Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost

Pull virtio updates from Michael Tsirkin:
 "virtio,vhost,vdpa: features, fixes, and cleanups:

   - reduction in interrupt rate in virtio

   - perf improvement for VDUSE

   - scalability for vhost-scsi

   - non power of 2 ring support for packed rings

   - better management for mlx5 vdpa

   - suspend for snet

   - VIRTIO_F_NOTIFICATION_DATA

   - shared backend with vdpa-sim-blk

   - user VA support in vdpa-sim

   - better struct packing for virtio

  and fixes, cleanups all over the place"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost: (52 commits)
  vhost_vdpa: fix unmap process in no-batch mode
  MAINTAINERS: make me a reviewer of VIRTIO CORE AND NET DRIVERS
  tools/virtio: fix build caused by virtio_ring changes
  virtio_ring: add a struct device forward declaration
  vdpa_sim_blk: support shared backend
  vdpa_sim: move buffer allocation in the devices
  vdpa/snet: use likely/unlikely macros in hot functions
  vdpa/snet: implement kick_vq_with_data callback
  virtio-vdpa: add VIRTIO_F_NOTIFICATION_DATA feature support
  virtio: add VIRTIO_F_NOTIFICATION_DATA feature support
  vdpa/snet: support the suspend vDPA callback
  vdpa/snet: support getting and setting VQ state
  MAINTAINERS: add vringh.h to Virtio Core and Net Drivers
  vringh: address kdoc warnings
  vdpa: address kdoc warnings
  virtio_ring: don't update event idx on get_buf
  vdpa_sim: add support for user VA
  vdpa_sim: replace the spinlock with a mutex to protect the state
  vdpa_sim: use kthread worker
  vdpa_sim: make devices agnostic for work management
  ...

32 files changed, 1761 insertions, 481 deletions

diff --git a/MAINTAINERS b/MAINTAINERS
index a53a1a29b10c..23d628857400 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -22212,6 +22212,7 @@ F:	include/uapi/linux/virtio_console.h
 VIRTIO CORE AND NET DRIVERS
 M:	"Michael S. Tsirkin" <mst@redhat.com>
 M:	Jason Wang <jasowang@redhat.com>
+R:	Xuan Zhuo <xuanzhuo@linux.alibaba.com>
 L:	virtualization@lists.linux-foundation.org
 S:	Maintained
 F:	Documentation/ABI/testing/sysfs-bus-vdpa
@@ -22225,6 +22226,7 @@ F:	drivers/vdpa/
 F:	drivers/virtio/
 F:	include/linux/vdpa.h
 F:	include/linux/virtio*.h
+F:	include/linux/vringh.h
 F:	include/uapi/linux/virtio_*.h
 F:	tools/virtio/
 
diff --git a/drivers/s390/virtio/virtio_ccw.c b/drivers/s390/virtio/virtio_ccw.c
index 954fc31b4bc7..02922768b129 100644
--- a/drivers/s390/virtio/virtio_ccw.c
+++ b/drivers/s390/virtio/virtio_ccw.c
@@ -391,7 +391,7 @@ static void virtio_ccw_drop_indicator(struct virtio_ccw_device *vcdev,
 	ccw_device_dma_free(vcdev->cdev, thinint_area, sizeof(*thinint_area));
 }
 
-static bool virtio_ccw_kvm_notify(struct virtqueue *vq)
+static inline bool virtio_ccw_do_kvm_notify(struct virtqueue *vq, u32 data)
 {
 	struct virtio_ccw_vq_info *info = vq->priv;
 	struct virtio_ccw_device *vcdev;
@@ -402,12 +402,22 @@ static bool virtio_ccw_kvm_notify(struct virtqueue *vq)
 	BUILD_BUG_ON(sizeof(struct subchannel_id) != sizeof(unsigned int));
 	info->cookie = kvm_hypercall3(KVM_S390_VIRTIO_CCW_NOTIFY,
 				      *((unsigned int *)&schid),
-				      vq->index, info->cookie);
+				      data, info->cookie);
 	if (info->cookie < 0)
 		return false;
 	return true;
 }
 
+static bool virtio_ccw_kvm_notify(struct virtqueue *vq)
+{
+	return virtio_ccw_do_kvm_notify(vq, vq->index);
+}
+
+static bool virtio_ccw_kvm_notify_with_data(struct virtqueue *vq)
+{
+	return virtio_ccw_do_kvm_notify(vq, vring_notification_data(vq));
+}
+
 static int virtio_ccw_read_vq_conf(struct virtio_ccw_device *vcdev,
 				   struct ccw1 *ccw, int index)
 {
@@ -495,6 +505,7 @@ static struct virtqueue *virtio_ccw_setup_vq(struct virtio_device *vdev,
 					     struct ccw1 *ccw)
 {
 	struct virtio_ccw_device *vcdev = to_vc_device(vdev);
+	bool (*notify)(struct virtqueue *vq);
 	int err;
 	struct virtqueue *vq = NULL;
 	struct virtio_ccw_vq_info *info;
@@ -502,6 +513,11 @@ static struct virtqueue *virtio_ccw_setup_vq(struct virtio_device *vdev,
 	unsigned long flags;
 	bool may_reduce;
 
