On 03/10/2018 11:01 AM, Zhihong Wang wrote:
This patch adapts vhost lib for selective datapath by calling device ops
at the corresponding stage.
Signed-off-by: Zhihong Wang <zhihong.w...@intel.com>
---
Changes in v4:
1. Remove the "engine" concept in the lib.
---
Changes in v2:
1. Ensure negotiated capabilities are supported in vhost-user lib.
2. Configure the data path at the right time.
lib/librte_vhost/rte_vhost.h | 27 ++++++++++
lib/librte_vhost/rte_vhost_version.map | 2 +
lib/librte_vhost/socket.c | 94 ++++++++++++++++++++++++++++++++--
lib/librte_vhost/vhost.c | 3 ++
lib/librte_vhost/vhost.h | 2 +
lib/librte_vhost/vhost_user.c | 54 +++++++++++++++++--
6 files changed, 172 insertions(+), 10 deletions(-)
diff --git a/lib/librte_vhost/rte_vhost.h b/lib/librte_vhost/rte_vhost.h
index d50f4c67d..3c3334d3e 100644
--- a/lib/librte_vhost/rte_vhost.h
+++ b/lib/librte_vhost/rte_vhost.h
@@ -279,6 +279,33 @@ int rte_vhost_driver_disable_features(const char *path,
uint64_t features);
int rte_vhost_driver_get_features(const char *path, uint64_t *features);
/**
+ * Get the protocol feature bits before feature negotiation.
+ *
+ * @param path
+ * The vhost-user socket file path
+ * @param protocol_features
+ * A pointer to store the queried protocol feature bits
+ * @return
+ * 0 on success, -1 on failure
+ */
+int __rte_experimental
+rte_vhost_driver_get_protocol_features(const char *path,
+ uint64_t *protocol_features);
+
+/**
+ * Get the queue number bits before feature negotiation.
+ *
+ * @param path
+ * The vhost-user socket file path
+ * @param queue_num
+ * A pointer to store the queried queue number bits
+ * @return
+ * 0 on success, -1 on failure
+ */
+int __rte_experimental
+rte_vhost_driver_get_queue_num(const char *path, uint32_t *queue_num);
+
+/**
* Get the feature bits after negotiation
*
* @param vid
diff --git a/lib/librte_vhost/rte_vhost_version.map
b/lib/librte_vhost/rte_vhost_version.map
index 6e2d5364a..812ccd72b 100644
--- a/lib/librte_vhost/rte_vhost_version.map
+++ b/lib/librte_vhost/rte_vhost_version.map
@@ -67,4 +67,6 @@ EXPERIMENTAL {
rte_vhost_driver_set_vdpa_did;
rte_vhost_driver_get_vdpa_did;
rte_vhost_get_vdpa_did;
+ rte_vhost_driver_get_protocol_features;
+ rte_vhost_driver_get_queue_num;
} DPDK_18.02;
diff --git a/lib/librte_vhost/socket.c b/lib/librte_vhost/socket.c
index 3d58da94e..ba7b422a0 100644
--- a/lib/librte_vhost/socket.c
+++ b/lib/librte_vhost/socket.c
@@ -216,6 +216,8 @@ vhost_user_add_connection(int fd, struct vhost_user_socket
*vsocket)
vhost_set_builtin_virtio_net(vid, vsocket->use_builtin_virtio_net);
+ vhost_set_vdpa_did(vid, vsocket->did);
+
if (vsocket->dequeue_zero_copy)
vhost_enable_dequeue_zero_copy(vid);
@@ -648,20 +650,102 @@ int
rte_vhost_driver_get_features(const char *path, uint64_t *features)
{
struct vhost_user_socket *vsocket;
+ uint64_t vdpa_features;
+ int did = -1;
+ int ret = 0;
pthread_mutex_lock(&vhost_user.mutex);
vsocket = find_vhost_user_socket(path);
- if (vsocket)
- *features = vsocket->features;
+ if (vsocket) {
+ did = vsocket->did;
+ if (did < 0 || vdpa_devices[did]->ops->feature_get == NULL)
+ *features = vsocket->features;
+ else if (vdpa_devices[did]->ops->feature_get(did,
+ &vdpa_features) < 0) {
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "failed to get vdpa features "
+ "for socket file %s.\n", path);
+ ret = -1;
+ } else
+ *features = vsocket->features & vdpa_features;
It seems correct but it is not very intuitive.
