request interrupts. */
- if (!rxq || !rxq->channel) {
- /* Use invalid intr_vec[] index to disable entry. */
- intr_handle->intr_vec[i] =
- RTE_INTR_VEC_RXTX_OFFSET +
- RTE_MAX_RXTX_INTR_VEC_ID;
- continue;
- }
- if (count >= RTE_MAX_RXTX_INTR_VEC_ID) {
- rte_errno = E2BIG;
- ERROR("too many Rx queues for interrupt vector size"
- " (%d), Rx interrupts cannot be enabled",
- RTE_MAX_RXTX_INTR_VEC_ID);
- priv_rx_intr_vec_disable(priv);
- return -rte_errno;
- }
- fd = rxq->channel->fd;
- flags = fcntl(fd, F_GETFL);
- rc = fcntl(fd, F_SETFL, flags | O_NONBLOCK);
- if (rc < 0) {
- rte_errno = errno;
- ERROR("failed to make Rx interrupt file descriptor"
- " %d non-blocking for queue index %d", fd, i);
- priv_rx_intr_vec_disable(priv);
- return -rte_errno;
- }
- intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + count;
- intr_handle->efds[count] = fd;
- count++;
- }
- if (!count)
- priv_rx_intr_vec_disable(priv);
- else
- intr_handle->nb_efd = count;
- return 0;
-}
-
-/**
- * Clean up Rx interrupts handler.
- *
- * @param priv
- * Pointer to private structure.
- */
-static void
-priv_rx_intr_vec_disable(struct priv *priv)
-{
- struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
-
- rte_intr_free_epoll_fd(intr_handle);
- free(intr_handle->intr_vec);
- intr_handle->nb_efd = 0;
- intr_handle->intr_vec = NULL;
-}
-
-/**
- * DPDK callback for Rx queue interrupt enable.
- *
- * @param dev
- * Pointer to Ethernet device structure.
- * @param idx
- * Rx queue index.
- *
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx)
-{
- struct priv *priv = dev->data->dev_private;
- struct rxq *rxq = (*priv->rxqs)[idx];
- int ret;
-
- if (!rxq || !rxq->channel)
- ret = EINVAL;
- else
- ret = ibv_req_notify_cq(rxq->cq, 0);
- if (ret) {
- rte_errno = ret;
- WARN("unable to arm interrupt on rx queue %d", idx);
- }
- return -ret;
-}
-
-/**
- * DPDK callback for Rx queue interrupt disable.
- *
- * @param dev
- * Pointer to Ethernet device structure.
- * @param idx
- * Rx queue index.
- *
- * @return
- * 0 on success, negative errno value otherwise and rte_errno is set.
- */
-static int
-mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx)
-{
- struct priv *priv = dev->data->dev_private;
- struct rxq *rxq = (*priv->rxqs)[idx];
- struct ibv_cq *ev_cq;
- void *ev_ctx;
- int ret;
-
- if (!rxq || !rxq->channel) {
- ret = EINVAL;
- } else {
- ret = ibv_get_cq_event(rxq->cq->channel, &ev_cq, &ev_ctx);
- if (ret || ev_cq != rxq->cq)
- ret = EINVAL;
- }
- if (ret) {
- rte_errno = ret;
- WARN("unable to disable interrupt on rx queue %d",
- idx);
- } else {
- ibv_ack_cq_events(rxq->cq, 1);
- }
- return -ret;
-}
-
/**
* Verify and store value for device argument.
