[PATCH net-next V2 0/7] E-Switch vport sharing & delegation

[PATCH net-next V2 0/7] E-Switch vport sharing & delegation

[Saeed Mahameed]

From: Saeed Mahameed <saeedm@nvidia.com>

v1->v2:
 - rename goto labels after targets. Jakub.

An mlx5 E-Switch FDB table can manage vports belonging to other sibling
physical functions, such as ECPF (ARM embedded cores) and Host PF (x86).
This enables a single source of truth for SDN software to manage network
pipelines from one host. While such functionality already exists in mlx5,
it is currently limited by static vport allocation,
meaning the number of vports shared between multi-host functions
must be known pre-boot.

This patchset enables delegated/external vports to be discovered
dynamically when switchdev mode is enabled, leveraging new firmware
capabilities for dynamic vport creation.

Adjacent functions that delegate their SR-IOV VFs to sibling PFs, can be
dynamically discovered on the sibling PF's switchdev mode enabling,
after sriov was enabled on the originating PF, allowing for more
flexible and scalable management in multi-host and ECPF-to-host
scenarios.

The patchset consists of the following changes:

- Refactoring of ACL root namespace handling: The storage of vport ACL root
  namespaces is converted from a linear array to an xarray, allowing dynamic
  creation of ACLs per individual vport.
- Improvements for vhca_id to vport mapping.
- Dynamic querying and creation of delegated functions/vports.


Adithya Jayachandran (2):
  net/mlx5: E-Switch, Add support for adjacent functions vports
    discovery
  net/mlx5: E-switch, Set representor attributes for adjacent VFs

Saeed Mahameed (5):
  net/mlx5: FS, Convert vport acls root namespaces to xarray
  net/mlx5: E-Switch, Move vport acls root namespaces creation to
    eswitch
  net/mlx5: E-Switch, Create acls root namespace for adjacent vports
  net/mlx5: E-Switch, Register representors for adjacent vports
  net/mlx5: {DR,HWS}, Use the cached vhca_id for this device

 .../net/ethernet/mellanox/mlx5/core/Makefile  |   2 +-
 .../mellanox/mlx5/core/esw/adj_vport.c        | 209 ++++++++++++++++++
 .../mellanox/mlx5/core/esw/devlink_port.c     |  11 +-
 .../net/ethernet/mellanox/mlx5/core/eswitch.c | 131 ++++++++++-
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |  17 ++
 .../mellanox/mlx5/core/eswitch_offloads.c     |  37 +++-
 .../net/ethernet/mellanox/mlx5/core/fs_core.c | 183 ++++++++-------
 .../net/ethernet/mellanox/mlx5/core/fs_core.h |  18 +-
 .../mellanox/mlx5/core/steering/hws/cmd.c     |  34 +--
 .../mellanox/mlx5/core/steering/sws/dr_cmd.c  |  34 +--
 10 files changed, 500 insertions(+), 176 deletions(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c

-- 
2.50.1

[PATCH net-next V2 1/7] net/mlx5: FS, Convert vport acls root namespaces to xarray

[Saeed Mahameed]

From: Saeed Mahameed <saeedm@nvidia.com>

Before this patch it was a linear array and could only support a certain
number of vports, in the next patches, vport numbers are not bound to a
well known limit, thus convert acl root name space storage to xarray.

In addition create fs_core public API to add/remove vport acl namespaces
as it is the eswitch responsibility to create the vports and their
root name spaces for acls, in the next patch we will move
mlx5_fs_ingress_acls_{init,cleanup} to eswitch and will use
the individual mlx5_fs_vport_{egress,ingresS}_acl_ns_{add,remove}
APIs for dynamically create vports.

Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/fs_core.c | 169 ++++++++++++------
 .../net/ethernet/mellanox/mlx5/core/fs_core.h |  13 +-
 2 files changed, 123 insertions(+), 59 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
index d87392360dbd..7d9de2f65ce6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
@@ -2793,30 +2793,32 @@ struct mlx5_flow_namespace *mlx5_get_flow_namespace(struct mlx5_core_dev *dev,
 }
 EXPORT_SYMBOL(mlx5_get_flow_namespace);
 
+struct mlx5_vport_acl_root_ns {
+	u16 vport_idx;
+	struct mlx5_flow_root_namespace *root_ns;
+};
+
 struct mlx5_flow_namespace *
 mlx5_get_flow_vport_namespace(struct mlx5_core_dev *dev,
 			      enum mlx5_flow_namespace_type type, int vport_idx)
 {
 	struct mlx5_flow_steering *steering = dev->priv.steering;
+	struct mlx5_vport_acl_root_ns *vport_ns;
 
 	if (!steering)
 		return NULL;
 
 	switch (type) {
 	case MLX5_FLOW_NAMESPACE_ESW_EGRESS:
-		if (vport_idx >= steering->esw_egress_acl_vports)
-			return NULL;
-		if (steering->esw_egress_root_ns &&
-		    steering->esw_egress_root_ns[vport_idx])
-			return &steering->esw_egress_root_ns[vport_idx]->ns;
+		vport_ns = xa_load(&steering->esw_egress_root_ns, vport_idx);
+		if (vport_ns)
+			return &vport_ns->root_ns->ns;
 		else
 			return NULL;
 	case MLX5_FLOW_NAMESPACE_ESW_INGRESS:
-		if (vport_idx >= steering->esw_ingress_acl_vports)
-			return NULL;
-		if (steering->esw_ingress_root_ns &&
-		    steering->esw_ingress_root_ns[vport_idx])
-			return &steering->esw_ingress_root_ns[vport_idx]->ns;
+		vport_ns = xa_load(&steering->esw_ingress_root_ns, vport_idx);
+		if (vport_ns)
+			return &vport_ns->root_ns->ns;
 		else
 			return NULL;
 	case MLX5_FLOW_NAMESPACE_RDMA_TRANSPORT_RX:
@@ -3575,30 +3577,102 @@ static int init_fdb_root_ns(struct mlx5_flow_steering *steering)
 	return err;
 }
 
-static int init_egress_acl_root_ns(struct mlx5_flow_steering *steering, int vport)
+static void
+mlx5_fs_remove_vport_acl_root_ns(struct xarray *esw_acl_root_ns, u16 vport_idx)
+{
+	struct mlx5_vport_acl_root_ns *vport_ns;
+
+	vport_ns = xa_erase(esw_acl_root_ns, vport_idx);
+	if (vport_ns) {
+		cleanup_root_ns(vport_ns->root_ns);
+		kfree(vport_ns);
+	}
+}
+
+static int
+mlx5_fs_add_vport_acl_root_ns(struct mlx5_flow_steering *steering,
+			      struct xarray *esw_acl_root_ns,
+			      enum fs_flow_table_type table_type,
+			      u16 vport_idx)
 {
+	struct mlx5_vport_acl_root_ns *vport_ns;
 	struct fs_prio *prio;
+	int err;
+
+	/* sanity check, intended xarrays are used */
+	if (WARN_ON(esw_acl_root_ns != &steering->esw_egress_root_ns &&
+		    esw_acl_root_ns != &steering->esw_ingress_root_ns))
+		return -EINVAL;
 
-	steering->esw_egress_root_ns[vport] = create_root_ns(steering, FS_FT_ESW_EGRESS_ACL);
-	if (!steering->esw_egress_root_ns[vport])
+	if (table_type != FS_FT_ESW_EGRESS_ACL &&
+	    table_type != FS_FT_ESW_INGRESS_ACL) {
+		mlx5_core_err(steering->dev,
+			      "Invalid table type %d for egress/ingress ACLs\n",
+			      table_type);
+		return -EINVAL;
+	}
+
+	if (xa_load(esw_acl_root_ns, vport_idx))
+		return -EEXIST;
+
+	vport_ns = kzalloc(sizeof(*vport_ns), GFP_KERNEL);
+	if (!vport_ns)
 		return -ENOMEM;
 
+	vport_ns->root_ns = create_root_ns(steering, table_type);
+	if (!vport_ns->root_ns) {
+		err = -ENOMEM;
+		goto kfree_vport_ns;
+	}
+
 	/* create 1 prio*/
-	prio = fs_create_prio(&steering->esw_egress_root_ns[vport]->ns, 0, 1);
-	return PTR_ERR_OR_ZERO(prio);
+	prio = fs_create_prio(&vport_ns->root_ns->ns, 0, 1);
+	if (IS_ERR(prio)) {
+		err = PTR_ERR(prio);
+		goto cleanup_root_ns;
+	}
+
+	vport_ns->vport_idx = vport_idx;
+	err = xa_insert(esw_acl_root_ns, vport_idx, vport_ns, GFP_KERNEL);
+	if (err)
+		goto cleanup_root_ns;
+	return 0;
+
+cleanup_root_ns:
+	cleanup_root_ns(vport_ns->root_ns);
+kfree_vport_ns:
+	kfree(vport_ns);
+	return err;
 }
 
-static int init_ingress_acl_root_ns(struct mlx5_flow_steering *steering, int vport)
+int mlx5_fs_vport_egress_acl_ns_add(struct mlx5_flow_steering *steering,
+				    u16 vport_idx)
 {
-	struct fs_prio *prio;
+	return mlx5_fs_add_vport_acl_root_ns(steering,
+					     &steering->esw_egress_root_ns,
+					     FS_FT_ESW_EGRESS_ACL, vport_idx);
+}
 