+	if (__virtio_test_bit(vdev, VIRTIO_F_NOTIFICATION_DATA))
+		notify = virtio_ccw_kvm_notify_with_data;
+	else
+		notify = virtio_ccw_kvm_notify;
+
 	/* Allocate queue. */
 	info = kzalloc(sizeof(struct virtio_ccw_vq_info), GFP_KERNEL);
 	if (!info) {
@@ -524,7 +540,7 @@ static struct virtqueue *virtio_ccw_setup_vq(struct virtio_device *vdev,
 	may_reduce = vcdev->revision > 0;
 	vq = vring_create_virtqueue(i, info->num, KVM_VIRTIO_CCW_RING_ALIGN,
 				    vdev, true, may_reduce, ctx,
-				    virtio_ccw_kvm_notify, callback, name);
+				    notify, callback, name);
 
 	if (!vq) {
 		/* For now, we fail if we can't get the requested size. */
diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index 195963b82b63..e29e32b306ad 100644
--- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
@@ -778,12 +778,28 @@ static bool vq_is_tx(u16 idx)
 	return idx % 2;
 }
 
-static u16 get_features_12_3(u64 features)
+enum {
+	MLX5_VIRTIO_NET_F_MRG_RXBUF = 2,
+	MLX5_VIRTIO_NET_F_HOST_ECN = 4,
+	MLX5_VIRTIO_NET_F_GUEST_ECN = 6,
+	MLX5_VIRTIO_NET_F_GUEST_TSO6 = 7,
+	MLX5_VIRTIO_NET_F_GUEST_TSO4 = 8,
+	MLX5_VIRTIO_NET_F_GUEST_CSUM = 9,
+	MLX5_VIRTIO_NET_F_CSUM = 10,
+	MLX5_VIRTIO_NET_F_HOST_TSO6 = 11,
+	MLX5_VIRTIO_NET_F_HOST_TSO4 = 12,
+};
+
+static u16 get_features(u64 features)
 {
-	return (!!(features & BIT_ULL(VIRTIO_NET_F_HOST_TSO4)) << 9) |
-	       (!!(features & BIT_ULL(VIRTIO_NET_F_HOST_TSO6)) << 8) |
-	       (!!(features & BIT_ULL(VIRTIO_NET_F_CSUM)) << 7) |
-	       (!!(features & BIT_ULL(VIRTIO_NET_F_GUEST_CSUM)) << 6);
+	return (!!(features & BIT_ULL(VIRTIO_NET_F_MRG_RXBUF)) << MLX5_VIRTIO_NET_F_MRG_RXBUF) |
+	       (!!(features & BIT_ULL(VIRTIO_NET_F_HOST_ECN)) << MLX5_VIRTIO_NET_F_HOST_ECN) |
+	       (!!(features & BIT_ULL(VIRTIO_NET_F_GUEST_ECN)) << MLX5_VIRTIO_NET_F_GUEST_ECN) |
+	       (!!(features & BIT_ULL(VIRTIO_NET_F_GUEST_TSO6)) << MLX5_VIRTIO_NET_F_GUEST_TSO6) |
+	       (!!(features & BIT_ULL(VIRTIO_NET_F_GUEST_TSO4)) << MLX5_VIRTIO_NET_F_GUEST_TSO4) |
+	       (!!(features & BIT_ULL(VIRTIO_NET_F_CSUM)) << MLX5_VIRTIO_NET_F_CSUM) |
+	       (!!(features & BIT_ULL(VIRTIO_NET_F_HOST_TSO6)) << MLX5_VIRTIO_NET_F_HOST_TSO6) |
+	       (!!(features & BIT_ULL(VIRTIO_NET_F_HOST_TSO4)) << MLX5_VIRTIO_NET_F_HOST_TSO4);
 }
 
 static bool counters_supported(const struct mlx5_vdpa_dev *mvdev)
@@ -797,6 +813,7 @@ static int create_virtqueue(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtque
 	int inlen = MLX5_ST_SZ_BYTES(create_virtio_net_q_in);
 	u32 out[MLX5_ST_SZ_DW(create_virtio_net_q_out)] = {};
 	void *obj_context;
+	u16 mlx_features;
 	void *cmd_hdr;
 	void *vq_ctx;
 	void *in;
@@ -812,6 +829,7 @@ static int create_virtqueue(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtque
 		goto err_alloc;
 	}
 
