X-Git-Url: http://git.cascardo.eti.br/?a=blobdiff_plain;f=lib%2Fnetdev-dpdk.c;h=e0cacec37df6fc096b4536208bde93b9c6deaafa;hb=585a5beaa2a45dd43f41543b7c9800d13083bc1a;hp=1e43daec90778ea3bc1bff7c3badb77ee0e97dc9;hpb=e5c0f5a4452e368bbe4f24e9f5a591965399d0fa;p=cascardo%2Fovs.git

diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
index 1e43daec9..e0cacec37 100644
--- a/lib/netdev-dpdk.c
+++ b/lib/netdev-dpdk.c
@@ -100,6 +100,8 @@ BUILD_ASSERT_DECL((MAX_NB_MBUF / ROUND_DOWN_POW2(MAX_NB_MBUF/MIN_NB_MBUF))
 #define NIC_PORT_RX_Q_SIZE 2048  /* Size of Physical NIC RX Queue, Max (n+32<=4096)*/
 #define NIC_PORT_TX_Q_SIZE 2048  /* Size of Physical NIC TX Queue, Max (n+32<=4096)*/
 
+#define OVS_VHOST_MAX_QUEUE_NUM 1024  /* Maximum number of vHost TX queues. */
+
 static char *cuse_dev_name = NULL;    /* Character device cuse_dev_name. */
 static char *vhost_sock_dir = NULL;   /* Location of vhost-user sockets */
 
@@ -173,6 +175,8 @@ struct dpdk_tx_queue {
                                     * from concurrent access.  It is used only
                                     * if the queue is shared among different
                                     * pmd threads (see 'txq_needs_locking'). */
+    int map;                       /* Mapping of configured vhost-user queues
+                                    * to enabled by guest. */
     uint64_t tsc;
     struct rte_mbuf *burst_pkts[MAX_TX_QUEUE_LEN];
 };
@@ -188,7 +192,7 @@ struct dpdk_ring {
     /* For the client rings */
     struct rte_ring *cring_tx;
     struct rte_ring *cring_rx;
-    int user_port_id; /* User given port no, parsed from port name */
+    unsigned int user_port_id; /* User given port no, parsed from port name */
     int eth_port_id; /* ethernet device port id */
     struct ovs_list list_node OVS_GUARDED_BY(dpdk_mutex);
 };
@@ -572,6 +576,9 @@ netdev_dpdk_alloc_txq(struct netdev_dpdk *netdev, unsigned int n_txqs)
             /* Queues are shared among CPUs. Always flush */
             netdev->tx_q[i].flush_tx = true;
         }
+
+        /* Initialize map for vhost devices. */
+        netdev->tx_q[i].map = -1;
         rte_spinlock_init(&netdev->tx_q[i].tx_lock);
     }
 }
@@ -614,6 +621,7 @@ netdev_dpdk_init(struct netdev *netdev_, unsigned int port_no,
 
     netdev_->n_txq = NR_QUEUE;
     netdev_->n_rxq = NR_QUEUE;
+    netdev_->requested_n_rxq = NR_QUEUE;
     netdev->real_n_txq = NR_QUEUE;
 
     if (type == DPDK_DEV_ETH) {
@@ -622,6 +630,8 @@ netdev_dpdk_init(struct netdev *netdev_, unsigned int port_no,
         if (err) {
             goto unlock;
         }
+    } else {
+        netdev_dpdk_alloc_txq(netdev, OVS_VHOST_MAX_QUEUE_NUM);
     }
 
     list_push_back(&dpdk_list, &netdev->list_node);
@@ -634,6 +644,8 @@ unlock:
     return err;
 }
 
+/* dev_name must be the prefix followed by a positive decimal number.
+ * (no leading + or - signs are allowed) */
 static int
 dpdk_dev_parse_name(const char dev_name[], const char prefix[],
                     unsigned int *port_no)
@@ -645,8 +657,12 @@ dpdk_dev_parse_name(const char dev_name[], const char prefix[],
     }
 
     cport = dev_name + strlen(prefix);
-    *port_no = strtol(cport, NULL, 0); /* string must be null terminated */
-    return 0;
+
+    if (str_to_uint(cport, 10, port_no)) {
+        return 0;
+    } else {
+        return ENODEV;
+    }
 }
 
 static int
@@ -676,24 +692,37 @@ static int
 netdev_dpdk_vhost_user_construct(struct netdev *netdev_)
 {
     struct netdev_dpdk *netdev = netdev_dpdk_cast(netdev_);
+    const char *name = netdev_->name;
     int err;
 
