Merging upstream version 6.7.7.

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

6 files changed, 247 insertions, 200 deletions

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.c b/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.c
index 05e148db98..c93492b677 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.c
@@ -14,17 +14,22 @@
 
 struct mlx5_sf_dev_table {
 	struct xarray devices;
-	unsigned int max_sfs;
 	phys_addr_t base_address;
 	u64 sf_bar_length;
 	struct notifier_block nb;
-	struct mutex table_lock; /* Serializes sf life cycle and vhca state change handler */
 	struct workqueue_struct *active_wq;
 	struct work_struct work;
 	u8 stop_active_wq:1;
 	struct mlx5_core_dev *dev;
 };
 
+struct mlx5_sf_dev_active_work_ctx {
+	struct work_struct work;
+	struct mlx5_vhca_state_event event;
+	struct mlx5_sf_dev_table *table;
+	int sf_index;
+};
+
 static bool mlx5_sf_dev_supported(const struct mlx5_core_dev *dev)
 {
 	return MLX5_CAP_GEN(dev, sf) && mlx5_vhca_event_supported(dev);
@@ -110,12 +115,6 @@ static void mlx5_sf_dev_add(struct mlx5_core_dev *dev, u16 sf_index, u16 fn_id,
 	sf_dev->parent_mdev = dev;
 	sf_dev->fn_id = fn_id;
 
-	if (!table->max_sfs) {
-		mlx5_adev_idx_free(id);
-		kfree(sf_dev);
-		err = -EOPNOTSUPP;
-		goto add_err;
-	}
 	sf_dev->bar_base_addr = table->base_address + (sf_index * table->sf_bar_length);
 
 	trace_mlx5_sf_dev_add(dev, sf_dev, id);
@@ -172,7 +171,6 @@ mlx5_sf_dev_state_change_handler(struct notifier_block *nb, unsigned long event_
 		return 0;
 
 	sf_index = event->function_id - base_id;
-	mutex_lock(&table->table_lock);
 	sf_dev = xa_load(&table->devices, sf_index);
 	switch (event->new_vhca_state) {
 	case MLX5_VHCA_STATE_INVALID:
@@ -196,7 +194,6 @@ mlx5_sf_dev_state_change_handler(struct notifier_block *nb, unsigned long event_
 	default:
 		break;
 	}
-	mutex_unlock(&table->table_lock);
 	return 0;
 }
 
@@ -221,15 +218,44 @@ static int mlx5_sf_dev_vhca_arm_all(struct mlx5_sf_dev_table *table)
 	return 0;
 }
 
-static void mlx5_sf_dev_add_active_work(struct work_struct *work)
+static void mlx5_sf_dev_add_active_work(struct work_struct *_work)
 {
-	struct mlx5_sf_dev_table *table = container_of(work, struct mlx5_sf_dev_table, work);
+	struct mlx5_sf_dev_active_work_ctx *work_ctx;
+
+	work_ctx = container_of(_work, struct mlx5_sf_dev_active_work_ctx, work);
+	if (work_ctx->table->stop_active_wq)
+		goto out;
+	/* Don't probe device which is already probe */
+	if (!xa_load(&work_ctx->table->devices, work_ctx->sf_index))
+		mlx5_sf_dev_add(work_ctx->table->dev, work_ctx->sf_index,
+				work_ctx->event.function_id, work_ctx->event.sw_function_id);
+	/* There is a race where SF got inactive after the query
+	 * above. e.g.: the query returns that the state of the
+	 * SF is active, and after that the eswitch manager set it to
+	 * inactive.
+	 * This case cannot be managed in SW, since the probing of the
+	 * SF is on one system, and the inactivation is on a different
+	 * system.
+	 * If the inactive is done after the SF perform init_hca(),
+	 * the SF will fully probe and then removed. If it was
+	 * done before init_hca(), the SF probe will fail.
+	 */
+out:
+	kfree(work_ctx);
+}
+
+/* In case SFs are generated externally, probe active SFs */
+static void mlx5_sf_dev_queue_active_works(struct work_struct *_work)
+{
+	struct mlx5_sf_dev_table *table = container_of(_work, struct mlx5_sf_dev_table, work);
 	u32 out[MLX5_ST_SZ_DW(query_vhca_state_out)] = {};
+	struct mlx5_sf_dev_active_work_ctx *work_ctx;
 	struct mlx5_core_dev *dev = table->dev;
 	u16 max_functions;
 	u16 function_id;
 	u16 sw_func_id;
 	int err = 0;
+	int wq_idx;
 	u8 state;
 	int i;
 