+	mlx_features = get_features(ndev->mvdev.actual_features);
 	cmd_hdr = MLX5_ADDR_OF(create_virtio_net_q_in, in, general_obj_in_cmd_hdr);
 
 	MLX5_SET(general_obj_in_cmd_hdr, cmd_hdr, opcode, MLX5_CMD_OP_CREATE_GENERAL_OBJECT);
@@ -822,7 +840,9 @@ static int create_virtqueue(struct mlx5_vdpa_net *ndev, struct mlx5_vdpa_virtque
 	MLX5_SET(virtio_net_q_object, obj_context, hw_available_index, mvq->avail_idx);
 	MLX5_SET(virtio_net_q_object, obj_context, hw_used_index, mvq->used_idx);
 	MLX5_SET(virtio_net_q_object, obj_context, queue_feature_bit_mask_12_3,
-		 get_features_12_3(ndev->mvdev.actual_features));
+		 mlx_features >> 3);
+	MLX5_SET(virtio_net_q_object, obj_context, queue_feature_bit_mask_2_0,
+		 mlx_features & 7);
 	vq_ctx = MLX5_ADDR_OF(virtio_net_q_object, obj_context, virtio_q_context);
 	MLX5_SET(virtio_q, vq_ctx, virtio_q_type, get_queue_type(ndev));
 
@@ -2171,23 +2191,27 @@ static u32 mlx5_vdpa_get_vq_group(struct vdpa_device *vdev, u16 idx)
 	return MLX5_VDPA_DATAVQ_GROUP;
 }
 
-enum { MLX5_VIRTIO_NET_F_GUEST_CSUM = 1 << 9,
-	MLX5_VIRTIO_NET_F_CSUM = 1 << 10,
-	MLX5_VIRTIO_NET_F_HOST_TSO6 = 1 << 11,
-	MLX5_VIRTIO_NET_F_HOST_TSO4 = 1 << 12,
-};
-
 static u64 mlx_to_vritio_features(u16 dev_features)
 {
 	u64 result = 0;
 
-	if (dev_features & MLX5_VIRTIO_NET_F_GUEST_CSUM)
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_MRG_RXBUF))
+		result |= BIT_ULL(VIRTIO_NET_F_MRG_RXBUF);
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_HOST_ECN))
+		result |= BIT_ULL(VIRTIO_NET_F_HOST_ECN);
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_GUEST_ECN))
+		result |= BIT_ULL(VIRTIO_NET_F_GUEST_ECN);
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_GUEST_TSO6))
+		result |= BIT_ULL(VIRTIO_NET_F_GUEST_TSO6);
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_GUEST_TSO4))
+		result |= BIT_ULL(VIRTIO_NET_F_GUEST_TSO4);
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_GUEST_CSUM))
 		result |= BIT_ULL(VIRTIO_NET_F_GUEST_CSUM);
-	if (dev_features & MLX5_VIRTIO_NET_F_CSUM)
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_CSUM))
 		result |= BIT_ULL(VIRTIO_NET_F_CSUM);
-	if (dev_features & MLX5_VIRTIO_NET_F_HOST_TSO6)
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_HOST_TSO6))
 		result |= BIT_ULL(VIRTIO_NET_F_HOST_TSO6);
-	if (dev_features & MLX5_VIRTIO_NET_F_HOST_TSO4)
+	if (dev_features & BIT_ULL(MLX5_VIRTIO_NET_F_HOST_TSO4))
 		result |= BIT_ULL(VIRTIO_NET_F_HOST_TSO4);
 
 	return result;
@@ -2298,6 +2322,113 @@ static void update_cvq_info(struct mlx5_vdpa_dev *mvdev)
 	}
 }
 