*
(void)pci_drv;
assert(pci_drv == &mlx4_driver);
-
list = ibv_get_device_list(&i);
if (list == NULL) {
rte_errno = errno;
return -rte_errno;
}
ibv_dev = list[i];
-
DEBUG("device opened");
if (ibv_query_device(attr_ctx, &device_attr)) {
rte_errno = ENODEV;
goto error;
}
INFO("%u port(s) detected", device_attr.phys_port_cnt);
-
conf.ports.present |= (UINT64_C(1) << device_attr.phys_port_cnt) - 1;
if (mlx4_args(pci_dev->device.devargs, &conf)) {
ERROR("failed to process device arguments");
/* If port is not enabled, skip. */
if (!(conf.ports.enabled & (1 << i)))
continue;
-
DEBUG("using port %u", port);
-
ctx = ibv_open_device(ibv_dev);
if (ctx == NULL) {
rte_errno = ENODEV;
goto port_error;
}
-
/* Check port status. */
err = ibv_query_port(ctx, port, &port_attr);
if (err) {
ERROR("port query failed: %s", strerror(rte_errno));
goto port_error;
}
-
if (port_attr.link_layer != IBV_LINK_LAYER_ETHERNET) {
rte_errno = ENOTSUP;
ERROR("port %d is not configured in Ethernet mode",
port);
goto port_error;
}
-
if (port_attr.state != IBV_PORT_ACTIVE)
DEBUG("port %d is not active: \"%s\" (%d)",
port, ibv_port_state_str(port_attr.state),
port_attr.state);
-
+ /* Make asynchronous FD non-blocking to handle interrupts. */
+ if (mlx4_fd_set_non_blocking(ctx->async_fd) < 0) {
+ ERROR("cannot make asynchronous FD non-blocking: %s",
+ strerror(rte_errno));
+ goto port_error;
+ }
/* Allocate protection domain. */
pd = ibv_alloc_pd(ctx);
if (pd == NULL) {
ERROR("PD allocation failure");
goto port_error;
}
-
/* from rte_ethdev.c */
priv = rte_zmalloc("ethdev private structure",
sizeof(*priv),
ERROR("priv allocation failure");
goto port_error;
}
-
priv->ctx = ctx;
priv->device_attr = device_attr;
priv->port = port;
priv->pd = pd;
priv->mtu = ETHER_MTU;
-
priv->vf = vf;
/* Configure the first MAC address by default. */
if (priv_get_mac(priv, &mac.addr_bytes)) {
/* Get actual MTU if possible. */
priv_get_mtu(priv, &priv->mtu);
DEBUG("port %u MTU is %u", priv->port, priv->mtu);
-
/* from rte_ethdev.c */
{
char name[RTE_ETH_NAME_MAX_LEN];
rte_errno = ENOMEM;
goto port_error;
}
-
eth_dev->data->dev_private = priv;
eth_dev->data->mac_addrs = &priv->mac;
eth_dev->device = &pci_dev->device;
-
rte_eth_copy_pci_info(eth_dev, pci_dev);
-
eth_dev->device->driver = &mlx4_driver.driver;
-
+ /* Initialize local interrupt handle for current port. */
+ priv->intr_handle = (struct rte_intr_handle){
+ .fd = -1,
+ .type = RTE_INTR_HANDLE_EXT,
+ };
/*
- * Copy and override interrupt handle to prevent it from
- * being shared between all ethdev instances of a given PCI
- * device. This is required to properly handle Rx interrupts
- * on all ports.
+ * Override ethdev interrupt handle pointer with private
+ * handle instead of that of the parent PCI device used by
+ * default. This prevents it from being shared between all
+ * ports of the same PCI device since each of them is
+ * associated its own Verbs context.
+ *
+ * Rx interrupts in particular require this as the PMD has
+ * no control over the registration of queue interrupts
+ * besides setting up eth_dev->intr_handle, the rest is
+ * handled by rte_intr_rx_ctl().
*/
- priv->intr_handle_dev = *eth_dev->intr_handle;
- eth_dev->intr_handle = &priv->intr_handle_dev;
-
+ eth_dev->intr_handle = &priv->intr_handle;
priv->dev = eth_dev;
eth_dev->dev_ops = &mlx4_dev_ops;
eth_dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE;
-
/* Bring Ethernet device up. */
DEBUG("forcing Ethernet interface up");
priv_set_flags(priv, ~IFF_UP, IFF_UP);
if (eth_dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC)
mlx4_link_update(eth_dev, 0);
continue;
-
port_error:
rte_free(priv);
if (pd)
}
if (i == device_attr.phys_port_cnt)
return 0;
-
/*
* XXX if something went wrong in the loop above, there is a resource
* leak (ctx, pd, priv, dpdk ethdev) but we can do nothing about it as
* long as the dpdk does not provide a way to deallocate a ethdev and a
* way to enumerate the registered ethdevs to free the previous ones.
*/
-
error:
if (attr_ctx)
claim_zero(ibv_close_device(attr_ctx));