-	steering->esw_ingress_root_ns[vport] = create_root_ns(steering, FS_FT_ESW_INGRESS_ACL);
-	if (!steering->esw_ingress_root_ns[vport])
-		return -ENOMEM;
+int mlx5_fs_vport_ingress_acl_ns_add(struct mlx5_flow_steering *steering,
+				     u16 vport_idx)
+{
+	return mlx5_fs_add_vport_acl_root_ns(steering,
+					     &steering->esw_ingress_root_ns,
+					     FS_FT_ESW_INGRESS_ACL, vport_idx);
+}
 
-	/* create 1 prio*/
-	prio = fs_create_prio(&steering->esw_ingress_root_ns[vport]->ns, 0, 1);
-	return PTR_ERR_OR_ZERO(prio);
+void mlx5_fs_vport_egress_acl_ns_remove(struct mlx5_flow_steering *steering,
+					int vport_idx)
+{
+	mlx5_fs_remove_vport_acl_root_ns(&steering->esw_egress_root_ns,
+					 vport_idx);
+}
+
+void mlx5_fs_vport_ingress_acl_ns_remove(struct mlx5_flow_steering *steering,
+					 int vport_idx)
+{
+	mlx5_fs_remove_vport_acl_root_ns(&steering->esw_ingress_root_ns,
+					 vport_idx);
 }
 
 int mlx5_fs_egress_acls_init(struct mlx5_core_dev *dev, int total_vports)
@@ -3607,15 +3681,10 @@ int mlx5_fs_egress_acls_init(struct mlx5_core_dev *dev, int total_vports)
 	int err;
 	int i;
 
-	steering->esw_egress_root_ns =
-			kcalloc(total_vports,
-				sizeof(*steering->esw_egress_root_ns),
-				GFP_KERNEL);
-	if (!steering->esw_egress_root_ns)
-		return -ENOMEM;
+	xa_init(&steering->esw_egress_root_ns);
 
 	for (i = 0; i < total_vports; i++) {
-		err = init_egress_acl_root_ns(steering, i);
+		err = mlx5_fs_vport_egress_acl_ns_add(steering, i);
 		if (err)
 			goto cleanup_root_ns;
 	}
@@ -3623,10 +3692,9 @@ int mlx5_fs_egress_acls_init(struct mlx5_core_dev *dev, int total_vports)
 	return 0;
 
 cleanup_root_ns:
-	for (i--; i >= 0; i--)
-		cleanup_root_ns(steering->esw_egress_root_ns[i]);
-	kfree(steering->esw_egress_root_ns);
-	steering->esw_egress_root_ns = NULL;
+	while (i--)
+		mlx5_fs_vport_egress_acl_ns_remove(steering, i);
+	xa_destroy(&steering->esw_egress_root_ns);
 	return err;
 }
 
@@ -3635,14 +3703,10 @@ void mlx5_fs_egress_acls_cleanup(struct mlx5_core_dev *dev)
 	struct mlx5_flow_steering *steering = dev->priv.steering;
 	int i;
 
-	if (!steering->esw_egress_root_ns)
-		return;
-
 	for (i = 0; i < steering->esw_egress_acl_vports; i++)
-		cleanup_root_ns(steering->esw_egress_root_ns[i]);
+		mlx5_fs_vport_egress_acl_ns_remove(steering, i);
 
-	kfree(steering->esw_egress_root_ns);
-	steering->esw_egress_root_ns = NULL;
+	xa_destroy(&steering->esw_egress_root_ns);
 }
 
 int mlx5_fs_ingress_acls_init(struct mlx5_core_dev *dev, int total_vports)
@@ -3651,15 +3715,10 @@ int mlx5_fs_ingress_acls_init(struct mlx5_core_dev *dev, int total_vports)
 	int err;
 	int i;
 
-	steering->esw_ingress_root_ns =
-			kcalloc(total_vports,
-				sizeof(*steering->esw_ingress_root_ns),
-				GFP_KERNEL);
-	if (!steering->esw_ingress_root_ns)
-		return -ENOMEM;
+	xa_init(&steering->esw_ingress_root_ns);
 
 	for (i = 0; i < total_vports; i++) {
-		err = init_ingress_acl_root_ns(steering, i);
+		err = mlx5_fs_vport_ingress_acl_ns_add(steering, i);
 		if (err)
 			goto cleanup_root_ns;
 	}
@@ -3667,10 +3726,10 @@ int mlx5_fs_ingress_acls_init(struct mlx5_core_dev *dev, int total_vports)
 	return 0;
 
 cleanup_root_ns:
-	for (i--; i >= 0; i--)
-		cleanup_root_ns(steering->esw_ingress_root_ns[i]);
-	kfree(steering->esw_ingress_root_ns);
-	steering->esw_ingress_root_ns = NULL;
+	while (i--)
+		mlx5_fs_vport_ingress_acl_ns_remove(steering, i);
+
+	xa_destroy(&steering->esw_ingress_root_ns);
 	return err;
 }
 
@@ -3679,14 +3738,10 @@ void mlx5_fs_ingress_acls_cleanup(struct mlx5_core_dev *dev)
 	struct mlx5_flow_steering *steering = dev->priv.steering;
 	int i;
 
-	if (!steering->esw_ingress_root_ns)
-		return;
-
 	for (i = 0; i < steering->esw_ingress_acl_vports; i++)
-		cleanup_root_ns(steering->esw_ingress_root_ns[i]);
+		mlx5_fs_vport_ingress_acl_ns_remove(steering, i);
 
-	kfree(steering->esw_ingress_root_ns);
-	steering->esw_ingress_root_ns = NULL;
+	xa_destroy(&steering->esw_ingress_root_ns);
 }
 
 u32 mlx5_fs_get_capabilities(struct mlx5_core_dev *dev, enum mlx5_flow_namespace_type type)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
index 500826229b0b..a7642d9fc118 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
@@ -151,8 +151,8 @@ struct mlx5_flow_steering {
 	struct mlx5_flow_root_namespace *root_ns;
 	struct mlx5_flow_root_namespace *fdb_root_ns;
 	struct mlx5_flow_namespace	**fdb_sub_ns;
-	struct mlx5_flow_root_namespace **esw_egress_root_ns;
-	struct mlx5_flow_root_namespace **esw_ingress_root_ns;
+	struct xarray			esw_egress_root_ns;
+	struct xarray			esw_ingress_root_ns;
 	struct mlx5_flow_root_namespace	*sniffer_tx_root_ns;
 	struct mlx5_flow_root_namespace	*sniffer_rx_root_ns;
 	struct mlx5_flow_root_namespace	*rdma_rx_root_ns;
@@ -383,6 +383,15 @@ void mlx5_fs_egress_acls_cleanup(struct mlx5_core_dev *dev);
 int mlx5_fs_ingress_acls_init(struct mlx5_core_dev *dev, int total_vports);
 void mlx5_fs_ingress_acls_cleanup(struct mlx5_core_dev *dev);
 
+int mlx5_fs_vport_egress_acl_ns_add(struct mlx5_flow_steering *steering,
+				    u16 vport_idx);
+int mlx5_fs_vport_ingress_acl_ns_add(struct mlx5_flow_steering *steering,
+				     u16 vport_idx);
+void mlx5_fs_vport_egress_acl_ns_remove(struct mlx5_flow_steering *steering,
+					int vport_idx);
+void mlx5_fs_vport_ingress_acl_ns_remove(struct mlx5_flow_steering *steering,
+					 int vport_idx);
+
 u32 mlx5_fs_get_capabilities(struct mlx5_core_dev *dev, enum mlx5_flow_namespace_type type);
 
 struct mlx5_flow_root_namespace *find_root(struct fs_node *node);
-- 
2.50.1

Re: [PATCH net-next V2 1/7] net/mlx5: FS, Convert vport acls root namespaces to xarray

[Simon Horman]

On Tue, Aug 26, 2025 at 09:45:10PM -0700, Saeed Mahameed wrote:
> From: Saeed Mahameed <saeedm@nvidia.com>
> 
> Before this patch it was a linear array and could only support a certain
> number of vports, in the next patches, vport numbers are not bound to a
> well known limit, thus convert acl root name space storage to xarray.
> 
> In addition create fs_core public API to add/remove vport acl namespaces
> as it is the eswitch responsibility to create the vports and their
> root name spaces for acls, in the next patch we will move
> mlx5_fs_ingress_acls_{init,cleanup} to eswitch and will use
> the individual mlx5_fs_vport_{egress,ingresS}_acl_ns_{add,remove}
> APIs for dynamically create vports.
> 
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Reviewed-by: Simon Horman <horms@kernel.org>

[PATCH net-next V2 2/7] net/mlx5: E-Switch, Move vport acls root namespaces creation to eswitch

[Saeed Mahameed]

From: Saeed Mahameed <saeedm@nvidia.com>

Move the loop that creates the vports ACLs root name spaces to eswitch,
since it is the eswitch responsibility to decide when and how many
vports ACLs root namespaces to create, in the next patches we will use
the fs_core vport ACL root namespace APIs to create/remove root ns
ACLs dynamically for dynamically created vports.

Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/eswitch.c | 73 ++++++++++++++++--
 .../net/ethernet/mellanox/mlx5/core/fs_core.c | 76 ++-----------------
 .../net/ethernet/mellanox/mlx5/core/fs_core.h |  7 --
 3 files changed, 72 insertions(+), 84 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index 9fe5a45124fd..900650a1a66e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -1439,19 +1439,76 @@ static void mlx5_esw_mode_change_notify(struct mlx5_eswitch *esw, u16 mode)
 	blocking_notifier_call_chain(&esw->n_head, 0, &info);
 }
 
+static int mlx5_esw_egress_acls_init(struct mlx5_core_dev *dev)
+{
+	struct mlx5_flow_steering *steering = dev->priv.steering;
+	int total_vports = mlx5_eswitch_get_total_vports(dev);
+	int err;
+	int i;
+
+	for (i = 0; i < total_vports; i++) {
+		err = mlx5_fs_vport_egress_acl_ns_add(steering, i);
+		if (err)
+			goto acl_ns_remove;
+	}
+	return 0;
+
+acl_ns_remove:
+	while (i--)
+		mlx5_fs_vport_egress_acl_ns_remove(steering, i);
+	return err;
+}
+
+static void mlx5_esw_egress_acls_cleanup(struct mlx5_core_dev *dev)
+{
+	struct mlx5_flow_steering *steering = dev->priv.steering;
+	int total_vports = mlx5_eswitch_get_total_vports(dev);
+	int i;
+
+	for (i = total_vports - 1; i >= 0; i--)
+		mlx5_fs_vport_egress_acl_ns_remove(steering, i);
+}
+
+static int mlx5_esw_ingress_acls_init(struct mlx5_core_dev *dev)
+{
+	struct mlx5_flow_steering *steering = dev->priv.steering;
+	int total_vports = mlx5_eswitch_get_total_vports(dev);
+	int err;
+	int i;
+
+	for (i = 0; i < total_vports; i++) {
+		err = mlx5_fs_vport_ingress_acl_ns_add(steering, i);
+		if (err)
+			goto acl_ns_remove;
+	}
+	return 0;
+
+acl_ns_remove:
+	while (i--)
+		mlx5_fs_vport_ingress_acl_ns_remove(steering, i);
+	return err;
+}
+
+static void mlx5_esw_ingress_acls_cleanup(struct mlx5_core_dev *dev)
+{
+	struct mlx5_flow_steering *steering = dev->priv.steering;
+	int total_vports = mlx5_eswitch_get_total_vports(dev);
+	int i;
+
+	for (i = total_vports - 1; i >= 0; i--)
+		mlx5_fs_vport_ingress_acl_ns_remove(steering, i);
+}
+
 static int mlx5_esw_acls_ns_init(struct mlx5_eswitch *esw)
 {
 	struct mlx5_core_dev *dev = esw->dev;
-	int total_vports;
 	int err;
 
 	if (esw->flags & MLX5_ESWITCH_VPORT_ACL_NS_CREATED)
 		return 0;
 
-	total_vports = mlx5_eswitch_get_total_vports(dev);
-
 	if (MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support)) {
-		err = mlx5_fs_egress_acls_init(dev, total_vports);
+		err = mlx5_esw_egress_acls_init(dev);
 		if (err)
 			return err;
 	} else {
@@ -1459,7 +1516,7 @@ static int mlx5_esw_acls_ns_init(struct mlx5_eswitch *esw)
 	}
 
 	if (MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support)) {
-		err = mlx5_fs_ingress_acls_init(dev, total_vports);
+		err = mlx5_esw_ingress_acls_init(dev);
 		if (err)
 			goto err;
 	} else {
@@ -1470,7 +1527,7 @@ static int mlx5_esw_acls_ns_init(struct mlx5_eswitch *esw)
 
 err:
 	if (MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
-		mlx5_fs_egress_acls_cleanup(dev);
+		mlx5_esw_egress_acls_cleanup(dev);
 	return err;
 }
 
@@ -1480,9 +1537,9 @@ static void mlx5_esw_acls_ns_cleanup(struct mlx5_eswitch *esw)
 
 	esw->flags &= ~MLX5_ESWITCH_VPORT_ACL_NS_CREATED;
 	if (MLX5_CAP_ESW_INGRESS_ACL(dev, ft_support))
-		mlx5_fs_ingress_acls_cleanup(dev);
+		mlx5_esw_ingress_acls_cleanup(dev);
 	if (MLX5_CAP_ESW_EGRESS_ACL(dev, ft_support))
-		mlx5_fs_egress_acls_cleanup(dev);
+		mlx5_esw_egress_acls_cleanup(dev);
 }
 
 /**
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
index 7d9de2f65ce6..e4d24e6f3903 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.c
@@ -3675,75 +3675,6 @@ void mlx5_fs_vport_ingress_acl_ns_remove(struct mlx5_flow_steering *steering,
 					 vport_idx);
 }
 
-int mlx5_fs_egress_acls_init(struct mlx5_core_dev *dev, int total_vports)
-{
-	struct mlx5_flow_steering *steering = dev->priv.steering;
-	int err;
-	int i;
-
-	xa_init(&steering->esw_egress_root_ns);
-
-	for (i = 0; i < total_vports; i++) {
-		err = mlx5_fs_vport_egress_acl_ns_add(steering, i);
-		if (err)
-			goto cleanup_root_ns;
-	}
-	steering->esw_egress_acl_vports = total_vports;
-	return 0;
-
-cleanup_root_ns:
-	while (i--)
-		mlx5_fs_vport_egress_acl_ns_remove(steering, i);
-	xa_destroy(&steering->esw_egress_root_ns);
-	return err;
-}
-
-void mlx5_fs_egress_acls_cleanup(struct mlx5_core_dev *dev)
-{
-	struct mlx5_flow_steering *steering = dev->priv.steering;
-	int i;
-
-	for (i = 0; i < steering->esw_egress_acl_vports; i++)
-		mlx5_fs_vport_egress_acl_ns_remove(steering, i);
-
-	xa_destroy(&steering->esw_egress_root_ns);
-}
-
-int mlx5_fs_ingress_acls_init(struct mlx5_core_dev *dev, int total_vports)
-{
-	struct mlx5_flow_steering *steering = dev->priv.steering;
-	int err;
-	int i;
-
-	xa_init(&steering->esw_ingress_root_ns);
-
-	for (i = 0; i < total_vports; i++) {
-		err = mlx5_fs_vport_ingress_acl_ns_add(steering, i);
-		if (err)
-			goto cleanup_root_ns;
-	}
-	steering->esw_ingress_acl_vports = total_vports;
-	return 0;
-
-cleanup_root_ns:
-	while (i--)
-		mlx5_fs_vport_ingress_acl_ns_remove(steering, i);
-
-	xa_destroy(&steering->esw_ingress_root_ns);
-	return err;
-}
-
-void mlx5_fs_ingress_acls_cleanup(struct mlx5_core_dev *dev)
-{
-	struct mlx5_flow_steering *steering = dev->priv.steering;
-	int i;
-
-	for (i = 0; i < steering->esw_ingress_acl_vports; i++)
-		mlx5_fs_vport_ingress_acl_ns_remove(steering, i);
-
-	xa_destroy(&steering->esw_ingress_root_ns);
-}
-
 u32 mlx5_fs_get_capabilities(struct mlx5_core_dev *dev, enum mlx5_flow_namespace_type type)
 {
 	struct mlx5_flow_root_namespace *root;
@@ -3874,6 +3805,11 @@ void mlx5_fs_core_cleanup(struct mlx5_core_dev *dev)
 {
 	struct mlx5_flow_steering *steering = dev->priv.steering;
 
+	WARN_ON(!xa_empty(&steering->esw_egress_root_ns));
+	WARN_ON(!xa_empty(&steering->esw_ingress_root_ns));
+	xa_destroy(&steering->esw_egress_root_ns);
+	xa_destroy(&steering->esw_ingress_root_ns);
+
 	cleanup_root_ns(steering->root_ns);
 	cleanup_fdb_root_ns(steering);
 	cleanup_root_ns(steering->port_sel_root_ns);
@@ -3964,6 +3900,8 @@ int mlx5_fs_core_init(struct mlx5_core_dev *dev)
 			goto err;
 	}
 
+	xa_init(&steering->esw_egress_root_ns);
+	xa_init(&steering->esw_ingress_root_ns);
 	return 0;
 
 err:
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
index a7642d9fc118..7877d9a2118d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_core.h
@@ -159,8 +159,6 @@ struct mlx5_flow_steering {
 	struct mlx5_flow_root_namespace	*rdma_tx_root_ns;
 	struct mlx5_flow_root_namespace	*egress_root_ns;
 	struct mlx5_flow_root_namespace	*port_sel_root_ns;
-	int esw_egress_acl_vports;
-	int esw_ingress_acl_vports;
 	struct mlx5_flow_root_namespace **rdma_transport_rx_root_ns;
 	struct mlx5_flow_root_namespace **rdma_transport_tx_root_ns;
 	int rdma_transport_rx_vports;
@@ -378,11 +376,6 @@ void mlx5_fs_core_free(struct mlx5_core_dev *dev);
 int mlx5_fs_core_init(struct mlx5_core_dev *dev);
 void mlx5_fs_core_cleanup(struct mlx5_core_dev *dev);
 
-int mlx5_fs_egress_acls_init(struct mlx5_core_dev *dev, int total_vports);
-void mlx5_fs_egress_acls_cleanup(struct mlx5_core_dev *dev);
-int mlx5_fs_ingress_acls_init(struct mlx5_core_dev *dev, int total_vports);
-void mlx5_fs_ingress_acls_cleanup(struct mlx5_core_dev *dev);
-
 int mlx5_fs_vport_egress_acl_ns_add(struct mlx5_flow_steering *steering,
 				    u16 vport_idx);
 int mlx5_fs_vport_ingress_acl_ns_add(struct mlx5_flow_steering *steering,
-- 
2.50.1

Re: [PATCH net-next V2 2/7] net/mlx5: E-Switch, Move vport acls root namespaces creation to eswitch

[Simon Horman]

On Tue, Aug 26, 2025 at 09:45:11PM -0700, Saeed Mahameed wrote:
> From: Saeed Mahameed <saeedm@nvidia.com>
> 
> Move the loop that creates the vports ACLs root name spaces to eswitch,
> since it is the eswitch responsibility to decide when and how many
> vports ACLs root namespaces to create, in the next patches we will use
> the fs_core vport ACL root namespace APIs to create/remove root ns
> ACLs dynamically for dynamically created vports.
> 
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Reviewed-by: Simon Horman <horms@kernel.org>

[PATCH net-next V2 3/7] net/mlx5: E-Switch, Add support for adjacent functions vports discovery

[Saeed Mahameed]

From: Adithya Jayachandran <ajayachandra@nvidia.com>

Adding driver support to query adjacent functions vports, AKA
delegated vports.