+    /* 'name' is appended to 'vhost_sock_dir' and used to create a socket in
+     * the file system. '/' or '\' would traverse directories, so they're not
+     * acceptable in 'name'. */
+    if (strchr(name, '/') || strchr(name, '\\')) {
+        VLOG_ERR("\"%s\" is not a valid name for a vhost-user port. "
+                 "A valid name must not include '/' or '\\'",
+                 name);
+        return EINVAL;
+    }
+
     ovs_mutex_lock(&dpdk_mutex);
     /* Take the name of the vhost-user port and append it to the location where
      * the socket is to be created, then register the socket.
      */
     snprintf(netdev->vhost_id, sizeof(netdev->vhost_id), "%s/%s",
-            vhost_sock_dir, netdev_->name);
+             vhost_sock_dir, name);
+
     err = rte_vhost_driver_register(netdev->vhost_id);
     if (err) {
         VLOG_ERR("vhost-user socket device setup failure for socket %s\n",
                  netdev->vhost_id);
     } else {
         fatal_signal_add_file_to_unlink(netdev->vhost_id);
+        VLOG_INFO("Socket %s created for vhost-user port %s\n",
+                  netdev->vhost_id, name);
+        err = vhost_construct_helper(netdev_);
     }
 
-    VLOG_INFO("Socket %s created for vhost-user port %s\n", netdev->vhost_id, netdev_->name);
-    err = vhost_construct_helper(netdev_);
     ovs_mutex_unlock(&dpdk_mutex);
     return err;
 }
@@ -768,20 +797,35 @@ netdev_dpdk_dealloc(struct netdev *netdev_)
 }
 
 static int
-netdev_dpdk_get_config(const struct netdev *netdev_, struct smap *args)
+netdev_dpdk_get_config(const struct netdev *netdev, struct smap *args)
 {
-    struct netdev_dpdk *dev = netdev_dpdk_cast(netdev_);
+    struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
 
     ovs_mutex_lock(&dev->mutex);
 
-    smap_add_format(args, "configured_rx_queues", "%d", netdev_->n_rxq);
-    smap_add_format(args, "requested_tx_queues", "%d", netdev_->n_txq);
+    smap_add_format(args, "requested_rx_queues", "%d", netdev->requested_n_rxq);
+    smap_add_format(args, "configured_rx_queues", "%d", netdev->n_rxq);
+    smap_add_format(args, "requested_tx_queues", "%d", netdev->n_txq);
     smap_add_format(args, "configured_tx_queues", "%d", dev->real_n_txq);
     ovs_mutex_unlock(&dev->mutex);
 
     return 0;
 }
 
+static int
+netdev_dpdk_set_config(struct netdev *netdev, const struct smap *args)
+{
+    struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
+
+    ovs_mutex_lock(&dev->mutex);
+    netdev->requested_n_rxq = MAX(smap_get_int(args, "n_rxq",
+                                               netdev->requested_n_rxq), 1);
+    netdev_change_seq_changed(netdev);
+    ovs_mutex_unlock(&dev->mutex);
+
+    return 0;
+}
+
 static int
 netdev_dpdk_get_numa_id(const struct netdev *netdev_)
 {
@@ -872,10 +916,8 @@ netdev_dpdk_vhost_set_multiq(struct netdev *netdev_, unsigned int n_txq,
     ovs_mutex_lock(&dpdk_mutex);
     ovs_mutex_lock(&netdev->mutex);
 
-    rte_free(netdev->tx_q);
     netdev->up.n_txq = n_txq;
     netdev->up.n_rxq = n_rxq;
-    netdev_dpdk_alloc_txq(netdev, netdev->up.n_txq);
 
     ovs_mutex_unlock(&netdev->mutex);
     ovs_mutex_unlock(&dpdk_mutex);
@@ -1100,17 +1142,16 @@ __netdev_dpdk_vhost_send(struct netdev *netdev, int qid,
     unsigned int total_pkts = cnt;
     uint64_t start = 0;
 
-    if (OVS_UNLIKELY(!is_vhost_running(virtio_dev))) {
+    qid = vhost_dev->tx_q[qid % vhost_dev->real_n_txq].map;
+
+    if (OVS_UNLIKELY(!is_vhost_running(virtio_dev) || qid == -1)) {
         rte_spinlock_lock(&vhost_dev->stats_lock);
         vhost_dev->stats.tx_dropped+= cnt;
         rte_spinlock_unlock(&vhost_dev->stats_lock);
         goto out;
     }
 