@@ -249,27 +275,22 @@ static void mlx5_sf_dev_add_active_work(struct work_struct *work)
 			continue;
 
 		sw_func_id = MLX5_GET(query_vhca_state_out, out, vhca_state_context.sw_function_id);
-		mutex_lock(&table->table_lock);
-		/* Don't probe device which is already probe */
-		if (!xa_load(&table->devices, i))
-			mlx5_sf_dev_add(dev, i, function_id, sw_func_id);
-		/* There is a race where SF got inactive after the query
-		 * above. e.g.: the query returns that the state of the
-		 * SF is active, and after that the eswitch manager set it to
-		 * inactive.
-		 * This case cannot be managed in SW, since the probing of the
-		 * SF is on one system, and the inactivation is on a different
-		 * system.
-		 * If the inactive is done after the SF perform init_hca(),
-		 * the SF will fully probe and then removed. If it was
-		 * done before init_hca(), the SF probe will fail.
-		 */
-		mutex_unlock(&table->table_lock);
+		work_ctx = kzalloc(sizeof(*work_ctx), GFP_KERNEL);
+		if (!work_ctx)
+			return;
+
+		INIT_WORK(&work_ctx->work, &mlx5_sf_dev_add_active_work);
+		work_ctx->event.function_id = function_id;
+		work_ctx->event.sw_function_id = sw_func_id;
+		work_ctx->table = table;
+		work_ctx->sf_index = i;
+		wq_idx = work_ctx->event.function_id % MLX5_DEV_MAX_WQS;
+		mlx5_vhca_events_work_enqueue(dev, wq_idx, &work_ctx->work);
 	}
 }
 
 /* In case SFs are generated externally, probe active SFs */
-static int mlx5_sf_dev_queue_active_work(struct mlx5_sf_dev_table *table)
+static int mlx5_sf_dev_create_active_works(struct mlx5_sf_dev_table *table)
 {
 	if (MLX5_CAP_GEN(table->dev, eswitch_manager))
 		return 0; /* the table is local */
@@ -280,12 +301,12 @@ static int mlx5_sf_dev_queue_active_work(struct mlx5_sf_dev_table *table)
 	table->active_wq = create_singlethread_workqueue("mlx5_active_sf");
 	if (!table->active_wq)
 		return -ENOMEM;
-	INIT_WORK(&table->work, &mlx5_sf_dev_add_active_work);
+	INIT_WORK(&table->work, &mlx5_sf_dev_queue_active_works);
 	queue_work(table->active_wq, &table->work);
 	return 0;
 }
 
-static void mlx5_sf_dev_destroy_active_work(struct mlx5_sf_dev_table *table)
+static void mlx5_sf_dev_destroy_active_works(struct mlx5_sf_dev_table *table)
 {
 	if (table->active_wq) {
 		table->stop_active_wq = true;
@@ -296,7 +317,6 @@ static void mlx5_sf_dev_destroy_active_work(struct mlx5_sf_dev_table *table)
 void mlx5_sf_dev_table_create(struct mlx5_core_dev *dev)
 {
 	struct mlx5_sf_dev_table *table;
-	unsigned int max_sfs;
 	int err;
 
 	if (!mlx5_sf_dev_supported(dev))
@@ -310,37 +330,30 @@ void mlx5_sf_dev_table_create(struct mlx5_core_dev *dev)
 
 	table->nb.notifier_call = mlx5_sf_dev_state_change_handler;
 	table->dev = dev;
-	if (MLX5_CAP_GEN(dev, max_num_sf))
-		max_sfs = MLX5_CAP_GEN(dev, max_num_sf);
-	else
-		max_sfs = 1 << MLX5_CAP_GEN(dev, log_max_sf);
 	table->sf_bar_length = 1 << (MLX5_CAP_GEN(dev, log_min_sf_size) + 12);
 	table->base_address = pci_resource_start(dev->pdev, 2);
-	table->max_sfs = max_sfs;
 	xa_init(&table->devices);
-	mutex_init(&table->table_lock);
 	dev->priv.sf_dev_table = table;
 
 	err = mlx5_vhca_event_notifier_register(dev, &table->nb);
 	if (err)
 		goto vhca_err;
 
