net/virtio: support vhost-user protocol features
authorMaxime Coquelin <maxime.coquelin@redhat.com>
Fri, 3 Jul 2020 15:57:26 +0000 (17:57 +0200)
committerFerruh Yigit <ferruh.yigit@intel.com>
Sat, 11 Jul 2020 04:18:52 +0000 (06:18 +0200)
This patch adds support for Vhost-user protocol features.
It is required to support protocol features that were not in
initial Vhost-user specification, such as reply-ack, MTU...

Also, this patch prevents Virtio multiqueue feature negotiation
if the slave does not support MQ protocol feature as stated
in Vhost-user specification:
"The multiple queues feature is supported only when the protocol
feature ``VHOST_USER_PROTOCOL_F_MQ`` (bit 0) is set."

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
drivers/net/virtio/virtio_user/vhost.h
drivers/net/virtio/virtio_user/vhost_user.c
drivers/net/virtio/virtio_user/virtio_user_dev.c
drivers/net/virtio/virtio_user/virtio_user_dev.h
drivers/net/virtio/virtio_user_ethdev.c

index 1e784e5..9ace1a9 100644 (file)
@@ -44,6 +44,15 @@ struct vhost_vring_addr {
        uint64_t log_guest_addr;
 };
 
+#ifndef VHOST_USER_F_PROTOCOL_FEATURES
+#define VHOST_USER_F_PROTOCOL_FEATURES 30
+#endif
+
+/** Protocol features. */
+#ifndef VHOST_USER_PROTOCOL_F_MQ
+#define VHOST_USER_PROTOCOL_F_MQ       0
+#endif
+
 enum vhost_user_request {
        VHOST_USER_NONE = 0,
        VHOST_USER_GET_FEATURES = 1,
index 74b82e5..e330675 100644 (file)
@@ -241,6 +241,8 @@ const char * const vhost_msg_strings[] = {
        [VHOST_USER_SET_VRING_KICK] = "VHOST_SET_VRING_KICK",
        [VHOST_USER_SET_MEM_TABLE] = "VHOST_SET_MEM_TABLE",
        [VHOST_USER_SET_VRING_ENABLE] = "VHOST_SET_VRING_ENABLE",
+       [VHOST_USER_GET_PROTOCOL_FEATURES] = "VHOST_USER_GET_PROTOCOL_FEATURES",
+       [VHOST_USER_SET_PROTOCOL_FEATURES] = "VHOST_USER_SET_PROTOCOL_FEATURES",
 };
 
 static int
@@ -269,10 +271,12 @@ vhost_user_sock(struct virtio_user_dev *dev,
 
        switch (req) {
        case VHOST_USER_GET_FEATURES:
+       case VHOST_USER_GET_PROTOCOL_FEATURES:
                need_reply = 1;
                break;
 
        case VHOST_USER_SET_FEATURES:
+       case VHOST_USER_SET_PROTOCOL_FEATURES:
        case VHOST_USER_SET_LOG_BASE:
                msg.payload.u64 = *((__u64 *)arg);
                msg.size = sizeof(m.payload.u64);
@@ -351,6 +355,7 @@ vhost_user_sock(struct virtio_user_dev *dev,
 
                switch (req) {
                case VHOST_USER_GET_FEATURES:
+               case VHOST_USER_GET_PROTOCOL_FEATURES:
                        if (msg.size != sizeof(m.payload.u64)) {
                                PMD_DRV_LOG(ERR, "Received bad msg size");
                                return -1;
index 7fb135f..105536d 100644 (file)
@@ -151,8 +151,10 @@ virtio_user_start_device(struct virtio_user_dev *dev)
        if (virtio_user_queue_setup(dev, virtio_user_create_queue) < 0)
                goto error;
 
-       /* Step 1: set features */
+       /* Step 1: negotiate protocol features & set features */
        features = dev->features;
+
+
        /* Strip VIRTIO_NET_F_MAC, as MAC address is handled in vdev init */
        features &= ~(1ull << VIRTIO_NET_F_MAC);
        /* Strip VIRTIO_NET_F_CTRL_VQ, as devices do not really need to know */
@@ -417,13 +419,19 @@ virtio_user_dev_setup(struct virtio_user_dev *dev)
         1ULL << VIRTIO_NET_F_GUEST_TSO6        |       \
         1ULL << VIRTIO_F_IN_ORDER              |       \
         1ULL << VIRTIO_F_VERSION_1             |       \
-        1ULL << VIRTIO_F_RING_PACKED)
+        1ULL << VIRTIO_F_RING_PACKED           |       \
+        1ULL << VHOST_USER_F_PROTOCOL_FEATURES)
+
+#define VIRTIO_USER_SUPPORTED_PROTOCOL_FEATURES                \
+       (1ULL << VHOST_USER_PROTOCOL_F_MQ)
 