Also, you have to put braces everywhere if one of the if/else if/else
have some.
What about something like this:
rte_vhost_driver_get_features(const char *path, uint64_t *features)
{
struct vhost_user_socket *vsocket;
uint64_t vdpa_features;
int did = -1;
int ret = 0;
struct rte_vdpa_device *vdpa_dev;
pthread_mutex_lock(&vhost_user.mutex);
vsocket = find_vhost_user_socket(path);
if (!vsocket) {
RTE_LOG(ERR, VHOST_CONFIG,
"socket file %s is not registered yet.\n"
, path);
ret = -1;
goto out_unlock;
}
did = vsocket->did;
vdpa_dev = rte_vdpa_device_get(did);
if (!vdpa_dev || !vdpa->ops->feature_get) {
*features = vsocket->features;
goto out_unlock;
}
if (vdpa_dev->ops->feature_get(did, &vdpa_features) < 0) {
RTE_LOG(ERR, VHOST_CONFIG,
"failed to get vdpa features "
"for socket file %s.\n", path);
ret = -1;
goto out_unlock;
}
*features = vsocket->features & vdpa_features;
out_unlock:
pthread_mutex_unlock(&vhost_user.mutex);
return ret;
}
with in rte_vdpa.h:
static inline struct rte_vdpa_device *
rte_vdpa_device_get(int did)
{
if (did < 0 || did >= MAX_VHOST_DEVICE)
return NULL;
return vdpa_devices[did];
}
Doing this you have 3 checks in 1, so you avoid NULL pointer
de-referencing if did > 0 but invalid.
Same logic applies to functions below:
+ } else {
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "socket file %s is not registered yet.\n", path);
+ ret = -1;
+ }
pthread_mutex_unlock(&vhost_user.mutex);
- if (!vsocket) {
+ return ret;
+}
+
+int
+rte_vhost_driver_get_protocol_features(const char *path,
+ uint64_t *protocol_features)
+{
+ struct vhost_user_socket *vsocket;
+ uint64_t vdpa_protocol_features;
+ int did = -1;
+ int ret = 0;
+
+ pthread_mutex_lock(&vhost_user.mutex);
+ vsocket = find_vhost_user_socket(path);
+ if (vsocket) {
+ did = vsocket->did;
+ if (did < 0 || vdpa_devices[did]->ops->protocol_feature_get
+ == NULL)
+ *protocol_features = VHOST_USER_PROTOCOL_FEATURES;
+ else if (vdpa_devices[did]->ops->protocol_feature_get(did,
+ &vdpa_protocol_features) < 0) {
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "failed to get vdpa protocol features "
+ "for socket file %s.\n", path);
+ ret = -1;
+ } else
+ *protocol_features = VHOST_USER_PROTOCOL_FEATURES
+ & vdpa_protocol_features;
+ } else {
RTE_LOG(ERR, VHOST_CONFIG,
"socket file %s is not registered yet.\n", path);
- return -1;
+ ret = -1;
+ }
+ pthread_mutex_unlock(&vhost_user.mutex);
+
+ return ret;
+}
+
+int
+rte_vhost_driver_get_queue_num(const char *path,
+ uint32_t *queue_num)
+{
+ struct vhost_user_socket *vsocket;
+ uint32_t vdpa_queue_num;
+ int did = -1;
+ int ret = 0;
+
+ pthread_mutex_lock(&vhost_user.mutex);
+ vsocket = find_vhost_user_socket(path);
+ if (vsocket) {
+ did = vsocket->did;
+ if (did < 0 || vdpa_devices[did]->ops->queue_num_get == NULL)
+ *queue_num = VHOST_MAX_QUEUE_PAIRS;
+ else if (vdpa_devices[did]->ops->queue_num_get(
+ did, &vdpa_queue_num) < 0) {
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "failed to get vdpa queue number "
+ "for socket file %s.\n", path);
+ ret = -1;
+ } else if (vdpa_queue_num > VHOST_MAX_QUEUE_PAIRS)
+ *queue_num = VHOST_MAX_QUEUE_PAIRS;
+ else
+ *queue_num = vdpa_queue_num;
} else {
- return 0;
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "socket file %s is not registered yet.\n", path);
+ ret = -1;
}
+ pthread_mutex_unlock(&vhost_user.mutex);
+
+ return ret;
}
/*
diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c
index 1740cc1ab..66b6b492f 100644
--- a/lib/librte_vhost/vhost.c
+++ b/lib/librte_vhost/vhost.c
@@ -296,11 +296,14 @@ void
vhost_destroy_device(int vid)
{
struct virtio_net *dev = get_device(vid);
+ int did = dev->did;
if (dev == NULL)
return;
if (dev->flags & VIRTIO_DEV_RUNNING) {
+ if (did >= 0 && vdpa_devices[did]->ops->dev_close)
+ vdpa_devices[did]->ops->dev_close(dev->vid);
Ditto.