+static u8 query_vport_state(struct mlx5_core_dev *mdev, u8 opmod, u16 vport)
+{
+	u32 out[MLX5_ST_SZ_DW(query_vport_state_out)] = {};
+	u32 in[MLX5_ST_SZ_DW(query_vport_state_in)] = {};
+	int err;
+
+	MLX5_SET(query_vport_state_in, in, opcode, MLX5_CMD_OP_QUERY_VPORT_STATE);
+	MLX5_SET(query_vport_state_in, in, op_mod, opmod);
+	MLX5_SET(query_vport_state_in, in, vport_number, vport);
+	if (vport)
+		MLX5_SET(query_vport_state_in, in, other_vport, 1);
+
+	err = mlx5_cmd_exec_inout(mdev, query_vport_state, in, out);
+	if (err)
+		return 0;
+
+	return MLX5_GET(query_vport_state_out, out, state);
+}
+
+static bool get_link_state(struct mlx5_vdpa_dev *mvdev)
+{
+	if (query_vport_state(mvdev->mdev, MLX5_VPORT_STATE_OP_MOD_VNIC_VPORT, 0) ==
+	    VPORT_STATE_UP)
+		return true;
+
+	return false;
+}
+
+static void update_carrier(struct work_struct *work)
+{
+	struct mlx5_vdpa_wq_ent *wqent;
+	struct mlx5_vdpa_dev *mvdev;
+	struct mlx5_vdpa_net *ndev;
+
+	wqent = container_of(work, struct mlx5_vdpa_wq_ent, work);
+	mvdev = wqent->mvdev;
+	ndev = to_mlx5_vdpa_ndev(mvdev);
+	if (get_link_state(mvdev))
+		ndev->config.status |= cpu_to_mlx5vdpa16(mvdev, VIRTIO_NET_S_LINK_UP);
+	else
+		ndev->config.status &= cpu_to_mlx5vdpa16(mvdev, ~VIRTIO_NET_S_LINK_UP);
+
+	if (ndev->config_cb.callback)
+		ndev->config_cb.callback(ndev->config_cb.private);
+
+	kfree(wqent);
+}
+
+static int queue_link_work(struct mlx5_vdpa_net *ndev)
+{
+	struct mlx5_vdpa_wq_ent *wqent;
+
+	wqent = kzalloc(sizeof(*wqent), GFP_ATOMIC);
+	if (!wqent)
+		return -ENOMEM;
+
+	wqent->mvdev = &ndev->mvdev;
+	INIT_WORK(&wqent->work, update_carrier);
+	queue_work(ndev->mvdev.wq, &wqent->work);
+	return 0;
+}
+
+static int event_handler(struct notifier_block *nb, unsigned long event, void *param)
+{
+	struct mlx5_vdpa_net *ndev = container_of(nb, struct mlx5_vdpa_net, nb);
+	struct mlx5_eqe *eqe = param;
+	int ret = NOTIFY_DONE;
+
+	if (event == MLX5_EVENT_TYPE_PORT_CHANGE) {
+		switch (eqe->sub_type) {
+		case MLX5_PORT_CHANGE_SUBTYPE_DOWN:
+		case MLX5_PORT_CHANGE_SUBTYPE_ACTIVE:
+			if (queue_link_work(ndev))
+				return NOTIFY_DONE;
+
+			ret = NOTIFY_OK;
+			break;
+		default:
+			return NOTIFY_DONE;
+		}
+		return ret;
+	}
+	return ret;
+}
+
+static void register_link_notifier(struct mlx5_vdpa_net *ndev)
+{
+	if (!(ndev->mvdev.actual_features & BIT_ULL(VIRTIO_NET_F_STATUS)))
+		return;
+
+	ndev->nb.notifier_call = event_handler;
+	mlx5_notifier_register(ndev->mvdev.mdev, &ndev->nb);
+	ndev->nb_registered = true;
+	queue_link_work(ndev);
+}
+
+static void unregister_link_notifier(struct mlx5_vdpa_net *ndev)
+{
+	if (!ndev->nb_registered)
+		return;
+
+	ndev->nb_registered = false;
+	mlx5_notifier_unregister(ndev->mvdev.mdev, &ndev->nb);
+	if (ndev->mvdev.wq)
+		flush_workqueue(ndev->mvdev.wq);
+}
+
 static int mlx5_vdpa_set_driver_features(struct vdpa_device *vdev, u64 features)
 {
 	struct mlx5_vdpa_dev *mvdev = to_mvdev(vdev);
@@ -2567,10 +2698,11 @@ static void mlx5_vdpa_set_status(struct vdpa_device *vdev, u8 status)
 				mlx5_vdpa_warn(mvdev, "failed to setup control VQ vring\n");
 				goto err_setup;
 			}
+			register_link_notifier(ndev);
 			err = setup_driver(mvdev);
 			if (err) {
 				mlx5_vdpa_warn(mvdev, "failed to setup driver\n");
-				goto err_setup;
+				goto err_driver;
 			}
 		} else {
 			mlx5_vdpa_warn(mvdev, "did not expect DRIVER_OK to be cleared\n");
@@ -2582,6 +2714,8 @@ static void mlx5_vdpa_set_status(struct vdpa_device *vdev, u8 status)
 	up_write(&ndev->reslock);
 	return;
 