Adjacent functions can delegate their sriov vfs to other sibling PF in
the system, to be managed by the eswitch capable sibling PF.
E.g, ECPF to Host PF, multi host PF between each other, etc.

Only supported in switchdev mode.

Signed-off-by: Adithya Jayachandran <ajayachandra@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/Makefile  |   2 +-
 .../mellanox/mlx5/core/esw/adj_vport.c        | 185 ++++++++++++++++++
 .../net/ethernet/mellanox/mlx5/core/eswitch.c |  58 +++++-
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |   7 +
 .../mellanox/mlx5/core/eswitch_offloads.c     |   4 +
 5 files changed, 251 insertions(+), 5 deletions(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/Makefile b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
index a253c73db9e5..36ef2a47e757 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/Makefile
+++ b/drivers/net/ethernet/mellanox/mlx5/core/Makefile
@@ -69,7 +69,7 @@ mlx5_core-$(CONFIG_MLX5_TC_SAMPLE)   += en/tc/sample.o
 # Core extra
 #
 mlx5_core-$(CONFIG_MLX5_ESWITCH)   += eswitch.o eswitch_offloads.o eswitch_offloads_termtbl.o \
-				      ecpf.o rdma.o esw/legacy.o \
+				      ecpf.o rdma.o esw/legacy.o esw/adj_vport.o \
 				      esw/devlink_port.o esw/vporttbl.o esw/qos.o esw/ipsec.o
 
 mlx5_core-$(CONFIG_MLX5_ESWITCH)   += esw/acl/helper.o \
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
new file mode 100644
index 000000000000..37a06c0949d5
--- /dev/null
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
@@ -0,0 +1,185 @@
+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+// Copyright (c) 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+
+#include "eswitch.h"
+
+enum {
+	MLX5_ADJ_VPORT_DISCONNECT = 0x0,
+	MLX5_ADJ_VPORT_CONNECT = 0x1,
+};
+
+static int mlx5_esw_adj_vport_modify(struct mlx5_core_dev *dev,
+				     u16 vport, bool connect)
+{
+	u32 in[MLX5_ST_SZ_DW(modify_vport_state_in)] = {};
+
+	MLX5_SET(modify_vport_state_in, in, opcode,
+		 MLX5_CMD_OP_MODIFY_VPORT_STATE);
+	MLX5_SET(modify_vport_state_in, in, op_mod,
+		 MLX5_VPORT_STATE_OP_MOD_ESW_VPORT);
+	MLX5_SET(modify_vport_state_in, in, other_vport, 1);
+	MLX5_SET(modify_vport_state_in, in, vport_number, vport);
+	MLX5_SET(modify_vport_state_in, in, ingress_connect_valid, 1);
+	MLX5_SET(modify_vport_state_in, in, egress_connect_valid, 1);
+	MLX5_SET(modify_vport_state_in, in, ingress_connect, connect);
+	MLX5_SET(modify_vport_state_in, in, egress_connect, connect);
+
+	return mlx5_cmd_exec_in(dev, modify_vport_state, in);
+}
+
+static void mlx5_esw_destroy_esw_vport(struct mlx5_core_dev *dev, u16 vport)
+{
+	u32 in[MLX5_ST_SZ_DW(destroy_esw_vport_in)] = {};
+
+	MLX5_SET(destroy_esw_vport_in, in, opcode,
+		 MLX5_CMD_OPCODE_DESTROY_ESW_VPORT);
+	MLX5_SET(destroy_esw_vport_in, in, vport_num, vport);
+
+	mlx5_cmd_exec_in(dev, destroy_esw_vport, in);
+}
+
+static int mlx5_esw_create_esw_vport(struct mlx5_core_dev *dev, u16 vhca_id,
+				     u16 *vport_num)
+{
+	u32 out[MLX5_ST_SZ_DW(create_esw_vport_out)] = {};
+	u32 in[MLX5_ST_SZ_DW(create_esw_vport_in)] = {};
+	int err;
+
+	MLX5_SET(create_esw_vport_in, in, opcode,
+		 MLX5_CMD_OPCODE_CREATE_ESW_VPORT);
+	MLX5_SET(create_esw_vport_in, in, managed_vhca_id, vhca_id);
+
+	err = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+	if (!err)
+		*vport_num = MLX5_GET(create_esw_vport_out, out, vport_num);
+
+	return err;
+}
+
+static int mlx5_esw_adj_vport_create(struct mlx5_eswitch *esw, u16 vhca_id)
+{
+	struct mlx5_vport *vport;
+	u16 vport_num;
+	int err;
+
+	err = mlx5_esw_create_esw_vport(esw->dev, vhca_id, &vport_num);
+	if (err) {
+		esw_warn(esw->dev,
+			 "Failed to create adjacent vport for vhca_id %d, err %d\n",
+			 vhca_id, err);
+		return err;
+	}
+
+	esw_debug(esw->dev, "Created adjacent vport[%d] %d for vhca_id 0x%x\n",
+		  esw->last_vport_idx, vport_num, vhca_id);
+
+	err = mlx5_esw_vport_alloc(esw, esw->last_vport_idx++, vport_num);
+	if (err)
+		goto destroy_esw_vport;
+
+	xa_set_mark(&esw->vports, vport_num, MLX5_ESW_VPT_VF);
+	vport = mlx5_eswitch_get_vport(esw, vport_num);
+	vport->adjacent = true;
+	vport->vhca_id = vhca_id;
+
+	mlx5_esw_adj_vport_modify(esw->dev, vport_num, MLX5_ADJ_VPORT_CONNECT);
+	return 0;
+
+destroy_esw_vport:
+	mlx5_esw_destroy_esw_vport(esw->dev, vport_num);
+	return err;
+}
+
+static void mlx5_esw_adj_vport_destroy(struct mlx5_eswitch *esw,
+				       struct mlx5_vport *vport)
+{
+	u16 vport_num = vport->vport;
+
+	esw_debug(esw->dev, "Destroying adjacent vport %d for vhca_id 0x%x\n",
+		  vport_num, vport->vhca_id);
+	mlx5_esw_adj_vport_modify(esw->dev, vport_num,
+				  MLX5_ADJ_VPORT_DISCONNECT);
+	mlx5_esw_vport_free(esw, vport);
+	/* Reset the vport index back so new adj vports can use this index.
+	 * When vport count can incrementally change, this needs to be modified.
+	 */
+	esw->last_vport_idx--;
+	mlx5_esw_destroy_esw_vport(esw->dev, vport_num);
+}
+
+void mlx5_esw_adjacent_vhcas_cleanup(struct mlx5_eswitch *esw)
+{
+	struct mlx5_vport *vport;
+	unsigned long i;
+
+	if (!MLX5_CAP_GEN_2(esw->dev, delegated_vhca_max))
+		return;
+
+	mlx5_esw_for_each_vf_vport(esw, i, vport, U16_MAX) {
+		if (!vport->adjacent)
+			continue;
+		mlx5_esw_adj_vport_destroy(esw, vport);
+	}
+}
+
+void mlx5_esw_adjacent_vhcas_setup(struct mlx5_eswitch *esw)
+{
+	u32 delegated_vhca_max = MLX5_CAP_GEN_2(esw->dev, delegated_vhca_max);
+	u32 in[MLX5_ST_SZ_DW(query_delegated_vhca_in)] = {};
+	int outlen, err, i = 0;
+	u8 *out;
+	u32 count;
+
+	if (!delegated_vhca_max)
+		return;
+
+	outlen = MLX5_ST_SZ_BYTES(query_delegated_vhca_out) +
+		 delegated_vhca_max *
+		 MLX5_ST_SZ_BYTES(delegated_function_vhca_rid_info);
+
+	esw_debug(esw->dev, "delegated_vhca_max=%d\n", delegated_vhca_max);
+
+	out = kvzalloc(outlen, GFP_KERNEL);
+	if (!out)
+		return;
+
+	MLX5_SET(query_delegated_vhca_in, in, opcode,
+		 MLX5_CMD_OPCODE_QUERY_DELEGATED_VHCA);
+
+	err = mlx5_cmd_exec(esw->dev, in, sizeof(in), out, outlen);
+	if (err) {
+		kvfree(out);
+		esw_warn(esw->dev, "Failed to query delegated vhca, err %d\n",
+			 err);
+		return;
+	}
+
+	count = MLX5_GET(query_delegated_vhca_out, out, functions_count);
+	esw_debug(esw->dev, "Delegated vhca functions count %d\n", count);
+
+	for (i = 0; i < count; i++) {
+		void *rid_info, *rid_info_reg;
+		u16 vhca_id;
+
+		rid_info = MLX5_ADDR_OF(query_delegated_vhca_out, out,
+					delegated_function_vhca_rid_info[i]);
+
+		rid_info_reg = MLX5_ADDR_OF(delegated_function_vhca_rid_info,
+					    rid_info, function_vhca_rid_info);
+
+		vhca_id = MLX5_GET(function_vhca_rid_info_reg, rid_info_reg,
+				   vhca_id);
+		esw_debug(esw->dev, "Delegating vhca_id 0x%x rid info:\n",
+			  vhca_id);
+
+		err = mlx5_esw_adj_vport_create(esw, vhca_id);
+		if (err) {
+			esw_warn(esw->dev,
+				 "Failed to init adjacent vhca 0x%x, err %d\n",
+				 vhca_id, err);
+			break;
+		}
+	}
+
+	kvfree(out);
+}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index 900650a1a66e..ef2e1d30b9b4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -1217,7 +1217,8 @@ void mlx5_eswitch_unload_vf_vports(struct mlx5_eswitch *esw, u16 num_vfs)
 	unsigned long i;
 