dev->flags &= ~VIRTIO_DEV_RUNNING;
dev->notify_ops->destroy_device(vid);
}
diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h
index 2970c1eab..30e8a0cfe 100644
--- a/lib/librte_vhost/vhost.h
+++ b/lib/librte_vhost/vhost.h
@@ -27,6 +27,8 @@
#define VIRTIO_DEV_READY 2
/* Used to indicate that the built-in vhost net device backend is enabled */
#define VIRTIO_DEV_BUILTIN_VIRTIO_NET 4
+/* Used to indicate that the device has its own data path and configured */
+#define VIRTIO_DEV_VDPA_CONFIGURED 8
/* Backend value set by guest. */
#define VIRTIO_DEV_STOPPED -1
diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
index 5c5361066..5049c8e55 100644
--- a/lib/librte_vhost/vhost_user.c
+++ b/lib/librte_vhost/vhost_user.c
@@ -133,7 +133,11 @@ vhost_user_set_owner(void)
static int
vhost_user_reset_owner(struct virtio_net *dev)
{
+ int did = dev->did;
+
if (dev->flags & VIRTIO_DEV_RUNNING) {
+ if (did >= 0 && vdpa_devices[did]->ops->dev_close)
+ vdpa_devices[did]->ops->dev_close(dev->vid);
Ditto.
dev->flags &= ~VIRTIO_DEV_RUNNING;
dev->notify_ops->destroy_device(dev->vid);
}
@@ -156,12 +160,25 @@ vhost_user_get_features(struct virtio_net *dev)
}
/*
+ * The queue number that we support are requested.
+ */
+static uint32_t
+vhost_user_get_queue_num(struct virtio_net *dev)
+{
+ uint32_t queue_num = 0;
+
+ rte_vhost_driver_get_queue_num(dev->ifname, &queue_num);
+ return (uint64_t)queue_num;
+}
+
+/*
* We receive the negotiated features supported by us and the virtio device.
