On 03/06/2018 11:43 AM, Tiwei Bie wrote:
This commit adds the VFIO based accelerator support to
vhost. A new API is provided to support asking QEMU to
do further setup to allow notifications and interrupts
being delivered directly between the driver in guest
and the vDPA device in host.
Signed-off-by: Tiwei Bie <tiwei....@intel.com>
---
lib/librte_vhost/rte_vhost.h | 28 ++++++
lib/librte_vhost/rte_vhost_version.map | 1 +
lib/librte_vhost/vhost_user.c | 166 +++++++++++++++++++++++++++++++++
lib/librte_vhost/vhost_user.h | 9 ++
4 files changed, 204 insertions(+)
diff --git a/lib/librte_vhost/rte_vhost.h b/lib/librte_vhost/rte_vhost.h
index d5589c543..68842e908 100644
--- a/lib/librte_vhost/rte_vhost.h
+++ b/lib/librte_vhost/rte_vhost.h
@@ -35,6 +35,7 @@ extern "C" {
#define RTE_VHOST_USER_PROTOCOL_F_REPLY_ACK 3
#define RTE_VHOST_USER_PROTOCOL_F_NET_MTU 4
#define RTE_VHOST_USER_PROTOCOL_F_SLAVE_REQ 5
+#define RTE_VHOST_USER_PROTOCOL_F_VFIO 8
#define RTE_VHOST_USER_F_PROTOCOL_FEATURES 30
/**
@@ -591,6 +592,33 @@ rte_vhost_get_vdpa_eid(int vid);
int __rte_experimental
rte_vhost_get_vdpa_did(int vid);
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice
+ *
+ * Enable or disable the VFIO based accelerator for vhost-user.
+ *
+ * This function is to ask QEMU to do further setup to better
+ * support the vDPA device at vhost user backend. With this
+ * setup, the notifications and interrupts will be delivered
+ * directly between the driver in guest and the vDPA device
+ * in host if platform supports e.g. EPT and Posted interrupt.
+ * It's nice to have, and not mandatory.
+ *
+ * @param vid
+ * vhost device ID
+ * @param int
+ * Enable or disable
+ *
+ * @return
+ * 0: success
+ * -ENODEV: no such vhost device
+ * -ENOTSUP: device does not support VFIO based accelerator feature
+ * -EINVAL: there is no accelerator assigned to this vhost device
+ * -EFAULT: failed to talk with QEMU
+ */
+int rte_vhost_vfio_accelerator_ctrl(int vid, int enable);
+
#ifdef __cplusplus
}
#endif
diff --git a/lib/librte_vhost/rte_vhost_version.map
b/lib/librte_vhost/rte_vhost_version.map
index 36257e51b..ca970170f 100644
--- a/lib/librte_vhost/rte_vhost_version.map
+++ b/lib/librte_vhost/rte_vhost_version.map
@@ -72,6 +72,7 @@ EXPERIMENTAL {
rte_vhost_set_vring_base;
rte_vhost_get_vdpa_eid;
rte_vhost_get_vdpa_did;
+ rte_vhost_vfio_accelerator_ctrl;
rte_vdpa_register_engine;
rte_vdpa_unregister_engine;
rte_vdpa_find_engine_id;
diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
index e3a1dfbfb..a65598d80 100644
--- a/lib/librte_vhost/vhost_user.c
+++ b/lib/librte_vhost/vhost_user.c
@@ -35,6 +35,7 @@
#include <rte_common.h>
#include <rte_malloc.h>
#include <rte_log.h>
+#include <rte_vhost.h>
#include "iotlb.h"
#include "vhost.h"
@@ -1628,6 +1629,27 @@ vhost_user_msg_handler(int vid, int fd)
return 0;
}
+static int process_slave_message_reply(struct virtio_net *dev,
+ const VhostUserMsg *msg)
+{
+ VhostUserMsg msg_reply;
+
+ if ((msg->flags & VHOST_USER_NEED_REPLY) == 0)
+ return 0;
+
+ if (read_vhost_message(dev->slave_req_fd, &msg_reply) < 0)
+ return -1;
+
+ if (msg_reply.request.slave != msg->request.slave) {
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "received unexpected msg type (%u), expected %u\n",
+ msg_reply.request.slave, msg->request.slave);
+ return -1;
+ }
+
+ return msg_reply.payload.u64;
+}
+
int
vhost_user_iotlb_miss(struct virtio_net *dev, uint64_t iova, uint8_t perm)
{
@@ -1653,3 +1675,147 @@ vhost_user_iotlb_miss(struct virtio_net *dev, uint64_t
iova, uint8_t perm)
return 0;
}
+
+static int vhost_user_slave_set_vring_file(struct virtio_net *dev,
+ uint32_t request,
+ struct vhost_vring_file *file)
Why passing the request as an argument?
It seems to be called only with the same request ID.
+{
+ int *fdp = NULL;
+ size_t fd_num = 0;
+ int ret;
+ struct VhostUserMsg msg = {
+ .request.slave = request,
+ .flags = VHOST_USER_VERSION | VHOST_USER_NEED_REPLY,
+ .payload.u64 = file->index & VHOST_USER_VRING_IDX_MASK,
+ .size = sizeof(msg.payload.u64),
+ };
+
+ if (file->fd < 0)
+ msg.payload.u64 |= VHOST_USER_VRING_NOFD_MASK;
+ else {
+ fdp = &file->fd;
+ fd_num = 1;
+ }
+
+ ret = send_vhost_message(dev->slave_req_fd, &msg, fdp, fd_num);
+ if (ret < 0) {
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "Failed to send slave message %u (%d)\n",
+ request, ret);
+ return ret;
+ }
+
+ return process_slave_message_reply(dev, &msg);
Maybe not needed right now, but we'll need a lock to avoid concurrent
requests sending and waiting for reply.
Thanks,
Maxime