+err_driver:
+	unregister_link_notifier(ndev);
 err_setup:
 	mlx5_vdpa_destroy_mr(&ndev->mvdev);
 	ndev->mvdev.status |= VIRTIO_CONFIG_S_FAILED;
@@ -2607,6 +2741,7 @@ static int mlx5_vdpa_reset(struct vdpa_device *vdev)
 	mlx5_vdpa_info(mvdev, "performing device reset\n");
 
 	down_write(&ndev->reslock);
+	unregister_link_notifier(ndev);
 	teardown_driver(ndev);
 	clear_vqs_ready(ndev);
 	mlx5_vdpa_destroy_mr(&ndev->mvdev);
@@ -2861,9 +2996,7 @@ static int mlx5_vdpa_suspend(struct vdpa_device *vdev)
 	mlx5_vdpa_info(mvdev, "suspending device\n");
 
 	down_write(&ndev->reslock);
-	ndev->nb_registered = false;
-	mlx5_notifier_unregister(mvdev->mdev, &ndev->nb);
-	flush_workqueue(ndev->mvdev.wq);
+	unregister_link_notifier(ndev);
 	for (i = 0; i < ndev->cur_num_vqs; i++) {
 		mvq = &ndev->vqs[i];
 		suspend_vq(ndev, mvq);
@@ -3000,84 +3133,6 @@ struct mlx5_vdpa_mgmtdev {
 	struct mlx5_vdpa_net *ndev;
 };
 
-static u8 query_vport_state(struct mlx5_core_dev *mdev, u8 opmod, u16 vport)
-{
-	u32 out[MLX5_ST_SZ_DW(query_vport_state_out)] = {};
-	u32 in[MLX5_ST_SZ_DW(query_vport_state_in)] = {};
-	int err;
-
-	MLX5_SET(query_vport_state_in, in, opcode, MLX5_CMD_OP_QUERY_VPORT_STATE);
-	MLX5_SET(query_vport_state_in, in, op_mod, opmod);
-	MLX5_SET(query_vport_state_in, in, vport_number, vport);
-	if (vport)
-		MLX5_SET(query_vport_state_in, in, other_vport, 1);
-
-	err = mlx5_cmd_exec_inout(mdev, query_vport_state, in, out);
-	if (err)
-		return 0;
-
-	return MLX5_GET(query_vport_state_out, out, state);
-}
-
-static bool get_link_state(struct mlx5_vdpa_dev *mvdev)
-{
-	if (query_vport_state(mvdev->mdev, MLX5_VPORT_STATE_OP_MOD_VNIC_VPORT, 0) ==
-	    VPORT_STATE_UP)
-		return true;
-
-	return false;
-}
-
-static void update_carrier(struct work_struct *work)
-{
-	struct mlx5_vdpa_wq_ent *wqent;
-	struct mlx5_vdpa_dev *mvdev;
-	struct mlx5_vdpa_net *ndev;
-
-	wqent = container_of(work, struct mlx5_vdpa_wq_ent, work);
-	mvdev = wqent->mvdev;
-	ndev = to_mlx5_vdpa_ndev(mvdev);
-	if (get_link_state(mvdev))
-		ndev->config.status |= cpu_to_mlx5vdpa16(mvdev, VIRTIO_NET_S_LINK_UP);
-	else
-		ndev->config.status &= cpu_to_mlx5vdpa16(mvdev, ~VIRTIO_NET_S_LINK_UP);
-
-	if (ndev->nb_registered && ndev->config_cb.callback)
-		ndev->config_cb.callback(ndev->config_cb.private);
-
-	kfree(wqent);
-}
-
-static int event_handler(struct notifier_block *nb, unsigned long event, void *param)
-{
-	struct mlx5_vdpa_net *ndev = container_of(nb, struct mlx5_vdpa_net, nb);
-	struct mlx5_eqe *eqe = param;
-	int ret = NOTIFY_DONE;
-	struct mlx5_vdpa_wq_ent *wqent;
-
-	if (event == MLX5_EVENT_TYPE_PORT_CHANGE) {
-		if (!(ndev->mvdev.actual_features & BIT_ULL(VIRTIO_NET_F_STATUS)))
-			return NOTIFY_DONE;
-		switch (eqe->sub_type) {
-		case MLX5_PORT_CHANGE_SUBTYPE_DOWN:
-		case MLX5_PORT_CHANGE_SUBTYPE_ACTIVE:
-			wqent = kzalloc(sizeof(*wqent), GFP_ATOMIC);
-			if (!wqent)
-				return NOTIFY_DONE;
-
-			wqent->mvdev = &ndev->mvdev;
-			INIT_WORK(&wqent->work, update_carrier);
-			queue_work(ndev->mvdev.wq, &wqent->work);
-			ret = NOTIFY_OK;
-			break;
-		default:
-			return NOTIFY_DONE;
-		}
-		return ret;
-	}
-	return ret;
-}
-
 static int config_func_mtu(struct mlx5_core_dev *mdev, u16 mtu)
 {
 	int inlen = MLX5_ST_SZ_BYTES(modify_nic_vport_context_in);
@@ -3127,6 +3182,8 @@ static int mlx5_vdpa_dev_add(struct vdpa_mgmt_dev *v_mdev, const char *name,
 			return -EINVAL;
 		}
 		device_features &= add_config->device_features;
+	} else {
+		device_features &= ~BIT_ULL(VIRTIO_NET_F_MRG_RXBUF);
 	}
 	if (!(device_features & BIT_ULL(VIRTIO_F_VERSION_1) &&
 	      device_features & BIT_ULL(VIRTIO_F_ACCESS_PLATFORM))) {
@@ -3258,9 +3315,6 @@ static int mlx5_vdpa_dev_add(struct vdpa_mgmt_dev *v_mdev, const char *name,
 		goto err_res2;
 	}
 