 	mlx5_esw_for_each_vf_vport(esw, i, vport, num_vfs) {
-		if (!vport->enabled)
+		/* Adjacent VFs are unloaded separately */
+		if (!vport->enabled || vport->adjacent)
 			continue;
 		mlx5_eswitch_unload_pf_vf_vport(esw, vport->vport);
 	}
@@ -1236,6 +1237,42 @@ static void mlx5_eswitch_unload_ec_vf_vports(struct mlx5_eswitch *esw,
 	}
 }
 
+static void mlx5_eswitch_unload_adj_vf_vports(struct mlx5_eswitch *esw)
+{
+	struct mlx5_vport *vport;
+	unsigned long i;
+
+	mlx5_esw_for_each_vf_vport(esw, i, vport, U16_MAX) {
+		if (!vport->enabled || !vport->adjacent)
+			continue;
+		mlx5_eswitch_unload_pf_vf_vport(esw, vport->vport);
+	}
+}
+
+static int
+mlx5_eswitch_load_adj_vf_vports(struct mlx5_eswitch *esw,
+				enum mlx5_eswitch_vport_event enabled_events)
+{
+	struct mlx5_vport *vport;
+	unsigned long i;
+	int err;
+
+	mlx5_esw_for_each_vf_vport(esw, i, vport, U16_MAX) {
+		if (!vport->adjacent)
+			continue;
+		err = mlx5_eswitch_load_pf_vf_vport(esw, vport->vport,
+						    enabled_events);
+		if (err)
+			goto unload_adj_vf_vport;
+	}
+
+	return 0;
+
+unload_adj_vf_vport:
+	mlx5_eswitch_unload_adj_vf_vports(esw);
+	return err;
+}
+
 int mlx5_eswitch_load_vf_vports(struct mlx5_eswitch *esw, u16 num_vfs,
 				enum mlx5_eswitch_vport_event enabled_events)
 {
@@ -1345,7 +1382,15 @@ mlx5_eswitch_enable_pf_vf_vports(struct mlx5_eswitch *esw,
 					  enabled_events);
 	if (ret)
 		goto vf_err;
+
+	/* Enable adjacent VF vports */
+	ret = mlx5_eswitch_load_adj_vf_vports(esw, enabled_events);
+	if (ret)
+		goto unload_adj_vf_vports;
+
 	return 0;
+unload_adj_vf_vports:
+	mlx5_eswitch_unload_adj_vf_vports(esw);
 
 vf_err:
 	if (mlx5_core_ec_sriov_enabled(esw->dev))
@@ -1367,6 +1412,8 @@ mlx5_eswitch_enable_pf_vf_vports(struct mlx5_eswitch *esw,
  */
 void mlx5_eswitch_disable_pf_vf_vports(struct mlx5_eswitch *esw)
 {
+	mlx5_eswitch_unload_adj_vf_vports(esw);
+
 	mlx5_eswitch_unload_vf_vports(esw, esw->esw_funcs.num_vfs);
 
 	if (mlx5_core_ec_sriov_enabled(esw->dev))
@@ -1791,8 +1838,7 @@ int mlx5_esw_sf_max_hpf_functions(struct mlx5_core_dev *dev, u16 *max_sfs, u16 *
 	return err;
 }
 
-static int mlx5_esw_vport_alloc(struct mlx5_eswitch *esw,
-				int index, u16 vport_num)
+int mlx5_esw_vport_alloc(struct mlx5_eswitch *esw, int index, u16 vport_num)
 {
 	struct mlx5_vport *vport;
 	int err;
@@ -1819,8 +1865,9 @@ static int mlx5_esw_vport_alloc(struct mlx5_eswitch *esw,
 	return err;
 }
 
-static void mlx5_esw_vport_free(struct mlx5_eswitch *esw, struct mlx5_vport *vport)
+void mlx5_esw_vport_free(struct mlx5_eswitch *esw, struct mlx5_vport *vport)
 {
+	esw->total_vports--;
 	xa_erase(&esw->vports, vport->vport);
 	kfree(vport);
 }
@@ -1904,6 +1951,9 @@ static int mlx5_esw_vports_init(struct mlx5_eswitch *esw)
 	err = mlx5_esw_vport_alloc(esw, idx, MLX5_VPORT_UPLINK);
 	if (err)
 		goto err;
+
+	/* Adjacent vports or other dynamically create vports will use this */
+	esw->last_vport_idx = ++idx;
 	return 0;
 
 err:
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index cfd6b1b8c6f4..9f8bb397eae5 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -216,6 +216,7 @@ struct mlx5_vport {
 	u32                     metadata;
 	int                     vhca_id;
 
+	bool adjacent; /* delegated vhca from adjacent function */
 	struct mlx5_vport_info  info;
 
 	/* Protected with the E-Switch qos domain lock. The Vport QoS can
@@ -384,6 +385,7 @@ struct mlx5_eswitch {
 
 	struct mlx5_esw_bridge_offloads *br_offloads;
 	struct mlx5_esw_offload offloads;
+	u32 last_vport_idx; /* ++ every time a vport is created */
 	int                     mode;
 	u16                     manager_vport;
 	u16                     first_host_vport;
@@ -417,6 +419,8 @@ int mlx5_esw_qos_modify_vport_rate(struct mlx5_eswitch *esw, u16 vport_num, u32
 /* E-Switch API */
 int mlx5_eswitch_init(struct mlx5_core_dev *dev);
 void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw);
+int mlx5_esw_vport_alloc(struct mlx5_eswitch *esw, int index, u16 vport_num);
+void mlx5_esw_vport_free(struct mlx5_eswitch *esw, struct mlx5_vport *vport);
 
 #define MLX5_ESWITCH_IGNORE_NUM_VFS (-1)
 int mlx5_eswitch_enable_locked(struct mlx5_eswitch *esw, int num_vfs);
@@ -622,6 +626,9 @@ bool mlx5_esw_multipath_prereq(struct mlx5_core_dev *dev0,
 
 const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev);
 
+void mlx5_esw_adjacent_vhcas_setup(struct mlx5_eswitch *esw);
+void mlx5_esw_adjacent_vhcas_cleanup(struct mlx5_eswitch *esw);
+
 #define MLX5_DEBUG_ESWITCH_MASK BIT(3)
 
 #define esw_info(__dev, format, ...)			\
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index cdba7bc448ee..fb03981d5036 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -3538,6 +3538,8 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)
 	int err;
 
 	mutex_init(&esw->offloads.termtbl_mutex);
+	mlx5_esw_adjacent_vhcas_setup(esw);
+
 	err = mlx5_rdma_enable_roce(esw->dev);
 	if (err)
 		goto err_roce;
@@ -3602,6 +3604,7 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)
 err_metadata:
 	mlx5_rdma_disable_roce(esw->dev);
 err_roce:
+	mlx5_esw_adjacent_vhcas_cleanup(esw);
 	mutex_destroy(&esw->offloads.termtbl_mutex);
 	return err;
 }
@@ -3635,6 +3638,7 @@ void esw_offloads_disable(struct mlx5_eswitch *esw)
 	mapping_destroy(esw->offloads.reg_c0_obj_pool);
 	esw_offloads_metadata_uninit(esw);
 	mlx5_rdma_disable_roce(esw->dev);
+	mlx5_esw_adjacent_vhcas_cleanup(esw);
 	mutex_destroy(&esw->offloads.termtbl_mutex);
 }
 
-- 
2.50.1

Re: [PATCH net-next V2 3/7] net/mlx5: E-Switch, Add support for adjacent functions vports discovery

[Simon Horman]

On Tue, Aug 26, 2025 at 09:45:12PM -0700, Saeed Mahameed wrote:
> From: Adithya Jayachandran <ajayachandra@nvidia.com>
> 
> Adding driver support to query adjacent functions vports, AKA
> delegated vports.
> 
> Adjacent functions can delegate their sriov vfs to other sibling PF in
> the system, to be managed by the eswitch capable sibling PF.
> E.g, ECPF to Host PF, multi host PF between each other, etc.
> 
> Only supported in switchdev mode.
> 
> Signed-off-by: Adithya Jayachandran <ajayachandra@nvidia.com>
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

...

> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c

...