-	err = mlx5_sf_dev_queue_active_work(table);
+	err = mlx5_sf_dev_create_active_works(table);
 	if (err)
 		goto add_active_err;
 
 	err = mlx5_sf_dev_vhca_arm_all(table);
 	if (err)
 		goto arm_err;
-	mlx5_core_dbg(dev, "SF DEV: max sf devices=%d\n", max_sfs);
 	return;
 
 arm_err:
-	mlx5_sf_dev_destroy_active_work(table);
+	mlx5_sf_dev_destroy_active_works(table);
 add_active_err:
 	mlx5_vhca_event_notifier_unregister(dev, &table->nb);
+	mlx5_vhca_event_work_queues_flush(dev);
 vhca_err:
-	table->max_sfs = 0;
 	kfree(table);
 	dev->priv.sf_dev_table = NULL;
 table_err:
@@ -365,9 +378,9 @@ void mlx5_sf_dev_table_destroy(struct mlx5_core_dev *dev)
 	if (!table)
 		return;
 
-	mlx5_sf_dev_destroy_active_work(table);
+	mlx5_sf_dev_destroy_active_works(table);
 	mlx5_vhca_event_notifier_unregister(dev, &table->nb);
-	mutex_destroy(&table->table_lock);
+	mlx5_vhca_event_work_queues_flush(dev);
 
 	/* Now that event handler is not running, it is safe to destroy
 	 * the sf device without race.
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.h b/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.h
index 2a66a427ef..b99131e95e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/dev.h
@@ -19,6 +19,12 @@ struct mlx5_sf_dev {
 	u16 fn_id;
 };
 
+struct mlx5_sf_peer_devlink_event_ctx {
+	u16 fn_id;
+	struct devlink *devlink;
+	int err;
+};
+
 void mlx5_sf_dev_table_create(struct mlx5_core_dev *dev);
 void mlx5_sf_dev_table_destroy(struct mlx5_core_dev *dev);
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/driver.c b/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/driver.c
index 8fe82f1191..169c2c68ed 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/driver.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sf/dev/driver.c
@@ -8,6 +8,20 @@
 #include "dev.h"
 #include "devlink.h"
 
+static int mlx5_core_peer_devlink_set(struct mlx5_sf_dev *sf_dev, struct devlink *devlink)
+{
+	struct mlx5_sf_peer_devlink_event_ctx event_ctx = {
+		.fn_id = sf_dev->fn_id,
+		.devlink = devlink,
+	};
+	int ret;
+
+	ret = mlx5_blocking_notifier_call_chain(sf_dev->parent_mdev,
+						MLX5_DRIVER_EVENT_SF_PEER_DEVLINK,
+						&event_ctx);
+	return ret == NOTIFY_OK ? event_ctx.err : 0;
+}
+
 static int mlx5_sf_dev_probe(struct auxiliary_device *adev, const struct auxiliary_device_id *id)
 {
 	struct mlx5_sf_dev *sf_dev = container_of(adev, struct mlx5_sf_dev, adev);
@@ -54,9 +68,21 @@ static int mlx5_sf_dev_probe(struct auxiliary_device *adev, const struct auxilia
 		mlx5_core_warn(mdev, "mlx5_init_one err=%d\n", err);
 		goto init_one_err;
 	}
+
+	err = mlx5_core_peer_devlink_set(sf_dev, devlink);
+	if (err) {
+		mlx5_core_warn(mdev, "mlx5_core_peer_devlink_set err=%d\n", err);
+		goto peer_devlink_set_err;
+	}
+
 	devlink_register(devlink);
 	return 0;
 
+peer_devlink_set_err:
+	if (mlx5_dev_is_lightweight(sf_dev->mdev))
+		mlx5_uninit_one_light(sf_dev->mdev);
+	else
+		mlx5_uninit_one(sf_dev->mdev);
 init_one_err:
 	iounmap(mdev->iseg);
 remap_err:
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sf/devlink.c b/drivers/net/ethernet/mellanox/mlx5/core/sf/devlink.c
index e34a8f88c5..6c11e075ca 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sf/devlink.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sf/devlink.c
@@ -20,43 +20,36 @@ struct mlx5_sf {
 	u16 hw_state;
 };
 
+static void *mlx5_sf_by_dl_port(struct devlink_port *dl_port)
+{
+	struct mlx5_devlink_port *mlx5_dl_port = mlx5_devlink_port_get(dl_port);
+
+	return container_of(mlx5_dl_port, struct mlx5_sf, dl_port);
+}
+
 struct mlx5_sf_table {
 	struct mlx5_core_dev *dev; /* To refer from notifier context. */
-	struct xarray port_indices; /* port index based lookup. */
-	refcount_t refcount;
-	struct completion disable_complete;
+	struct xarray function_ids; /* function id based lookup. */
 	struct mutex sf_state_lock; /* Serializes sf state among user cmds & vhca event handler. */
 	struct notifier_block esw_nb;
 	struct notifier_block vhca_nb;
+	struct notifier_block mdev_nb;
 };
 
 static struct mlx5_sf *
-mlx5_sf_lookup_by_index(struct mlx5_sf_table *table, unsigned int port_index)
-{
-	return xa_load(&table->port_indices, port_index);
-}
-
-static struct mlx5_sf *
 mlx5_sf_lookup_by_function_id(struct mlx5_sf_table *table, unsigned int fn_id)
 {
-	unsigned long index;
-	struct mlx5_sf *sf;
-
-	xa_for_each(&table->port_indices, index, sf) {
-		if (sf->hw_fn_id == fn_id)
-			return sf;
-	}
-	return NULL;
+	return xa_load(&table->function_ids, fn_id);
 }
 
