vDPA device driver will need callfd to setup device VFIO interrupt.
Looking at QEMU vhost message sequence, the real callfd comes after
the other vring messages and just before SET_VRING_ENABLE.
This patch is to make sure we configure device only when all the
guest information is in hand.
Note: virtio-user's vhost message sequence is slightly different from
QEMU, it has no SET_VRING_ENABLE in the ending. So this patch is not
applicable to virtio-user. We could align virtio-user's message
sequence to QEMU in future to address this issue.
Signed-off-by: Xiao Wang <xiao.w.w...@intel.com>
---
lib/librte_vhost/vhost_user.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
index 51e6443..65d52bd 100644
--- a/lib/librte_vhost/vhost_user.c
+++ b/lib/librte_vhost/vhost_user.c
@@ -1404,7 +1404,9 @@
send_vhost_reply(fd, &msg);
}
- if (!(dev->flags & VIRTIO_DEV_RUNNING) && virtio_is_ready(dev)) {
+ if (msg.request.master == VHOST_USER_SET_VRING_ENABLE &&
+ !(dev->flags & VIRTIO_DEV_RUNNING) &&
+ virtio_is_ready(dev)) {
dev->flags |= VIRTIO_DEV_READY;
if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
--
1.8.3.1
