net/mlx4: support user space Rx interrupt event
authorMoti Haimovsky <motih@mellanox.com>
Tue, 6 Jun 2017 14:48:29 +0000 (17:48 +0300)
committerFerruh Yigit <ferruh.yigit@intel.com>
Mon, 12 Jun 2017 09:41:29 +0000 (10:41 +0100)
Implement rxq interrupt callbacks

Signed-off-by: Moti Haimovsky <motih@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
doc/guides/nics/features/mlx4.ini
doc/guides/rel_notes/release_17_08.rst
drivers/net/mlx4/mlx4.c
drivers/net/mlx4/mlx4.h

index 285f0ec..320cc11 100644 (file)
@@ -7,6 +7,7 @@
 Link status          = Y
 Link status event    = Y
 Removal event        = Y
+Rx interrupt         = Y
 Queue start/stop     = Y
 MTU update           = Y
 Jumbo frame          = Y
index 7bf6d25..eb31da2 100644 (file)
@@ -66,6 +66,11 @@ New Features
   * Add PMD specific APIs mainly to control VF from PF.
   * Update HWRM version to 1.7.7
 
+* **Added support for Rx interrupts on mlx4 driver.**
+
+  Rx queues can be armed with an interrupt which will trigger on the
+  next packet arrival.
+
 
 Resolved Issues
 ---------------
index ea314b8..5bc2a50 100644 (file)
@@ -75,6 +75,7 @@
 #include <rte_memory.h>
 #include <rte_flow.h>
 #include <rte_kvargs.h>
+#include <rte_interrupts.h>
 
 /* Generated configuration header. */
 #include "mlx4_autoconf.h"
@@ -127,6 +128,24 @@ const char *pmd_mlx4_init_params[] = {
        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_intr_efd_enable(struct priv *priv);
+
+static void
+priv_intr_efd_disable(struct priv *priv);
+
+static int
+priv_create_intr_vec(struct priv *priv);
+
+static void
+priv_destroy_intr_vec(struct priv *priv);
+
 /**
  * Check if running as a secondary process.
  *
@@ -2756,6 +2775,8 @@ rxq_cleanup(struct rxq *rxq)
        }
        if (rxq->cq != NULL)
                claim_zero(ibv_destroy_cq(rxq->cq));
+       if (rxq->channel != NULL)
+               claim_zero(ibv_destroy_comp_channel(rxq->channel));
        if (rxq->rd != NULL) {
                struct ibv_exp_destroy_res_domain_attr attr = {
                        .comp_mask = 0,
@@ -3696,11 +3717,22 @@ skip_mr:
                      (void *)dev, strerror(ret));
                goto error;
        }
+       if (dev->data->dev_conf.intr_conf.rxq) {
+               tmpl.channel = ibv_create_comp_channel(priv->ctx);
+               if (tmpl.channel == NULL) {
+                       dev->data->dev_conf.intr_conf.rxq = 0;
+                       ret = ENOMEM;
+                       ERROR("%p: Comp Channel creation failure: %s",
+                             (void *)dev, strerror(ret));
+                       goto error;
+               }
+       }
        attr.cq = (struct ibv_exp_cq_init_attr){
                .comp_mask = IBV_EXP_CQ_INIT_ATTR_RES_DOMAIN,
                .res_domain = tmpl.rd,
        };
-       tmpl.cq = ibv_exp_create_cq(priv->ctx, desc, NULL, NULL, 0, &attr.cq);
+       tmpl.cq = ibv_exp_create_cq(priv->ctx, desc, NULL, tmpl.channel, 0,
+                                   &attr.cq);
        if (tmpl.cq == NULL) {
                ret = ENOMEM;
                ERROR("%p: CQ creation failure: %s",
@@ -4005,6 +4037,11 @@ mlx4_dev_start(struct rte_eth_dev *dev)
                     (void *)dev);
                goto err;
        }
+       if (dev->data->dev_conf.intr_conf.rxq) {
+               ret = priv_intr_efd_enable(priv);
+               if (!ret)
+                       ret = priv_create_intr_vec(priv);
+       }
        ret = mlx4_priv_flow_start(priv);
        if (ret) {
                ERROR("%p: flow start failed: %s",
@@ -4197,6 +4234,10 @@ mlx4_dev_close(struct rte_eth_dev *dev)
                assert(priv->ctx == NULL);
        priv_dev_removal_interrupt_handler_uninstall(priv, dev);
        priv_dev_link_interrupt_handler_uninstall(priv, dev);
+       if (priv->dev->data->dev_conf.intr_conf.rxq) {
+               priv_destroy_intr_vec(priv);
+               priv_intr_efd_disable(priv);
+       }
        priv_unlock(priv);
        memset(priv, 0, sizeof(*priv));
 }
@@ -5157,6 +5198,8 @@ static const struct eth_dev_ops mlx4_dev_ops = {
        .mac_addr_set = mlx4_mac_addr_set,
        .mtu_set = mlx4_dev_set_mtu,
        .filter_ctrl = mlx4_dev_filter_ctrl,
+       .rx_queue_intr_enable = mlx4_rx_intr_enable,
+       .rx_queue_intr_disable = mlx4_rx_intr_disable,
 };
 