-static int mlx5_sf_id_insert(struct mlx5_sf_table *table, struct mlx5_sf *sf)
+static int mlx5_sf_function_id_insert(struct mlx5_sf_table *table, struct mlx5_sf *sf)
 {
-	return xa_insert(&table->port_indices, sf->port_index, sf, GFP_KERNEL);
+	return xa_insert(&table->function_ids, sf->hw_fn_id, sf, GFP_KERNEL);
 }
 
-static void mlx5_sf_id_erase(struct mlx5_sf_table *table, struct mlx5_sf *sf)
+static void mlx5_sf_function_id_erase(struct mlx5_sf_table *table, struct mlx5_sf *sf)
 {
-	xa_erase(&table->port_indices, sf->port_index);
+	xa_erase(&table->function_ids, sf->hw_fn_id);
 }
 
 static struct mlx5_sf *
@@ -93,7 +86,7 @@ mlx5_sf_alloc(struct mlx5_sf_table *table, struct mlx5_eswitch *esw,
 	sf->hw_state = MLX5_VHCA_STATE_ALLOCATED;
 	sf->controller = controller;
 
-	err = mlx5_sf_id_insert(table, sf);
+	err = mlx5_sf_function_id_insert(table, sf);
 	if (err)
 		goto insert_err;
 
@@ -111,28 +104,11 @@ id_err:
 
 static void mlx5_sf_free(struct mlx5_sf_table *table, struct mlx5_sf *sf)
 {
-	mlx5_sf_id_erase(table, sf);
 	mlx5_sf_hw_table_sf_free(table->dev, sf->controller, sf->id);
 	trace_mlx5_sf_free(table->dev, sf->port_index, sf->controller, sf->hw_fn_id);
 	kfree(sf);
 }
 
-static struct mlx5_sf_table *mlx5_sf_table_try_get(struct mlx5_core_dev *dev)
-{
-	struct mlx5_sf_table *table = dev->priv.sf_table;
-
-	if (!table)
-		return NULL;
-
-	return refcount_inc_not_zero(&table->refcount) ? table : NULL;
-}
-
-static void mlx5_sf_table_put(struct mlx5_sf_table *table)
-{
-	if (refcount_dec_and_test(&table->refcount))
-		complete(&table->disable_complete);
-}
-
 static enum devlink_port_fn_state mlx5_sf_to_devlink_state(u8 hw_state)
 {
 	switch (hw_state) {
@@ -172,26 +148,14 @@ int mlx5_devlink_sf_port_fn_state_get(struct devlink_port *dl_port,
 				      struct netlink_ext_ack *extack)
 {
 	struct mlx5_core_dev *dev = devlink_priv(dl_port->devlink);
-	struct mlx5_sf_table *table;
-	struct mlx5_sf *sf;
-	int err = 0;
-
-	table = mlx5_sf_table_try_get(dev);
-	if (!table)
-		return -EOPNOTSUPP;
+	struct mlx5_sf_table *table = dev->priv.sf_table;
+	struct mlx5_sf *sf = mlx5_sf_by_dl_port(dl_port);
 
-	sf = mlx5_sf_lookup_by_index(table, dl_port->index);
-	if (!sf) {
-		err = -EOPNOTSUPP;
-		goto sf_err;
-	}
 	mutex_lock(&table->sf_state_lock);
 	*state = mlx5_sf_to_devlink_state(sf->hw_state);
 	*opstate = mlx5_sf_to_devlink_opstate(sf->hw_state);
 	mutex_unlock(&table->sf_state_lock);
-sf_err:
-	mlx5_sf_table_put(table);
-	return err;
+	return 0;
 }
 
