On Fri, Jan 10, 2020 at 08:22:03PM +0800, Yang Zhong wrote: > As the vhost-user-scsi did in f04724, if the vhost-user-blk backend > supports the VHOST_USER_F_RESET_DEVICE protocol feature, then the > device can be reset when requested. > > If this feature is not supported, this reset will directly return. > > Signed-off-by: Yang Zhong > --- > hw/block/vhost-user-blk.c | 19 +++++++++++++++++++ > 1 file changed, 19 insertions(+) > > diff --git a/hw/block/vhost-user-blk.c b/hw/block/vhost-user-blk.c > index 63da9bb619..16ddc9b70c 100644 > --- a/hw/block/vhost-user-blk.c > +++ b/hw/block/vhost-user-blk.c > @@ -50,6 +50,10 @@ static const int user_feature_bits[] = { > VHOST_INVALID_FEATURE_BIT > }; > > +enum VhostUserProtocolFeature { > + VHOST_USER_PROTOCOL_F_RESET_DEVICE = 13, > +}; vhost-user protocol constants should be defined in hw/virtio/vhost-user.h and not duplicated in device implementations. > + > static void vhost_user_blk_update_config(VirtIODevice *vdev, uint8_t *config) > { > VHostUserBlk *s = VHOST_USER_BLK(vdev); > @@ -290,8 +294,23 @@ static void vhost_user_blk_handle_output(VirtIODevice *vdev, VirtQueue *vq) > static void vhost_user_blk_reset(VirtIODevice *vdev) > { > VHostUserBlk *s = VHOST_USER_BLK(vdev); > + struct vhost_dev *dev = &s->dev; > > vhost_dev_free_inflight(s->inflight); > + > + /* > + * Historically, reset was not implemented so only reset devices > + * that are expecting it. > + */ > + if (!virtio_has_feature(dev->protocol_features, > + VHOST_USER_PROTOCOL_F_RESET_DEVICE)) { > + return; > + } > + > + if (dev->vhost_ops->vhost_reset_device) { > + dev->vhost_ops->vhost_reset_device(dev); > + } > + > } This should be a generic protocol feature that all vhost-user device implementations benefit from. Devices shouldn't have to explicitly implement it over and over again. Why isn't vhost_user_reset_device() called already? Then it wouldn't be necessary to modify vhost_user_blk_reset(). Stefan