> +static int
> +mlx5_eswitch_load_adj_vf_vports(struct mlx5_eswitch *esw,
> +				enum mlx5_eswitch_vport_event enabled_events)
> +{
> +	struct mlx5_vport *vport;
> +	unsigned long i;
> +	int err;
> +
> +	mlx5_esw_for_each_vf_vport(esw, i, vport, U16_MAX) {
> +		if (!vport->adjacent)
> +			continue;
> +		err = mlx5_eswitch_load_pf_vf_vport(esw, vport->vport,
> +						    enabled_events);
> +		if (err)
> +			goto unload_adj_vf_vport;
> +	}
> +
> +	return 0;
> +
> +unload_adj_vf_vport:
> +	mlx5_eswitch_unload_adj_vf_vports(esw);
> +	return err;
> +}
> +
>  int mlx5_eswitch_load_vf_vports(struct mlx5_eswitch *esw, u16 num_vfs,
>  				enum mlx5_eswitch_vport_event enabled_events)
>  {
> @@ -1345,7 +1382,15 @@ mlx5_eswitch_enable_pf_vf_vports(struct mlx5_eswitch *esw,
>  					  enabled_events);
>  	if (ret)
>  		goto vf_err;
> +
> +	/* Enable adjacent VF vports */
> +	ret = mlx5_eswitch_load_adj_vf_vports(esw, enabled_events);
> +	if (ret)
> +		goto unload_adj_vf_vports;
> +
>  	return 0;
> +unload_adj_vf_vports:
> +	mlx5_eswitch_unload_adj_vf_vports(esw);
>  

Hi Adithya and Saeed,

mlx5_eswitch_load_adj_vf_vports() already unwinds on error,
calling mlx5_eswitch_unload_adj_vf_vports().

While resources allocated by the preceding call to
mlx5_eswitch_load_vf_vports() (just before this hunk) seem to be leaked.

So I wonder if the above two lines should be:

unload_vf_vports:
	mlx5_eswitch_unload_vf_vports(esw, num_vfs);

>  vf_err:
>  	if (mlx5_core_ec_sriov_enabled(esw->dev))

...

> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
> index cfd6b1b8c6f4..9f8bb397eae5 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
> @@ -216,6 +216,7 @@ struct mlx5_vport {
>  	u32                     metadata;
>  	int                     vhca_id;
>  
> +	bool adjacent; /* delegated vhca from adjacent function */
>  	struct mlx5_vport_info  info;
>  
>  	/* Protected with the E-Switch qos domain lock. The Vport QoS can
> @@ -384,6 +385,7 @@ struct mlx5_eswitch {
>  
>  	struct mlx5_esw_bridge_offloads *br_offloads;
>  	struct mlx5_esw_offload offloads;
> +	u32 last_vport_idx; /* ++ every time a vport is created */

The comment above documents one operation that can occur
on this field. But the code also performs others: decrement and set.

So perhaps this is more appropriate?

	u32 last_vport_idx; /* Number of vports created */

Or dropping the comment entirely: it seems obvious enough.

>  	int                     mode;
>  	u16                     manager_vport;
>  	u16                     first_host_vport;

...

Re: [PATCH net-next V2 3/7] net/mlx5: E-Switch, Add support for adjacent functions vports discovery

[Saeed Mahameed]

On 28 Aug 12:34, Simon Horman wrote:
>On Tue, Aug 26, 2025 at 09:45:12PM -0700, Saeed Mahameed wrote:
>> From: Adithya Jayachandran <ajayachandra@nvidia.com>
>>
>> Adding driver support to query adjacent functions vports, AKA
>> delegated vports.
>>
>> Adjacent functions can delegate their sriov vfs to other sibling PF in
>> the system, to be managed by the eswitch capable sibling PF.
>> E.g, ECPF to Host PF, multi host PF between each other, etc.
>>
>> Only supported in switchdev mode.
>>
>> Signed-off-by: Adithya Jayachandran <ajayachandra@nvidia.com>
>> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
>
>...
>
>> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
>
>...
>
>> +static int
>> +mlx5_eswitch_load_adj_vf_vports(struct mlx5_eswitch *esw,
>> +				enum mlx5_eswitch_vport_event enabled_events)
>> +{
>> +	struct mlx5_vport *vport;
>> +	unsigned long i;
>> +	int err;
>> +
>> +	mlx5_esw_for_each_vf_vport(esw, i, vport, U16_MAX) {
>> +		if (!vport->adjacent)
>> +			continue;
>> +		err = mlx5_eswitch_load_pf_vf_vport(esw, vport->vport,
>> +						    enabled_events);
>> +		if (err)
>> +			goto unload_adj_vf_vport;
>> +	}
>> +
>> +	return 0;
>> +
>> +unload_adj_vf_vport:
>> +	mlx5_eswitch_unload_adj_vf_vports(esw);
>> +	return err;
>> +}
>> +
>>  int mlx5_eswitch_load_vf_vports(struct mlx5_eswitch *esw, u16 num_vfs,
>>  				enum mlx5_eswitch_vport_event enabled_events)
>>  {
>> @@ -1345,7 +1382,15 @@ mlx5_eswitch_enable_pf_vf_vports(struct mlx5_eswitch *esw,
>>  					  enabled_events);
>>  	if (ret)
>>  		goto vf_err;
>> +
>> +	/* Enable adjacent VF vports */
>> +	ret = mlx5_eswitch_load_adj_vf_vports(esw, enabled_events);
>> +	if (ret)
>> +		goto unload_adj_vf_vports;
>> +
>>  	return 0;
>> +unload_adj_vf_vports:
>> +	mlx5_eswitch_unload_adj_vf_vports(esw);
>>
>
>Hi Adithya and Saeed,
>
>mlx5_eswitch_load_adj_vf_vports() already unwinds on error,
>calling mlx5_eswitch_unload_adj_vf_vports().
>
>While resources allocated by the preceding call to
>mlx5_eswitch_load_vf_vports() (just before this hunk) seem to be leaked.
>
>So I wonder if the above two lines should be:
>
>unload_vf_vports:
>	mlx5_eswitch_unload_vf_vports(esw, num_vfs);
>

Yep, looks incorrect, let me double check and submit v3.

>>  vf_err:
>>  	if (mlx5_core_ec_sriov_enabled(esw->dev))
>
>...
>
>> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
>> index cfd6b1b8c6f4..9f8bb397eae5 100644
>> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
>> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
>> @@ -216,6 +216,7 @@ struct mlx5_vport {
>>  	u32                     metadata;
>>  	int                     vhca_id;
>>
>> +	bool adjacent; /* delegated vhca from adjacent function */
>>  	struct mlx5_vport_info  info;
>>
>>  	/* Protected with the E-Switch qos domain lock. The Vport QoS can
>> @@ -384,6 +385,7 @@ struct mlx5_eswitch {
>>
>>  	struct mlx5_esw_bridge_offloads *br_offloads;
>>  	struct mlx5_esw_offload offloads;
>> +	u32 last_vport_idx; /* ++ every time a vport is created */
>
>The comment above documents one operation that can occur
>on this field. But the code also performs others: decrement and set.
>
>So perhaps this is more appropriate?
>
>	u32 last_vport_idx; /* Number of vports created */
>
>Or dropping the comment entirely: it seems obvious enough.
>

Agreed, will drop.

[PATCH net-next V2 4/7] net/mlx5: E-Switch, Create acls root namespace for adjacent vports

[Saeed Mahameed]

From: Saeed Mahameed <saeedm@nvidia.com>

Use the new vport acl root namespace add/remove API to create the
missing acl root name spaces per each new adjacent function vport.

Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
index 37a06c0949d5..1d104b3fe9e0 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
 // Copyright (c) 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
+#include "fs_core.h"
 #include "eswitch.h"
 
 enum {
@@ -82,6 +83,9 @@ static int mlx5_esw_adj_vport_create(struct mlx5_eswitch *esw, u16 vhca_id)
 	vport->adjacent = true;
 	vport->vhca_id = vhca_id;
 
+	mlx5_fs_vport_egress_acl_ns_add(esw->dev->priv.steering, vport->index);
+	mlx5_fs_vport_ingress_acl_ns_add(esw->dev->priv.steering, vport->index);
+
 	mlx5_esw_adj_vport_modify(esw->dev, vport_num, MLX5_ADJ_VPORT_CONNECT);
 	return 0;
 
@@ -99,6 +103,10 @@ static void mlx5_esw_adj_vport_destroy(struct mlx5_eswitch *esw,
 		  vport_num, vport->vhca_id);
 	mlx5_esw_adj_vport_modify(esw->dev, vport_num,
 				  MLX5_ADJ_VPORT_DISCONNECT);
+	mlx5_fs_vport_egress_acl_ns_remove(esw->dev->priv.steering,
+					   vport->index);
+	mlx5_fs_vport_ingress_acl_ns_remove(esw->dev->priv.steering,
+					    vport->index);
 	mlx5_esw_vport_free(esw, vport);
 	/* Reset the vport index back so new adj vports can use this index.
 	 * When vport count can incrementally change, this needs to be modified.
-- 
2.50.1

Re: [PATCH net-next V2 4/7] net/mlx5: E-Switch, Create acls root namespace for adjacent vports

[Simon Horman]

On Tue, Aug 26, 2025 at 09:45:13PM -0700, Saeed Mahameed wrote:
> From: Saeed Mahameed <saeedm@nvidia.com>
> 
> Use the new vport acl root namespace add/remove API to create the
> missing acl root name spaces per each new adjacent function vport.
> 
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Reviewed-by: Simon Horman <horms@kernel.org>

[PATCH net-next V2 5/7] net/mlx5: E-Switch, Register representors for adjacent vports

[Saeed Mahameed]

From: Saeed Mahameed <saeedm@nvidia.com>

Register representors for adjacent vports dynamically when they are
discovered. Dynamically added representors state will now be set to
'REGISTERED' when the representor type was already registered,
otherwise they won't be loaded.

Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../mellanox/mlx5/core/esw/adj_vport.c        | 10 ++++++
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |  5 +++
 .../mellanox/mlx5/core/eswitch_offloads.c     | 33 ++++++++++++++++---
 3 files changed, 43 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
index 1d104b3fe9e0..3380f85678bc 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
@@ -85,10 +85,19 @@ static int mlx5_esw_adj_vport_create(struct mlx5_eswitch *esw, u16 vhca_id)
 
 	mlx5_fs_vport_egress_acl_ns_add(esw->dev->priv.steering, vport->index);
 	mlx5_fs_vport_ingress_acl_ns_add(esw->dev->priv.steering, vport->index);
+	err = mlx5_esw_offloads_rep_add(esw, vport);
+	if (err)
+		goto acl_ns_remove;
 
 	mlx5_esw_adj_vport_modify(esw->dev, vport_num, MLX5_ADJ_VPORT_CONNECT);
 	return 0;
 
+acl_ns_remove:
+	mlx5_fs_vport_ingress_acl_ns_remove(esw->dev->priv.steering,
+					    vport->index);
+	mlx5_fs_vport_egress_acl_ns_remove(esw->dev->priv.steering,
+					   vport->index);
+	mlx5_esw_vport_free(esw, vport);
 destroy_esw_vport:
 	mlx5_esw_destroy_esw_vport(esw->dev, vport_num);
 	return err;
@@ -103,6 +112,7 @@ static void mlx5_esw_adj_vport_destroy(struct mlx5_eswitch *esw,
 		  vport_num, vport->vhca_id);
 	mlx5_esw_adj_vport_modify(esw->dev, vport_num,
 				  MLX5_ADJ_VPORT_DISCONNECT);
+	mlx5_esw_offloads_rep_remove(esw, vport);
 	mlx5_fs_vport_egress_acl_ns_remove(esw->dev->priv.steering,
 					   vport->index);
 	mlx5_fs_vport_ingress_acl_ns_remove(esw->dev->priv.steering,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index 9f8bb397eae5..9fc020bc40cf 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -838,6 +838,11 @@ void mlx5_esw_vport_vhca_id_unmap(struct mlx5_eswitch *esw,
 int mlx5_eswitch_vhca_id_to_vport(struct mlx5_eswitch *esw, u16 vhca_id, u16 *vport_num);
 bool mlx5_esw_vport_vhca_id(struct mlx5_eswitch *esw, u16 vportn, u16 *vhca_id);
 
+void mlx5_esw_offloads_rep_remove(struct mlx5_eswitch *esw,
+				  const struct mlx5_vport *vport);
+int mlx5_esw_offloads_rep_add(struct mlx5_eswitch *esw,
+			      const struct mlx5_vport *vport);
+
 /**
  * struct mlx5_esw_event_info - Indicates eswitch mode changed/changing.
  *
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index fb03981d5036..d57f86d297ab 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -2378,7 +2378,20 @@ static int esw_offloads_start(struct mlx5_eswitch *esw,
 	return 0;
 }
 
-static int mlx5_esw_offloads_rep_init(struct mlx5_eswitch *esw, const struct mlx5_vport *vport)
+void mlx5_esw_offloads_rep_remove(struct mlx5_eswitch *esw,
+				  const struct mlx5_vport *vport)
+{
+	struct mlx5_eswitch_rep *rep = xa_load(&esw->offloads.vport_reps,
+					       vport->vport);
+
+	if (!rep)
+		return;
+	xa_erase(&esw->offloads.vport_reps, vport->vport);
+	kfree(rep);
+}
+
+int mlx5_esw_offloads_rep_add(struct mlx5_eswitch *esw,
+			      const struct mlx5_vport *vport)
 {
 	struct mlx5_eswitch_rep *rep;
 	int rep_type;
@@ -2390,9 +2403,19 @@ static int mlx5_esw_offloads_rep_init(struct mlx5_eswitch *esw, const struct mlx
 
 	rep->vport = vport->vport;
 	rep->vport_index = vport->index;
-	for (rep_type = 0; rep_type < NUM_REP_TYPES; rep_type++)
-		atomic_set(&rep->rep_data[rep_type].state, REP_UNREGISTERED);
-
+	for (rep_type = 0; rep_type < NUM_REP_TYPES; rep_type++) {
+		if (!esw->offloads.rep_ops[rep_type]) {
+			atomic_set(&rep->rep_data[rep_type].state,
+				   REP_UNREGISTERED);
+			continue;
+		}
+		/* Dynamic/delegated vports add their representors after
+		 * mlx5_eswitch_register_vport_reps, so mark them as registered
+		 * for them to be loaded later with the others.
+		 */
+		rep->esw = esw;
+		atomic_set(&rep->rep_data[rep_type].state, REP_REGISTERED);
+	}
 	err = xa_insert(&esw->offloads.vport_reps, rep->vport, rep, GFP_KERNEL);
 	if (err)
 		goto insert_err;
@@ -2430,7 +2453,7 @@ static int esw_offloads_init_reps(struct mlx5_eswitch *esw)
 	xa_init(&esw->offloads.vport_reps);
 
 	mlx5_esw_for_each_vport(esw, i, vport) {
-		err = mlx5_esw_offloads_rep_init(esw, vport);
+		err = mlx5_esw_offloads_rep_add(esw, vport);
 		if (err)
 			goto err;
 	}
-- 
2.50.1

Re: [PATCH net-next V2 5/7] net/mlx5: E-Switch, Register representors for adjacent vports

[Simon Horman]

On Tue, Aug 26, 2025 at 09:45:14PM -0700, Saeed Mahameed wrote:
> From: Saeed Mahameed <saeedm@nvidia.com>
> 
> Register representors for adjacent vports dynamically when they are
> discovered. Dynamically added representors state will now be set to
> 'REGISTERED' when the representor type was already registered,
> otherwise they won't be loaded.
> 
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Reviewed-by: Simon Horman <horms@kernel.org>

[PATCH net-next V2 6/7] net/mlx5: E-switch, Set representor attributes for adjacent VFs

[Saeed Mahameed]

From: Adithya Jayachandran <ajayachandra@nvidia.com>

Adjacent vfs get their devlink port information from firmware,
use the information (pfnum, function id) from FW when populating the
devlink port attributes.

Before:
$ devlink port show
pci/0000:00:03.0/180225: type eth netdev eth0 flavour pcivf controller 0 pfnum 0 vfnum 49152 external false splittable false
  function:
    hw_addr 00:00:00:00:00:00

After:
$ devlink port show
pci/0000:00:03.0/180225: type eth netdev enp0s3npf0vf2 flavour pcivf controller 0 pfnum 0 vfnum 2 external false splittable false
  function:
    hw_addr 00:00:00:00:00:00

Signed-off-by: Adithya Jayachandran <ajayachandra@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../ethernet/mellanox/mlx5/core/esw/adj_vport.c  | 16 +++++++++++-----
 .../mellanox/mlx5/core/esw/devlink_port.c        | 11 ++++++++++-
 .../net/ethernet/mellanox/mlx5/core/eswitch.h    |  5 +++++
 3 files changed, 26 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
index 3380f85678bc..0091ba697bae 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c
@@ -57,7 +57,8 @@ static int mlx5_esw_create_esw_vport(struct mlx5_core_dev *dev, u16 vhca_id,
 	return err;
 }
 
-static int mlx5_esw_adj_vport_create(struct mlx5_eswitch *esw, u16 vhca_id)
+static int mlx5_esw_adj_vport_create(struct mlx5_eswitch *esw, u16 vhca_id,
+				     const void *rid_info_reg)
 {
 	struct mlx5_vport *vport;
 	u16 vport_num;
@@ -83,6 +84,12 @@ static int mlx5_esw_adj_vport_create(struct mlx5_eswitch *esw, u16 vhca_id)
 	vport->adjacent = true;
 	vport->vhca_id = vhca_id;
 
+	vport->adj_info.parent_pci_devfn =
+		MLX5_GET(function_vhca_rid_info_reg, rid_info_reg,
+			 parent_pci_device_function);
+	vport->adj_info.function_id =
+		MLX5_GET(function_vhca_rid_info_reg, rid_info_reg, function_id);
+
 	mlx5_fs_vport_egress_acl_ns_add(esw->dev->priv.steering, vport->index);
 	mlx5_fs_vport_ingress_acl_ns_add(esw->dev->priv.steering, vport->index);
 	err = mlx5_esw_offloads_rep_add(esw, vport);
@@ -176,7 +183,7 @@ void mlx5_esw_adjacent_vhcas_setup(struct mlx5_eswitch *esw)
 	esw_debug(esw->dev, "Delegated vhca functions count %d\n", count);
 
 	for (i = 0; i < count; i++) {
-		void *rid_info, *rid_info_reg;
+		const void *rid_info, *rid_info_reg;
 		u16 vhca_id;
 
 		rid_info = MLX5_ADDR_OF(query_delegated_vhca_out, out,
@@ -187,10 +194,9 @@ void mlx5_esw_adjacent_vhcas_setup(struct mlx5_eswitch *esw)
 
 		vhca_id = MLX5_GET(function_vhca_rid_info_reg, rid_info_reg,
 				   vhca_id);
-		esw_debug(esw->dev, "Delegating vhca_id 0x%x rid info:\n",
-			  vhca_id);
+		esw_debug(esw->dev, "Delegating vhca_id 0x%x\n", vhca_id);
 