 static int mlx5_sf_activate(struct mlx5_core_dev *dev, struct mlx5_sf *sf,
@@ -257,26 +221,10 @@ int mlx5_devlink_sf_port_fn_state_set(struct devlink_port *dl_port,
 				      struct netlink_ext_ack *extack)
 {
 	struct mlx5_core_dev *dev = devlink_priv(dl_port->devlink);
-	struct mlx5_sf_table *table;
-	struct mlx5_sf *sf;
-	int err;
-
-	table = mlx5_sf_table_try_get(dev);
-	if (!table) {
-		NL_SET_ERR_MSG_MOD(extack,
-				   "Port state set is only supported in eswitch switchdev mode or SF ports are disabled.");
-		return -EOPNOTSUPP;
-	}
-	sf = mlx5_sf_lookup_by_index(table, dl_port->index);
-	if (!sf) {
-		err = -ENODEV;
-		goto out;
-	}
+	struct mlx5_sf_table *table = dev->priv.sf_table;
+	struct mlx5_sf *sf = mlx5_sf_by_dl_port(dl_port);
 
-	err = mlx5_sf_state_set(dev, table, sf, state, extack);
-out:
-	mlx5_sf_table_put(table);
-	return err;
+	return mlx5_sf_state_set(dev, table, sf, state, extack);
 }
 
 static int mlx5_sf_add(struct mlx5_core_dev *dev, struct mlx5_sf_table *table,
@@ -335,32 +283,45 @@ mlx5_sf_new_check_attr(struct mlx5_core_dev *dev, const struct devlink_port_new_
 	return 0;
 }
 
+static bool mlx5_sf_table_supported(const struct mlx5_core_dev *dev)
+{
+	return dev->priv.eswitch && MLX5_ESWITCH_MANAGER(dev) &&
+	       mlx5_sf_hw_table_supported(dev);
+}
+
 int mlx5_devlink_sf_port_new(struct devlink *devlink,
 			     const struct devlink_port_new_attrs *new_attr,
 			     struct netlink_ext_ack *extack,
 			     struct devlink_port **dl_port)
 {
 	struct mlx5_core_dev *dev = devlink_priv(devlink);
-	struct mlx5_sf_table *table;
+	struct mlx5_sf_table *table = dev->priv.sf_table;
 	int err;
 
 	err = mlx5_sf_new_check_attr(dev, new_attr, extack);
 	if (err)
 		return err;
 
-	table = mlx5_sf_table_try_get(dev);
-	if (!table) {
+	if (!mlx5_sf_table_supported(dev)) {
+		NL_SET_ERR_MSG_MOD(extack, "SF ports are not supported.");
+		return -EOPNOTSUPP;
+	}
+
+	if (!is_mdev_switchdev_mode(dev)) {
 		NL_SET_ERR_MSG_MOD(extack,
-				   "Port add is only supported in eswitch switchdev mode or SF ports are disabled.");
+				   "SF ports are only supported in eswitch switchdev mode.");
 		return -EOPNOTSUPP;
 	}
-	err = mlx5_sf_add(dev, table, new_attr, extack, dl_port);
-	mlx5_sf_table_put(table);
-	return err;
+
+	return mlx5_sf_add(dev, table, new_attr, extack, dl_port);
 }
 
 static void mlx5_sf_dealloc(struct mlx5_sf_table *table, struct mlx5_sf *sf)
 {
+	mutex_lock(&table->sf_state_lock);
+
+	mlx5_sf_function_id_erase(table, sf);
+
 	if (sf->hw_state == MLX5_VHCA_STATE_ALLOCATED) {
 		mlx5_sf_free(table, sf);
 	} else if (mlx5_sf_is_active(sf)) {
@@ -376,6 +337,16 @@ static void mlx5_sf_dealloc(struct mlx5_sf_table *table, struct mlx5_sf *sf)
 		mlx5_sf_hw_table_sf_deferred_free(table->dev, sf->controller, sf->id);
 		kfree(sf);
 	}
+
+	mutex_unlock(&table->sf_state_lock);
+}
+
+static void mlx5_sf_del(struct mlx5_sf_table *table, struct mlx5_sf *sf)
+{
+	struct mlx5_eswitch *esw = table->dev->priv.eswitch;
+
+	mlx5_eswitch_unload_sf_vport(esw, sf->hw_fn_id);
+	mlx5_sf_dealloc(table, sf);
 }
 