-	ndev->nb.notifier_call = event_handler;
-	mlx5_notifier_register(mdev, &ndev->nb);
-	ndev->nb_registered = true;
 	mvdev->vdev.mdev = &mgtdev->mgtdev;
 	err = _vdpa_register_device(&mvdev->vdev, max_vqs + 1);
 	if (err)
@@ -3294,10 +3348,7 @@ static void mlx5_vdpa_dev_del(struct vdpa_mgmt_dev *v_mdev, struct vdpa_device *
 
 	mlx5_vdpa_remove_debugfs(ndev->debugfs);
 	ndev->debugfs = NULL;
-	if (ndev->nb_registered) {
-		ndev->nb_registered = false;
-		mlx5_notifier_unregister(mvdev->mdev, &ndev->nb);
-	}
+	unregister_link_notifier(ndev);
 	wq = mvdev->wq;
 	mvdev->wq = NULL;
 	destroy_workqueue(wq);
diff --git a/drivers/vdpa/solidrun/Makefile b/drivers/vdpa/solidrun/Makefile
index c0aa3415bf7b..9116252cd5fa 100644
--- a/drivers/vdpa/solidrun/Makefile
+++ b/drivers/vdpa/solidrun/Makefile
@@ -1,6 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0
 obj-$(CONFIG_SNET_VDPA) += snet_vdpa.o
 snet_vdpa-$(CONFIG_SNET_VDPA) += snet_main.o
+snet_vdpa-$(CONFIG_SNET_VDPA) += snet_ctrl.o
 ifdef CONFIG_HWMON
 snet_vdpa-$(CONFIG_SNET_VDPA) += snet_hwmon.o
 endif
diff --git a/drivers/vdpa/solidrun/snet_ctrl.c b/drivers/vdpa/solidrun/snet_ctrl.c
new file mode 100644
index 000000000000..3858738643b4
--- /dev/null
+++ b/drivers/vdpa/solidrun/snet_ctrl.c
@@ -0,0 +1,330 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * SolidRun DPU driver for control plane
+ *
+ * Copyright (C) 2022-2023 SolidRun
+ *
+ * Author: Alvaro Karsz <alvaro.karsz@solid-run.com>
+ *
+ */
+
+#include <linux/iopoll.h>
+
+#include "snet_vdpa.h"
+
+enum snet_ctrl_opcodes {
+	SNET_CTRL_OP_DESTROY = 1,
+	SNET_CTRL_OP_READ_VQ_STATE,
+	SNET_CTRL_OP_SUSPEND,
+};
+
+#define SNET_CTRL_TIMEOUT	        2000000
+
+#define SNET_CTRL_DATA_SIZE_MASK	0x0000FFFF
+#define SNET_CTRL_IN_PROCESS_MASK	0x00010000
+#define SNET_CTRL_CHUNK_RDY_MASK	0x00020000
+#define SNET_CTRL_ERROR_MASK		0x0FFC0000
+
+#define SNET_VAL_TO_ERR(val)		(-(((val) & SNET_CTRL_ERROR_MASK) >> 18))
+#define SNET_EMPTY_CTRL(val)		(((val) & SNET_CTRL_ERROR_MASK) || \
+						!((val) & SNET_CTRL_IN_PROCESS_MASK))
+#define SNET_DATA_READY(val)		((val) & (SNET_CTRL_ERROR_MASK | SNET_CTRL_CHUNK_RDY_MASK))
+
+/* Control register used to read data from the DPU */
+struct snet_ctrl_reg_ctrl {
+	/* Chunk size in 4B words */
+	u16 data_size;
+	/* We are in the middle of a command */
+	u16 in_process:1;
+	/* A data chunk is ready and can be consumed */
+	u16 chunk_ready:1;
+	/* Error code */
+	u16 error:10;
+	/* Saved for future usage */
+	u16 rsvd:4;
+};
+
+/* Opcode register */
+struct snet_ctrl_reg_op {
+	u16 opcode;
+	/* Only if VQ index is relevant for the command */
+	u16 vq_idx;
+};
+
+struct snet_ctrl_regs {
+	struct snet_ctrl_reg_op op;
+	struct snet_ctrl_reg_ctrl ctrl;
+	u32 rsvd;
+	u32 data[];
+};
+
+static struct snet_ctrl_regs __iomem *snet_get_ctrl(struct snet *snet)
+{
+	return snet->bar + snet->psnet->cfg.ctrl_off;
+}
+
+static int snet_wait_for_empty_ctrl(struct snet_ctrl_regs __iomem *regs)
+{
+	u32 val;
+
+	return readx_poll_timeout(ioread32, &regs->ctrl, val, SNET_EMPTY_CTRL(val), 10,
+				  SNET_CTRL_TIMEOUT);
+}
+
+static int snet_wait_for_empty_op(struct snet_ctrl_regs __iomem *regs)