-		err = mlx5_esw_adj_vport_create(esw, vhca_id);
+		err = mlx5_esw_adj_vport_create(esw, vhca_id, rid_info_reg);
 		if (err) {
 			esw_warn(esw->dev,
 				 "Failed to init adjacent vhca 0x%x, err %d\n",
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c
index c33accadae0f..cf88a106d80d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/devlink_port.c
@@ -27,6 +27,7 @@ static void mlx5_esw_offloads_pf_vf_devlink_port_attrs_set(struct mlx5_eswitch *
 {
 	struct mlx5_core_dev *dev = esw->dev;
 	struct netdev_phys_item_id ppid = {};
+	struct mlx5_vport *vport;
 	u32 controller_num = 0;
 	bool external;
 	u16 pfnum;
@@ -42,10 +43,18 @@ static void mlx5_esw_offloads_pf_vf_devlink_port_attrs_set(struct mlx5_eswitch *
 		dl_port->attrs.switch_id.id_len = ppid.id_len;
 		devlink_port_attrs_pci_pf_set(dl_port, controller_num, pfnum, external);
 	} else if (mlx5_eswitch_is_vf_vport(esw, vport_num)) {
+		u16 func_id = vport_num - 1;
+
+		vport = mlx5_eswitch_get_vport(esw, vport_num);
 		memcpy(dl_port->attrs.switch_id.id, ppid.id, ppid.id_len);
 		dl_port->attrs.switch_id.id_len = ppid.id_len;
+		if (vport->adjacent) {
+			func_id = vport->adj_info.function_id;
+			pfnum = vport->adj_info.parent_pci_devfn;
+		}
+
 		devlink_port_attrs_pci_vf_set(dl_port, controller_num, pfnum,
-					      vport_num - 1, external);
+					      func_id, external);
 	}  else if (mlx5_core_is_ec_vf_vport(esw->dev, vport_num)) {
 		u16 base_vport = mlx5_core_ec_vf_vport_base(dev);
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index 9fc020bc40cf..6fb925949ed1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -217,6 +217,11 @@ struct mlx5_vport {
 	int                     vhca_id;
 
 	bool adjacent; /* delegated vhca from adjacent function */
+	struct {
+		u16 parent_pci_devfn; /* Adjacent parent PCI device function */
+		u16 function_id; /* Function ID of the delegated VPort */
+	} adj_info;
+
 	struct mlx5_vport_info  info;
 
 	/* Protected with the E-Switch qos domain lock. The Vport QoS can
-- 
2.50.1

Re: [PATCH net-next V2 6/7] net/mlx5: E-switch, Set representor attributes for adjacent VFs

[Simon Horman]

On Tue, Aug 26, 2025 at 09:45:15PM -0700, Saeed Mahameed wrote:
> From: Adithya Jayachandran <ajayachandra@nvidia.com>
> 
> Adjacent vfs get their devlink port information from firmware,
> use the information (pfnum, function id) from FW when populating the
> devlink port attributes.
> 
> Before:
> $ devlink port show
> pci/0000:00:03.0/180225: type eth netdev eth0 flavour pcivf controller 0 pfnum 0 vfnum 49152 external false splittable false
>   function:
>     hw_addr 00:00:00:00:00:00
> 
> After:
> $ devlink port show
> pci/0000:00:03.0/180225: type eth netdev enp0s3npf0vf2 flavour pcivf controller 0 pfnum 0 vfnum 2 external false splittable false
>   function:
>     hw_addr 00:00:00:00:00:00
> 
> Signed-off-by: Adithya Jayachandran <ajayachandra@nvidia.com>
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Reviewed-by: Simon Horman <horms@kernel.org>

[PATCH net-next V2 7/7] net/mlx5: {DR,HWS}, Use the cached vhca_id for this device

[Saeed Mahameed]

From: Saeed Mahameed <saeedm@nvidia.com>

The mlx5 driver caches many capabilities to be used by mlx5 layers.

In SW and HW steering we can use the cached vhca_id instead of invoking
FW commands.

Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
---
 .../mellanox/mlx5/core/steering/hws/cmd.c     | 34 ++++---------------
 .../mellanox/mlx5/core/steering/sws/dr_cmd.c  | 34 ++++---------------
 2 files changed, 14 insertions(+), 54 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/cmd.c
index acb0317f930b..f22eaf506d28 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/cmd.c
@@ -1200,40 +1200,20 @@ int mlx5hws_cmd_query_caps(struct mlx5_core_dev *mdev,
 int mlx5hws_cmd_query_gvmi(struct mlx5_core_dev *mdev, bool other_function,
 			   u16 vport_number, u16 *gvmi)
 {
-	u32 in[MLX5_ST_SZ_DW(query_hca_cap_in)] = {};
-	int out_size;
-	void *out;
 	int err;
 
-	if (other_function) {
-		err = mlx5_vport_get_vhca_id(mdev, vport_number, gvmi);
-		if  (!err)
-			return 0;
-
-		mlx5_core_err(mdev, "Failed to get vport vhca id for vport %d\n",
-			      vport_number);
-		return err;
+	if (!other_function) {
+		/* self vhca_id */
+		*gvmi = MLX5_CAP_GEN(mdev, vhca_id);
+		return 0;
 	}
 
-	/* get vhca_id for `this` function */
-	out_size = MLX5_ST_SZ_BYTES(query_hca_cap_out);
-	out = kzalloc(out_size, GFP_KERNEL);
-	if (!out)
-		return -ENOMEM;
-
-	MLX5_SET(query_hca_cap_in, in, opcode, MLX5_CMD_OP_QUERY_HCA_CAP);
-	MLX5_SET(query_hca_cap_in, in, op_mod,
-		 MLX5_SET_HCA_CAP_OP_MOD_GENERAL_DEVICE << 1 | HCA_CAP_OPMOD_GET_CUR);
-
-	err = mlx5_cmd_exec_inout(mdev, query_hca_cap, in, out);
+	err = mlx5_vport_get_vhca_id(mdev, vport_number, gvmi);
 	if (err) {
-		kfree(out);
+		mlx5_core_err(mdev, "Failed to get vport vhca id for vport %d\n",
+			      vport_number);
 		return err;
 	}
 
-	*gvmi = MLX5_GET(query_hca_cap_out, out, capability.cmd_hca_cap.vhca_id);
-
-	kfree(out);
-
 	return 0;
 }
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/sws/dr_cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/sws/dr_cmd.c
index bf99b933fd14..1ebb2b15c080 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/steering/sws/dr_cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/sws/dr_cmd.c
@@ -35,41 +35,21 @@ int mlx5dr_cmd_query_esw_vport_context(struct mlx5_core_dev *mdev,
 int mlx5dr_cmd_query_gvmi(struct mlx5_core_dev *mdev, bool other_vport,
 			  u16 vport_number, u16 *gvmi)
 {
-	u32 in[MLX5_ST_SZ_DW(query_hca_cap_in)] = {};
-	int out_size;
-	void *out;
 	int err;
 
-	if (other_vport) {
-		err = mlx5_vport_get_vhca_id(mdev, vport_number, gvmi);
-		if  (!err)
-			return 0;
-
-		mlx5_core_err(mdev, "Failed to get vport vhca id for vport %d\n",
-			      vport_number);
-		return err;
+	if (!other_vport) {
+		/* self vhca_id */
+		*gvmi = MLX5_CAP_GEN(mdev, vhca_id);
+		return 0;
 	}
 
-	/* get vhca_id for `this` function */
-	out_size = MLX5_ST_SZ_BYTES(query_hca_cap_out);
-	out = kzalloc(out_size, GFP_KERNEL);
-	if (!out)
-		return -ENOMEM;
-
-	MLX5_SET(query_hca_cap_in, in, opcode, MLX5_CMD_OP_QUERY_HCA_CAP);
-	MLX5_SET(query_hca_cap_in, in, op_mod,
-		 MLX5_SET_HCA_CAP_OP_MOD_GENERAL_DEVICE << 1 |
-		 HCA_CAP_OPMOD_GET_CUR);
-
-	err = mlx5_cmd_exec_inout(mdev, query_hca_cap, in, out);
+	err = mlx5_vport_get_vhca_id(mdev, vport_number, gvmi);
 	if (err) {
-		kfree(out);
+		mlx5_core_err(mdev, "Failed to get vport vhca id for vport %d\n",
+			      vport_number);
 		return err;
 	}
 
-	*gvmi = MLX5_GET(query_hca_cap_out, out, capability.cmd_hca_cap.vhca_id);
-
-	kfree(out);
 	return 0;
 }
 
-- 
2.50.1

Re: [PATCH net-next V2 7/7] net/mlx5: {DR,HWS}, Use the cached vhca_id for this device

[Simon Horman]

On Tue, Aug 26, 2025 at 09:45:16PM -0700, Saeed Mahameed wrote:
> From: Saeed Mahameed <saeedm@nvidia.com>
> 
> The mlx5 driver caches many capabilities to be used by mlx5 layers.
> 
> In SW and HW steering we can use the cached vhca_id instead of invoking
> FW commands.
> 
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Reviewed-by: Simon Horman <horms@kernel.org>