 int mlx5_devlink_sf_port_del(struct devlink *devlink,
@@ -383,32 +354,11 @@ int mlx5_devlink_sf_port_del(struct devlink *devlink,
 			     struct netlink_ext_ack *extack)
 {
 	struct mlx5_core_dev *dev = devlink_priv(devlink);
-	struct mlx5_eswitch *esw = dev->priv.eswitch;
-	struct mlx5_sf_table *table;
-	struct mlx5_sf *sf;
-	int err = 0;
-
-	table = mlx5_sf_table_try_get(dev);
-	if (!table) {
-		NL_SET_ERR_MSG_MOD(extack,
-				   "Port del is only supported in eswitch switchdev mode or SF ports are disabled.");
-		return -EOPNOTSUPP;
-	}
-	sf = mlx5_sf_lookup_by_index(table, dl_port->index);
-	if (!sf) {
-		err = -ENODEV;
-		goto sf_err;
-	}
-
-	mlx5_eswitch_unload_sf_vport(esw, sf->hw_fn_id);
-	mlx5_sf_id_erase(table, sf);
+	struct mlx5_sf_table *table = dev->priv.sf_table;
+	struct mlx5_sf *sf = mlx5_sf_by_dl_port(dl_port);
 
-	mutex_lock(&table->sf_state_lock);
-	mlx5_sf_dealloc(table, sf);
-	mutex_unlock(&table->sf_state_lock);
-sf_err:
-	mlx5_sf_table_put(table);
-	return err;
+	mlx5_sf_del(table, sf);
+	return 0;
 }
 
 static bool mlx5_sf_state_update_check(const struct mlx5_sf *sf, u8 new_state)
@@ -433,14 +383,10 @@ static int mlx5_sf_vhca_event(struct notifier_block *nb, unsigned long opcode, v
 	bool update = false;
 	struct mlx5_sf *sf;
 
-	table = mlx5_sf_table_try_get(table->dev);
-	if (!table)
-		return 0;
-
 	mutex_lock(&table->sf_state_lock);
 	sf = mlx5_sf_lookup_by_function_id(table, event->function_id);
 	if (!sf)
-		goto sf_err;
+		goto unlock;
 
 	/* When driver is attached or detached to a function, an event
 	 * notifies such state change.
@@ -450,46 +396,18 @@ static int mlx5_sf_vhca_event(struct notifier_block *nb, unsigned long opcode, v
 		sf->hw_state = event->new_vhca_state;
 	trace_mlx5_sf_update_state(table->dev, sf->port_index, sf->controller,
 				   sf->hw_fn_id, sf->hw_state);
-sf_err:
+unlock:
 	mutex_unlock(&table->sf_state_lock);
-	mlx5_sf_table_put(table);
 	return 0;
 }
 
-static void mlx5_sf_table_enable(struct mlx5_sf_table *table)
-{
-	init_completion(&table->disable_complete);
-	refcount_set(&table->refcount, 1);
-}
-
-static void mlx5_sf_deactivate_all(struct mlx5_sf_table *table)
+static void mlx5_sf_del_all(struct mlx5_sf_table *table)
 {
-	struct mlx5_eswitch *esw = table->dev->priv.eswitch;
 	unsigned long index;
 	struct mlx5_sf *sf;
 
-	/* At this point, no new user commands can start and no vhca event can
-	 * arrive. It is safe to destroy all user created SFs.
-	 */
-	xa_for_each(&table->port_indices, index, sf) {
-		mlx5_eswitch_unload_sf_vport(esw, sf->hw_fn_id);
-		mlx5_sf_id_erase(table, sf);
-		mlx5_sf_dealloc(table, sf);
-	}
-}
-
-static void mlx5_sf_table_disable(struct mlx5_sf_table *table)
-{
-	if (!refcount_read(&table->refcount))
-		return;
-
-	/* Balances with refcount_set; drop the reference so that new user cmd cannot start
-	 * and new vhca event handler cannot run.
-	 */
-	mlx5_sf_table_put(table);
-	wait_for_completion(&table->disable_complete);
-
-	mlx5_sf_deactivate_all(table);
+	xa_for_each(&table->function_ids, index, sf)
+		mlx5_sf_del(table, sf);
 }
 
 static int mlx5_sf_esw_event(struct notifier_block *nb, unsigned long event, void *data)
@@ -498,11 +416,8 @@ static int mlx5_sf_esw_event(struct notifier_block *nb, unsigned long event, voi
 	const struct mlx5_esw_event_info *mode = data;
 
 	switch (mode->new_mode) {
-	case MLX5_ESWITCH_OFFLOADS:
-		mlx5_sf_table_enable(table);
-		break;
 	case MLX5_ESWITCH_LEGACY:
-		mlx5_sf_table_disable(table);
+		mlx5_sf_del_all(table);
 		break;
 	default:
 		break;
@@ -511,10 +426,29 @@ static int mlx5_sf_esw_event(struct notifier_block *nb, unsigned long event, voi
 	return 0;
 }
 