 int
 virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues,
                     int cq, int queue_size, const char *mac, char **ifname,
                     int server, int mrg_rxbuf, int in_order, int packed_vq)
 {
+       uint64_t protocol_features = 0;
+
        pthread_mutex_init(&dev->mutex, NULL);
        strlcpy(dev->path, path, PATH_MAX);
        dev->started = 0;
@@ -434,6 +442,7 @@ virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues,
        dev->mac_specified = 0;
        dev->frontend_features = 0;
        dev->unsupported_features = ~VIRTIO_USER_SUPPORTED_FEATURES;
+       dev->protocol_features = VIRTIO_USER_SUPPORTED_PROTOCOL_FEATURES;
        parse_mac(dev, mac);
 
        if (*ifname) {
@@ -446,6 +455,10 @@ virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues,
                return -1;
        }
 
+       if (!is_vhost_user_by_type(dev->path))
+               dev->unsupported_features |=
+                       (1ULL << VHOST_USER_F_PROTOCOL_FEATURES);
+
        if (!dev->is_server) {
                if (dev->ops->send_request(dev, VHOST_USER_SET_OWNER,
                                           NULL) < 0) {
@@ -460,6 +473,27 @@ virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues,
                                     strerror(errno));
                        return -1;
                }
+
+
+               if (dev->device_features &
+                               (1ULL << VHOST_USER_F_PROTOCOL_FEATURES)) {
+                       if (dev->ops->send_request(dev,
+                                       VHOST_USER_GET_PROTOCOL_FEATURES,
+                                       &protocol_features))
+                               return -1;
+
+                       dev->protocol_features &= protocol_features;
+
+                       if (dev->ops->send_request(dev,
+                                       VHOST_USER_SET_PROTOCOL_FEATURES,
+                                       &dev->protocol_features))
+                               return -1;
+
+                       if (!(dev->protocol_features &
+                                       (1ULL << VHOST_USER_PROTOCOL_F_MQ)))
+                               dev->unsupported_features |=
+                                       (1ull << VIRTIO_NET_F_MQ);
+               }
        } else {
                /* We just pretend vhost-user can support all these features.
                 * Note that this could be problematic that if some feature is
@@ -469,6 +503,8 @@ virtio_user_dev_init(struct virtio_user_dev *dev, char *path, int queues,
                dev->device_features = VIRTIO_USER_SUPPORTED_FEATURES;
        }
 
+
+
        if (!mrg_rxbuf)
                dev->unsupported_features |= (1ull << VIRTIO_NET_F_MRG_RXBUF);
 
index 3b6b606..56e638f 100644 (file)
@@ -40,6 +40,9 @@ struct virtio_user_dev {
        uint64_t        device_features; /* supported features by device */
        uint64_t        frontend_features; /* enabled frontend features */
        uint64_t        unsupported_features; /* unsupported features mask */
+       uint64_t        protocol_features; /* negotiated protocol features
+                                           * (Vhost-user only)
+                                           */
        uint8_t         status;
        uint16_t        port_id;
        uint8_t         mac_addr[RTE_ETHER_ADDR_LEN];
index 798f191..e51425c 100644 (file)
@@ -68,6 +68,7 @@ virtio_user_server_reconnect(struct virtio_user_dev *dev)
        int connectfd;
        struct rte_eth_dev *eth_dev = &rte_eth_devices[dev->port_id];
        struct virtio_hw *hw = eth_dev->data->dev_private;
+       uint64_t protocol_features;
 
        connectfd = accept(dev->listenfd, NULL, NULL);
        if (connectfd < 0)
@@ -81,6 +82,25 @@ virtio_user_server_reconnect(struct virtio_user_dev *dev)
                return -1;
        }
 
+       if (dev->device_features &
+                       (1ULL << VHOST_USER_F_PROTOCOL_FEATURES)) {
+               if (dev->ops->send_request(dev,
+                                       VHOST_USER_GET_PROTOCOL_FEATURES,
+                                       &protocol_features))
+                       return -1;
+
+               dev->protocol_features &= protocol_features;
+
+               if (dev->ops->send_request(dev,
+                                       VHOST_USER_SET_PROTOCOL_FEATURES,
+                                       &dev->protocol_features))
+                       return -1;
+
+               if (!(dev->protocol_features &
+                               (1ULL << VHOST_USER_PROTOCOL_F_MQ)))
+                       dev->unsupported_features |= (1ull << VIRTIO_NET_F_MQ);
+       }
+
        dev->device_features |= dev->frontend_features;
 
        /* umask vhost-user unsupported features */