*/
static int
vhost_user_set_features(struct virtio_net *dev, uint64_t features)
{
uint64_t vhost_features = 0;
+ int did = dev->did;
rte_vhost_driver_get_features(dev->ifname, &vhost_features);
if (features & ~vhost_features) {
@@ -191,6 +208,9 @@ vhost_user_set_features(struct virtio_net *dev, uint64_t
features)
dev->notify_ops->features_changed(dev->vid, features);
}
+ if (did >= 0 && vdpa_devices[did]->ops->feature_set)
+ vdpa_devices[did]->ops->feature_set(dev->vid);
+
dev->features = features;
if (dev->features &
((1 << VIRTIO_NET_F_MRG_RXBUF) | (1ULL << VIRTIO_F_VERSION_1)))
{
@@ -933,14 +953,18 @@ vhost_user_get_vring_base(struct virtio_net *dev,
VhostUserMsg *msg)
{
struct vhost_virtqueue *vq = dev->virtqueue[msg->payload.state.index];
+ int did = dev->did;
/* We have to stop the queue (virtio) if it is running. */
if (dev->flags & VIRTIO_DEV_RUNNING) {
+ if (did >= 0 && vdpa_devices[did]->ops->dev_close)
+ vdpa_devices[did]->ops->dev_close(dev->vid);
dev->flags &= ~VIRTIO_DEV_RUNNING;
dev->notify_ops->destroy_device(dev->vid);
}
dev->flags &= ~VIRTIO_DEV_READY;
+ dev->flags &= ~VIRTIO_DEV_VDPA_CONFIGURED;
/* Here we are safe to get the last avail index */
msg->payload.state.num = vq->last_avail_idx;
@@ -983,16 +1007,22 @@ vhost_user_set_vring_enable(struct virtio_net *dev,
VhostUserMsg *msg)
{
int enable = (int)msg->payload.state.num;
+ int index = (int)msg->payload.state.index;
+ int did = dev->did;
RTE_LOG(INFO, VHOST_CONFIG,
"set queue enable: %d to qp idx: %d\n",
- enable, msg->payload.state.index);
+ enable, index);
+
+ if (did >= 0 && vdpa_devices[did]->ops->vring_state_set)
+ vdpa_devices[did]->ops->vring_state_set(dev->vid,
+ index, enable);
if (dev->notify_ops->vring_state_changed)
dev->notify_ops->vring_state_changed(dev->vid,
- msg->payload.state.index, enable);
+ index, enable);
- dev->virtqueue[msg->payload.state.index]->enabled = enable;
+ dev->virtqueue[index]->enabled = enable;
return 0;
}
@@ -1001,9 +1031,10 @@ static void
vhost_user_get_protocol_features(struct virtio_net *dev,
struct VhostUserMsg *msg)
{
- uint64_t features, protocol_features = VHOST_USER_PROTOCOL_FEATURES;
+ uint64_t features, protocol_features;
rte_vhost_driver_get_features(dev->ifname, &features);
+ rte_vhost_driver_get_protocol_features(dev->ifname, &protocol_features);
/*
* REPLY_ACK protocol feature is only mandatory for now
@@ -1099,6 +1130,7 @@ static int
vhost_user_send_rarp(struct virtio_net *dev, struct VhostUserMsg *msg)
{
uint8_t *mac = (uint8_t *)&msg->payload.u64;
+ int did = dev->did;
RTE_LOG(DEBUG, VHOST_CONFIG,
":: mac: %02x:%02x:%02x:%02x:%02x:%02x\n",
@@ -1114,6 +1146,8 @@ vhost_user_send_rarp(struct virtio_net *dev, struct
VhostUserMsg *msg)
*/
rte_smp_wmb();
rte_atomic16_set(&dev->broadcast_rarp, 1);
+ if (did >= 0 && vdpa_devices[did]->ops->migration_done)
+ vdpa_devices[did]->ops->migration_done(dev->vid);
return 0;
}
@@ -1375,6 +1409,7 @@ vhost_user_msg_handler(int vid, int fd)
{
struct virtio_net *dev;
struct VhostUserMsg msg;
+ int did;
int ret;
int unlock_required = 0;
@@ -1527,7 +1562,7 @@ vhost_user_msg_handler(int vid, int fd)
break;
case VHOST_USER_GET_QUEUE_NUM:
- msg.payload.u64 = VHOST_MAX_QUEUE_PAIRS;
+ msg.payload.u64 = (uint64_t)vhost_user_get_queue_num(dev);
msg.size = sizeof(msg.payload.u64);
send_vhost_reply(fd, &msg);
break;
@@ -1580,6 +1615,15 @@ vhost_user_msg_handler(int vid, int fd)
}
}
+ did = dev->did;
+ if (did >= 0 && virtio_is_ready(dev) &&
+ !(dev->flags & VIRTIO_DEV_VDPA_CONFIGURED) &&
+ msg.request.master == VHOST_USER_SET_VRING_ENABLE) {
+ if (vdpa_devices[did]->ops->dev_conf)
+ vdpa_devices[did]->ops->dev_conf(vid);
+ dev->flags |= VIRTIO_DEV_VDPA_CONFIGURED;
+ }
+
return 0;
}
Other than the code refactoring I propose, the patch looks good to me.
Thanks!
Maxime