-static bool mlx5_sf_table_supported(const struct mlx5_core_dev *dev)
+static int mlx5_sf_mdev_event(struct notifier_block *nb, unsigned long event, void *data)
 {
-	return dev->priv.eswitch && MLX5_ESWITCH_MANAGER(dev) &&
-	       mlx5_sf_hw_table_supported(dev);
+	struct mlx5_sf_table *table = container_of(nb, struct mlx5_sf_table, mdev_nb);
+	struct mlx5_sf_peer_devlink_event_ctx *event_ctx = data;
+	int ret = NOTIFY_DONE;
+	struct mlx5_sf *sf;
+
+	if (event != MLX5_DRIVER_EVENT_SF_PEER_DEVLINK)
+		return NOTIFY_DONE;
+
+
+	mutex_lock(&table->sf_state_lock);
+	sf = mlx5_sf_lookup_by_function_id(table, event_ctx->fn_id);
+	if (!sf)
+		goto out;
+
+	event_ctx->err = devl_port_fn_devlink_set(&sf->dl_port.dl_port,
+						  event_ctx->devlink);
+
+	ret = NOTIFY_OK;
+out:
+	mutex_unlock(&table->sf_state_lock);
+	return ret;
 }
 
 int mlx5_sf_table_init(struct mlx5_core_dev *dev)
@@ -531,9 +465,8 @@ int mlx5_sf_table_init(struct mlx5_core_dev *dev)
 
 	mutex_init(&table->sf_state_lock);
 	table->dev = dev;
-	xa_init(&table->port_indices);
+	xa_init(&table->function_ids);
 	dev->priv.sf_table = table;
-	refcount_set(&table->refcount, 0);
 	table->esw_nb.notifier_call = mlx5_sf_esw_event;
 	err = mlx5_esw_event_notifier_register(dev->priv.eswitch, &table->esw_nb);
 	if (err)
@@ -544,6 +477,9 @@ int mlx5_sf_table_init(struct mlx5_core_dev *dev)
 	if (err)
 		goto vhca_err;
 
+	table->mdev_nb.notifier_call = mlx5_sf_mdev_event;
+	mlx5_blocking_notifier_register(dev, &table->mdev_nb);
+
 	return 0;
 
 vhca_err:
@@ -562,10 +498,10 @@ void mlx5_sf_table_cleanup(struct mlx5_core_dev *dev)
 	if (!table)
 		return;
 
+	mlx5_blocking_notifier_unregister(dev, &table->mdev_nb);
 	mlx5_vhca_event_notifier_unregister(table->dev, &table->vhca_nb);
 	mlx5_esw_event_notifier_unregister(dev->priv.eswitch, &table->esw_nb);
-	WARN_ON(refcount_read(&table->refcount));
 	mutex_destroy(&table->sf_state_lock);
-	WARN_ON(!xa_empty(&table->port_indices));
+	WARN_ON(!xa_empty(&table->function_ids));
 	kfree(table);
 }
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.c b/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.c
index d908fba968..cda01ba441 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.c
@@ -21,6 +21,15 @@ struct mlx5_vhca_event_work {
 	struct mlx5_vhca_state_event event;
 };
 
+struct mlx5_vhca_event_handler {
+	struct workqueue_struct *wq;
+};
+
+struct mlx5_vhca_events {
+	struct mlx5_core_dev *dev;
+	struct mlx5_vhca_event_handler handler[MLX5_DEV_MAX_WQS];
+};
+
 int mlx5_cmd_query_vhca_state(struct mlx5_core_dev *dev, u16 function_id, u32 *out, u32 outlen)
 {
 	u32 in[MLX5_ST_SZ_DW(query_vhca_state_in)] = {};
@@ -99,6 +108,11 @@ static void mlx5_vhca_state_work_handler(struct work_struct *_work)
 	kfree(work);
 }
 
+void mlx5_vhca_events_work_enqueue(struct mlx5_core_dev *dev, int idx, struct work_struct *work)
+{
+	queue_work(dev->priv.vhca_events->handler[idx].wq, work);
+}
+
 static int
 mlx5_vhca_state_change_notifier(struct notifier_block *nb, unsigned long type, void *data)
 {
@@ -106,6 +120,7 @@ mlx5_vhca_state_change_notifier(struct notifier_block *nb, unsigned long type, v
 				mlx5_nb_cof(nb, struct mlx5_vhca_state_notifier, nb);
 	struct mlx5_vhca_event_work *work;
 	struct mlx5_eqe *eqe = data;
+	int wq_idx;
 
 	work = kzalloc(sizeof(*work), GFP_ATOMIC);
 	if (!work)
@@ -113,7 +128,8 @@ mlx5_vhca_state_change_notifier(struct notifier_block *nb, unsigned long type, v
 	INIT_WORK(&work->work, &mlx5_vhca_state_work_handler);
 	work->notifier = notifier;
 	work->event.function_id = be16_to_cpu(eqe->data.vhca_state.function_id);
-	mlx5_events_work_enqueue(notifier->dev, &work->work);
+	wq_idx = work->event.function_id % MLX5_DEV_MAX_WQS;
+	mlx5_vhca_events_work_enqueue(notifier->dev, wq_idx, &work->work);
 	return NOTIFY_OK;
 }
 