+{
+	u32 val;
+
+	return readx_poll_timeout(ioread32, &regs->op, val, !val, 10, SNET_CTRL_TIMEOUT);
+}
+
+static int snet_wait_for_data(struct snet_ctrl_regs __iomem *regs)
+{
+	u32 val;
+
+	return readx_poll_timeout(ioread32, &regs->ctrl, val, SNET_DATA_READY(val), 10,
+				  SNET_CTRL_TIMEOUT);
+}
+
+static u32 snet_read32_word(struct snet_ctrl_regs __iomem *ctrl_regs, u16 word_idx)
+{
+	return ioread32(&ctrl_regs->data[word_idx]);
+}
+
+static u32 snet_read_ctrl(struct snet_ctrl_regs __iomem *ctrl_regs)
+{
+	return ioread32(&ctrl_regs->ctrl);
+}
+
+static void snet_write_ctrl(struct snet_ctrl_regs __iomem *ctrl_regs, u32 val)
+{
+	iowrite32(val, &ctrl_regs->ctrl);
+}
+
+static void snet_write_op(struct snet_ctrl_regs __iomem *ctrl_regs, u32 val)
+{
+	iowrite32(val, &ctrl_regs->op);
+}
+
+static int snet_wait_for_dpu_completion(struct snet_ctrl_regs __iomem *ctrl_regs)
+{
+	/* Wait until the DPU finishes completely.
+	 * It will clear the opcode register.
+	 */
+	return snet_wait_for_empty_op(ctrl_regs);
+}
+
+/* Reading ctrl from the DPU:
+ * buf_size must be 4B aligned
+ *
+ * Steps:
+ *
+ * (1) Verify that the DPU is not in the middle of another operation by
+ *     reading the in_process and error bits in the control register.
+ * (2) Write the request opcode and the VQ idx in the opcode register
+ *     and write the buffer size in the control register.
+ * (3) Start readind chunks of data, chunk_ready bit indicates that a
+ *     data chunk is available, we signal that we read the data by clearing the bit.
+ * (4) Detect that the transfer is completed when the in_process bit
+ *     in the control register is cleared or when the an error appears.
+ */
+static int snet_ctrl_read_from_dpu(struct snet *snet, u16 opcode, u16 vq_idx, void *buffer,
+				   u32 buf_size)
+{
+	struct pci_dev *pdev = snet->pdev;
+	struct snet_ctrl_regs __iomem *regs = snet_get_ctrl(snet);
+	u32 *bfr_ptr = (u32 *)buffer;
+	u32 val;
+	u16 buf_words;
+	int ret;
+	u16 words, i, tot_words = 0;
+
+	/* Supported for config 2+ */
+	if (!SNET_CFG_VER(snet, 2))
+		return -EOPNOTSUPP;
+
+	if (!IS_ALIGNED(buf_size, 4))
+		return -EINVAL;
+
+	mutex_lock(&snet->ctrl_lock);
+
+	buf_words = buf_size / 4;
+
+	/* Make sure control register is empty */
+	ret = snet_wait_for_empty_ctrl(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for previous control data to be consumed\n");
+		goto exit;
+	}
+
+	/* We need to write the buffer size in the control register, and the opcode + vq index in
+	 * the opcode register.
+	 * We use a spinlock to serialize the writes.
+	 */
+	spin_lock(&snet->ctrl_spinlock);
+
+	snet_write_ctrl(regs, buf_words);
+	snet_write_op(regs, opcode | (vq_idx << 16));
+
+	spin_unlock(&snet->ctrl_spinlock);
+
+	while (buf_words != tot_words) {
+		ret = snet_wait_for_data(regs);
+		if (ret) {
+			SNET_WARN(pdev, "Timeout waiting for control data\n");
+			goto exit;
+		}
+
+		val = snet_read_ctrl(regs);
+
+		/* Error? */
+		if (val & SNET_CTRL_ERROR_MASK) {
+			ret = SNET_VAL_TO_ERR(val);
+			SNET_WARN(pdev, "Error while reading control data from DPU, err %d\n", ret);
+			goto exit;
+		}