-    if (vhost_dev->txq_needs_locking) {
-        qid = qid % vhost_dev->real_n_txq;
-        rte_spinlock_lock(&vhost_dev->tx_q[qid].tx_lock);
-    }
+    rte_spinlock_lock(&vhost_dev->tx_q[qid].tx_lock);
 
     do {
         int vhost_qid = qid * VIRTIO_QNUM + VIRTIO_RXQ;
@@ -1148,9 +1189,7 @@ __netdev_dpdk_vhost_send(struct netdev *netdev, int qid,
         }
     } while (cnt);
 
-    if (vhost_dev->txq_needs_locking) {
-        rte_spinlock_unlock(&vhost_dev->tx_q[qid].tx_lock);
-    }
+    rte_spinlock_unlock(&vhost_dev->tx_q[qid].tx_lock);
 
     rte_spinlock_lock(&vhost_dev->stats_lock);
     netdev_dpdk_vhost_update_tx_counters(&vhost_dev->stats, pkts, total_pkts,
@@ -1810,9 +1849,50 @@ set_irq_status(struct virtio_net *dev)
     }
 }
 
+/*
+ * Fixes mapping for vhost-user tx queues. Must be called after each
+ * enabling/disabling of queues and real_n_txq modifications.
+ */
+static void
+netdev_dpdk_remap_txqs(struct netdev_dpdk *netdev)
+    OVS_REQUIRES(netdev->mutex)
+{
+    int *enabled_queues, n_enabled = 0;
+    int i, k, total_txqs = netdev->real_n_txq;
+
+    enabled_queues = dpdk_rte_mzalloc(total_txqs * sizeof *enabled_queues);
+
+    for (i = 0; i < total_txqs; i++) {
+        /* Enabled queues always mapped to themselves. */
+        if (netdev->tx_q[i].map == i) {
+            enabled_queues[n_enabled++] = i;
+        }
+    }
+
+    if (n_enabled == 0 && total_txqs != 0) {
+        enabled_queues[0] = -1;
+        n_enabled = 1;
+    }
+
+    k = 0;
+    for (i = 0; i < total_txqs; i++) {
+        if (netdev->tx_q[i].map != i) {
+            netdev->tx_q[i].map = enabled_queues[k];
+            k = (k + 1) % n_enabled;
+        }
+    }
+
+    VLOG_DBG("TX queue mapping for %s\n", netdev->vhost_id);
+    for (i = 0; i < total_txqs; i++) {
+        VLOG_DBG("%2d --> %2d", i, netdev->tx_q[i].map);
+    }
+
+    rte_free(enabled_queues);
+}
 
 static int
 netdev_dpdk_vhost_set_queues(struct netdev_dpdk *netdev, struct virtio_net *dev)
+    OVS_REQUIRES(netdev->mutex)
 {
     uint32_t qp_num;
 
@@ -1826,11 +1906,9 @@ netdev_dpdk_vhost_set_queues(struct netdev_dpdk *netdev, struct virtio_net *dev)
 
     netdev->real_n_rxq = qp_num;
     netdev->real_n_txq = qp_num;
-    if (netdev->up.n_txq > netdev->real_n_txq) {
-        netdev->txq_needs_locking = true;
-    } else {
-        netdev->txq_needs_locking = false;
-    }
+    netdev->txq_needs_locking = true;
+
+    netdev_dpdk_remap_txqs(netdev);
 
     return 0;
 }
@@ -1887,6 +1965,7 @@ static void
 destroy_device(volatile struct virtio_net *dev)
 {
     struct netdev_dpdk *vhost_dev;
+    bool exists = false;
 
     ovs_mutex_lock(&dpdk_mutex);
     LIST_FOR_EACH (vhost_dev, list_node, &dpdk_list) {
@@ -1895,24 +1974,73 @@ destroy_device(volatile struct virtio_net *dev)
             ovs_mutex_lock(&vhost_dev->mutex);
             dev->flags &= ~VIRTIO_DEV_RUNNING;
             ovsrcu_set(&vhost_dev->virtio_dev, NULL);
+            exists = true;
             ovs_mutex_unlock(&vhost_dev->mutex);
+            break;
+        }
+    }
 