@@ -132,28 +148,75 @@ void mlx5_vhca_state_cap_handle(struct mlx5_core_dev *dev, void *set_hca_cap)
 int mlx5_vhca_event_init(struct mlx5_core_dev *dev)
 {
 	struct mlx5_vhca_state_notifier *notifier;
+	char wq_name[MLX5_CMD_WQ_MAX_NAME];
+	struct mlx5_vhca_events *events;
+	int err, i;
 
 	if (!mlx5_vhca_event_supported(dev))
 		return 0;
 
-	notifier = kzalloc(sizeof(*notifier), GFP_KERNEL);
-	if (!notifier)
+	events = kzalloc(sizeof(*events), GFP_KERNEL);
+	if (!events)
 		return -ENOMEM;
 
+	events->dev = dev;
+	dev->priv.vhca_events = events;
+	for (i = 0; i < MLX5_DEV_MAX_WQS; i++) {
+		snprintf(wq_name, MLX5_CMD_WQ_MAX_NAME, "mlx5_vhca_event%d", i);
+		events->handler[i].wq = create_singlethread_workqueue(wq_name);
+		if (!events->handler[i].wq) {
+			err = -ENOMEM;
+			goto err_create_wq;
+		}
+	}
+
+	notifier = kzalloc(sizeof(*notifier), GFP_KERNEL);
+	if (!notifier) {
+		err = -ENOMEM;
+		goto err_notifier;
+	}
+
 	dev->priv.vhca_state_notifier = notifier;
 	notifier->dev = dev;
 	BLOCKING_INIT_NOTIFIER_HEAD(&notifier->n_head);
 	MLX5_NB_INIT(&notifier->nb, mlx5_vhca_state_change_notifier, VHCA_STATE_CHANGE);
 	return 0;
+
+err_notifier:
+err_create_wq:
+	for (--i; i >= 0; i--)
+		destroy_workqueue(events->handler[i].wq);
+	kfree(events);
+	return err;
+}
+
+void mlx5_vhca_event_work_queues_flush(struct mlx5_core_dev *dev)
+{
+	struct mlx5_vhca_events *vhca_events;
+	int i;
+
+	if (!mlx5_vhca_event_supported(dev))
+		return;
+
+	vhca_events = dev->priv.vhca_events;
+	for (i = 0; i < MLX5_DEV_MAX_WQS; i++)
+		flush_workqueue(vhca_events->handler[i].wq);
 }
 
 void mlx5_vhca_event_cleanup(struct mlx5_core_dev *dev)
 {
+	struct mlx5_vhca_events *vhca_events;
+	int i;
+
 	if (!mlx5_vhca_event_supported(dev))
 		return;
 
 	kfree(dev->priv.vhca_state_notifier);
 	dev->priv.vhca_state_notifier = NULL;
+	vhca_events = dev->priv.vhca_events;
+	for (i = 0; i < MLX5_DEV_MAX_WQS; i++)
+		destroy_workqueue(vhca_events->handler[i].wq);
+	kvfree(vhca_events);
 }
 
 void mlx5_vhca_event_start(struct mlx5_core_dev *dev)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.h b/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.h
index 013cdfe906..1725ba64f8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sf/vhca_event.h
@@ -28,6 +28,9 @@ int mlx5_modify_vhca_sw_id(struct mlx5_core_dev *dev, u16 function_id, u32 sw_fn
 int mlx5_vhca_event_arm(struct mlx5_core_dev *dev, u16 function_id);
 int mlx5_cmd_query_vhca_state(struct mlx5_core_dev *dev, u16 function_id,
 			      u32 *out, u32 outlen);
+void mlx5_vhca_events_work_enqueue(struct mlx5_core_dev *dev, int idx, struct work_struct *work);
+void mlx5_vhca_event_work_queues_flush(struct mlx5_core_dev *dev);
+
 #else
 
 static inline void mlx5_vhca_state_cap_handle(struct mlx5_core_dev *dev, void *set_hca_cap)