 /**
@@ -5591,6 +5634,168 @@ priv_dev_removal_interrupt_handler_install(struct priv *priv,
        return 0;
 }
 
+/**
+ * Fill epoll fd list for rxq interrupts.
+ *
+ * @param priv
+ *   Poinetr to private structure.
+ *
+ * @return
+ *   0 on success, negative on failure.
+ */
+static int
+priv_intr_efd_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);
+       struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
+
+       if (n == 0)
+               return 0;
+       if (n < rxqs_n) {
+               WARN("rxqs num is larger than EAL max interrupt vector "
+                    "%u > %u unable to supprt rxq interrupts",
+                    rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
+               return -EINVAL;
+       }
+       intr_handle->type = RTE_INTR_HANDLE_EXT;
+       for (i = 0; i != n; ++i) {
+               struct rxq *rxq = (*priv->rxqs)[i];
+               int fd = rxq->channel->fd;
+               int flags;
+               int rc;
+
+               flags = fcntl(fd, F_GETFL);
+               rc = fcntl(fd, F_SETFL, flags | O_NONBLOCK);
+               if (rc < 0) {
+                       WARN("failed to change rxq interrupt file "
+                            "descriptor %d for queue index %d", fd, i);
+                       return rc;
+               }
+               intr_handle->efds[i] = fd;
+       }
+       intr_handle->nb_efd = n;
+       return 0;
+}
+
+/**
+ * Clean epoll fd list for rxq interrupts.
+ *
+ * @param priv
+ *   Ponter to private structure.
+ */
+static void
+priv_intr_efd_disable(struct priv *priv)
+{
+       struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
+
+       rte_intr_free_epoll_fd(intr_handle);
+}
+
+/**
+ * Create and init interrupt vector array.
+ *
+ * @param priv
+ *   Pointer to private structure.
+ *
+ * @return
+ *   0 on success, negative on failure.
+ */
+static int
+priv_create_intr_vec(struct priv *priv)
+{
+       unsigned int rxqs_n = priv->rxqs_n;
+       unsigned int i;
+       struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
+
+       if (rxqs_n == 0)
+               return 0;
+       intr_handle->intr_vec = (int *)
+               rte_malloc("intr_vec", rxqs_n * sizeof(int), 0);
+       if (intr_handle->intr_vec == NULL) {
+               WARN("Failed to allocate memory for intr_vec "
+                    "rxq interrupt will not be supported");
+               return -ENOMEM;
+       }
+       for (i = 0; i != rxqs_n; ++i) {
+               /* 1:1 mapping between rxq and interrupt. */
+               intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + i;
+       }
+       return 0;
+}
+
+/**
+ * Destroy init interrupt vector array.
+ *
+ * @param priv
+ *   Pointer to private structure.
+ */
+static void
+priv_destroy_intr_vec(struct priv *priv)
+{
+       struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
+
+       rte_free(intr_handle->intr_vec);
+}
+
+/**
+ * DPDK callback for rx queue interrupt enable.
+ *
+ * @param dev
+ *   Pointer to Ethernet device structure.
+ * @param idx
+ *   RX queue index.
+ *
+ * @return
+ *   0 on success, negative on failure.
+ */
+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];
+       struct ibv_cq *cq = rxq->cq;
+       int ret = 0;
+
+       ret = ibv_req_notify_cq(cq, 0);
+       if (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 on failure.
+ */
+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 *cq = rxq->cq;
+       struct ibv_cq *ev_cq;
+       void *ev_ctx;
+       int ret = 0;
+
+       ret = ibv_get_cq_event(cq->channel, &ev_cq, &ev_ctx);
+       if (ret || ev_cq != cq)
+               ret = -1;
+       else
+               ibv_ack_cq_events(cq, 1);
+       if (ret)
+               WARN("unable to disable interrupt on rx queue %d",
+                    idx);
+       return ret;
+}
+
 /**
  * Verify and store value for device argument.
  *
index 9a3bae9..c46fc23 100644 (file)
@@ -226,6 +226,7 @@ struct rxq {
        struct ibv_qp *qp; /* Queue Pair. */
        struct ibv_exp_qp_burst_family *if_qp; /* QP burst interface. */
        struct ibv_exp_cq_family *if_cq; /* CQ interface. */
+       struct ibv_comp_channel *channel;
        /*
         * Each VLAN ID requires a separate flow steering rule.
         */