-            /*
-             * Wait for other threads to quiesce before
-             * setting the virtio_dev to NULL.
-             */
-            ovsrcu_synchronize();
-            /*
-             * As call to ovsrcu_synchronize() will end the quiescent state,
-             * put thread back into quiescent state before returning.
-             */
-            ovsrcu_quiesce_start();
+    ovs_mutex_unlock(&dpdk_mutex);
+
+    if (exists == true) {
+        /*
+         * Wait for other threads to quiesce after setting the 'virtio_dev'
+         * to NULL, before returning.
+         */
+        ovsrcu_synchronize();
+        /*
+         * As call to ovsrcu_synchronize() will end the quiescent state,
+         * put thread back into quiescent state before returning.
+         */
+        ovsrcu_quiesce_start();
+        VLOG_INFO("vHost Device '%s' %"PRIu64" has been removed", dev->ifname,
+                  dev->device_fh);
+    } else {
+        VLOG_INFO("vHost Device '%s' %"PRIu64" not found", dev->ifname,
+                  dev->device_fh);
+    }
+
+}
+
+static int
+vring_state_changed(struct virtio_net *dev, uint16_t queue_id, int enable)
+{
+    struct netdev_dpdk *vhost_dev;
+    bool exists = false;
+    int qid = queue_id / VIRTIO_QNUM;
+
+    if (queue_id % VIRTIO_QNUM == VIRTIO_TXQ) {
+        return 0;
+    }
+
+    ovs_mutex_lock(&dpdk_mutex);
+    LIST_FOR_EACH (vhost_dev, list_node, &dpdk_list) {
+        if (strncmp(dev->ifname, vhost_dev->vhost_id, IF_NAME_SZ) == 0) {
+            ovs_mutex_lock(&vhost_dev->mutex);
+            if (enable) {
+                vhost_dev->tx_q[qid].map = qid;
+            } else {
+                vhost_dev->tx_q[qid].map = -1;
+            }
+            netdev_dpdk_remap_txqs(vhost_dev);
+            exists = true;
+            ovs_mutex_unlock(&vhost_dev->mutex);
+            break;
         }
     }
     ovs_mutex_unlock(&dpdk_mutex);
 
-    VLOG_INFO("vHost Device '%s' %"PRIu64" has been removed", dev->ifname,
-              dev->device_fh);
+    if (exists) {
+        VLOG_INFO("State of queue %d ( tx_qid %d ) of vhost device '%s' %"
+                  PRIu64" changed to \'%s\'", queue_id, qid, dev->ifname,
+                  dev->device_fh, (enable == 1) ? "enabled" : "disabled");
+    } else {
+        VLOG_INFO("vHost Device '%s' %"PRIu64" not found", dev->ifname,
+                  dev->device_fh);
+        return -1;
+    }
+
+    return 0;
 }
 
 struct virtio_net *
@@ -1929,6 +2057,7 @@ static const struct virtio_net_device_ops virtio_net_device_ops =
 {
     .new_device =  new_device,
     .destroy_device = destroy_device,
+    .vring_state_changed = vring_state_changed
 };
 
 static void *
@@ -1994,7 +2123,7 @@ dpdk_ring_create(const char dev_name[], unsigned int port_no,
                  unsigned int *eth_port_id)
 {
     struct dpdk_ring *ivshmem;
-    char ring_name[10];
+    char ring_name[RTE_RING_NAMESIZE];
     int err;
 
     ivshmem = dpdk_rte_mzalloc(sizeof *ivshmem);
@@ -2003,7 +2132,7 @@ dpdk_ring_create(const char dev_name[], unsigned int port_no,
     }
 
     /* XXX: Add support for multiquque ring. */
-    err = snprintf(ring_name, 10, "%s_tx", dev_name);
+    err = snprintf(ring_name, sizeof(ring_name), "%s_tx", dev_name);
     if (err < 0) {
         return -err;
     }
@@ -2016,7 +2145,7 @@ dpdk_ring_create(const char dev_name[], unsigned int port_no,
         return ENOMEM;
     }
 
-    err = snprintf(ring_name, 10, "%s_rx", dev_name);
+    err = snprintf(ring_name, sizeof(ring_name), "%s_rx", dev_name);
     if (err < 0) {
         return -err;
     }
@@ -2126,7 +2255,7 @@ unlock_dpdk:
     DESTRUCT,                                                 \
     netdev_dpdk_dealloc,                                      \
     netdev_dpdk_get_config,                                   \
-    NULL,                       /* netdev_dpdk_set_config */  \
+    netdev_dpdk_set_config,                                   \
     NULL,                       /* get_tunnel_config */       \
     NULL,                       /* build header */            \
     NULL,                       /* push header */             \