+
+		words = min_t(u16, val & SNET_CTRL_DATA_SIZE_MASK, buf_words - tot_words);
+
+		for (i = 0; i < words; i++) {
+			*bfr_ptr = snet_read32_word(regs, i);
+			bfr_ptr++;
+		}
+
+		tot_words += words;
+
+		/* Is the job completed? */
+		if (!(val & SNET_CTRL_IN_PROCESS_MASK))
+			break;
+
+		/* Clear the chunk ready bit and continue */
+		val &= ~SNET_CTRL_CHUNK_RDY_MASK;
+		snet_write_ctrl(regs, val);
+	}
+
+	ret = snet_wait_for_dpu_completion(regs);
+	if (ret)
+		SNET_WARN(pdev, "Timeout waiting for the DPU to complete a control command\n");
+
+exit:
+	mutex_unlock(&snet->ctrl_lock);
+	return ret;
+}
+
+/* Send a control message to the DPU using the old mechanism
+ * used with config version 1.
+ */
+static int snet_send_ctrl_msg_old(struct snet *snet, u32 opcode)
+{
+	struct pci_dev *pdev = snet->pdev;
+	struct snet_ctrl_regs __iomem *regs = snet_get_ctrl(snet);
+	int ret;
+
+	mutex_lock(&snet->ctrl_lock);
+
+	/* Old mechanism uses just 1 register, the opcode register.
+	 * Make sure that the opcode register is empty, and that the DPU isn't
+	 * processing an old message.
+	 */
+	ret = snet_wait_for_empty_op(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for previous control message to be ACKed\n");
+		goto exit;
+	}
+
+	/* Write the message */
+	snet_write_op(regs, opcode);
+
+	/* DPU ACKs the message by clearing the opcode register */
+	ret = snet_wait_for_empty_op(regs);
+	if (ret)
+		SNET_WARN(pdev, "Timeout waiting for a control message to be ACKed\n");
+
+exit:
+	mutex_unlock(&snet->ctrl_lock);
+	return ret;
+}
+
+/* Send a control message to the DPU.
+ * A control message is a message without payload.
+ */
+static int snet_send_ctrl_msg(struct snet *snet, u16 opcode, u16 vq_idx)
+{
+	struct pci_dev *pdev = snet->pdev;
+	struct snet_ctrl_regs __iomem *regs = snet_get_ctrl(snet);
+	u32 val;
+	int ret;
+
+	/* If config version is not 2+, use the old mechanism */
+	if (!SNET_CFG_VER(snet, 2))
+		return snet_send_ctrl_msg_old(snet, opcode);
+
+	mutex_lock(&snet->ctrl_lock);
+
+	/* Make sure control register is empty */
+	ret = snet_wait_for_empty_ctrl(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for previous control data to be consumed\n");
+		goto exit;
+	}
+
+	/* We need to clear the control register and write the opcode + vq index in the opcode
+	 * register.
+	 * We use a spinlock to serialize the writes.
+	 */
+	spin_lock(&snet->ctrl_spinlock);
+
+	snet_write_ctrl(regs, 0);
+	snet_write_op(regs, opcode | (vq_idx << 16));
+
+	spin_unlock(&snet->ctrl_spinlock);
+
+	/* The DPU ACKs control messages by setting the chunk ready bit
+	 * without data.
+	 */
+	ret = snet_wait_for_data(regs);
+	if (ret) {
+		SNET_WARN(pdev, "Timeout waiting for control message to be ACKed\n");
+		goto exit;
+	}
+
+	/* Check for errors */
+	val = snet_read_ctrl(regs);
+	ret = SNET_VAL_TO_ERR(val);
+
+	/* Clear the chunk ready bit */
+	val &= ~SNET_CTRL_CHUNK_RDY_MASK;
+	snet_write_ctrl(regs, val);
+
+	ret = snet_wait_for_dpu_completion(regs);
+	if (ret)
+		SNET_WARN(pdev, "Timeout waiting for DPU to complete a control command, err %d\n",
+			  ret);
+
+exit:
+	mutex_unlock(&snet->ctrl_loc