[PATCH net-next V2 0/8] net/mlx5: Prepare eswitch infrastructure for satellite PF support

[PATCH net-next V2 0/8] net/mlx5: Prepare eswitch infrastructure for satellite PF support

[Tariq Toukan]

Hi,

See detailed description by Moshe below.

Regards,
Tariq

A satellite PF is a new SmartNIC configuration that adds another
physical function on the DPU that is not an eswitch manager and not a
page manager. The satellite PF can have its own SFs and can be passed
through to a VM on the DPU, providing an isolated function for users who
should not have access to the privileged ECPF. The ECPF handles the
satellite PF and the host PF in a similar way, using the same management
framework.

This series prepares the mlx5 eswitch command interface and vport
infrastructure for satellite PF support.

The first two patches abstract host PF data parsing behind a helper and
switch to the v1 response layout for query_esw_functions when supported,
so callers are insulated from layout differences.

The IPsec VF checks are tightened to use mlx5_eswitch_is_vf_vport()
instead of comparing against a specific vport number.

The remaining patches refactor SET_HCA_CAP and enable/disable_hca
command helpers to support vhca_id-based addressing, which is required
for managing functions that are not directly addressable by function_id.

A follow-up series will introduce satellite PF discovery and management
using this infrastructure.

V2:
- Add clarifications on vhca states of PFs.
- Add background on satellite PF feature.

V1:
https://lore.kernel.org/all/20260510053448.326823-1-tariqt@nvidia.com/

Moshe Shemesh (8):
  net/mlx5: Use helper to parse host PF info
  net/mlx5: Use v1 response layout for query_esw_functions
  net/mlx5: Use mlx5_eswitch_is_vf_vport() for IPsec VF checks
  net/mlx5: Switch vport HCA cap helpers to kvzalloc
  net/mlx5: Add mlx5_vport_set_other_func_general_cap macro
  net/mlx5: Refactor mlx5_set_msix_vec_count() SET_HCA_CAP
  net/mlx5: Use vport helper for IPsec eswitch set caps
  net/mlx5: Generalize enable/disable HCA for any PF vport

 .../net/ethernet/mellanox/mlx5/core/ecpf.c    |  24 ++-
 .../net/ethernet/mellanox/mlx5/core/ecpf.h    |   4 +-
 .../ethernet/mellanox/mlx5/core/esw/ipsec.c   |  83 +++------
 .../net/ethernet/mellanox/mlx5/core/eswitch.c | 157 +++++++++++++++---
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |  16 ++
 .../mellanox/mlx5/core/eswitch_offloads.c     |  38 ++---
 .../ethernet/mellanox/mlx5/core/mlx5_core.h   |   6 +
 .../net/ethernet/mellanox/mlx5/core/pci_irq.c |  27 +--
 .../net/ethernet/mellanox/mlx5/core/sriov.c   |   8 +-
 .../net/ethernet/mellanox/mlx5/core/vport.c   |  12 +-
 10 files changed, 227 insertions(+), 148 deletions(-)


base-commit: 627ac78f2741e2ebd2225e2e953b6964a8a9182f
-- 
2.44.0

[PATCH net-next V2 1/8] net/mlx5: Use helper to parse host PF info

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

Add a helper mlx5_esw_get_host_pf_info() to retrieve host PF data from
the query_esw_functions command output, so callers no longer need to
parse the layout to obtain the required information.

Convert all callers of mlx5_esw_query_functions() to use the new helper,
preparing for upcoming support of the new op_mod that returns data in
the network_function_params layout.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/eswitch.c | 43 ++++++++++++++-----
 .../net/ethernet/mellanox/mlx5/core/eswitch.h | 15 +++++++
 .../mellanox/mlx5/core/eswitch_offloads.c     | 34 ++++++---------
 .../net/ethernet/mellanox/mlx5/core/sriov.c   |  8 ++--
 4 files changed, 62 insertions(+), 38 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index 43c40353b2d8..861e79ddb489 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -1083,10 +1083,36 @@ const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev)
 	return ERR_PTR(err);
 }
 
+static struct mlx5_esw_pf_info
+mlx5_esw_host_pf_from_host_params(const void *entry)
+{
+	return (struct mlx5_esw_pf_info) {
+		.pf_not_exist = MLX5_GET(host_params_context, entry,
+					 host_pf_not_exist),
+		.pf_disabled = MLX5_GET(host_params_context, entry,
+					host_pf_disabled),
+		.num_of_vfs = MLX5_GET(host_params_context, entry,
+				       host_num_of_vfs),
+		.total_vfs = MLX5_GET(host_params_context, entry,
+				      host_total_vfs),
+		.host_number = MLX5_GET(host_params_context, entry,
+					host_number),
+	};
+}
+
+struct mlx5_esw_pf_info mlx5_esw_get_host_pf_info(const u32 *out)
+{
+	const void *entry;
+
+	entry = MLX5_ADDR_OF(query_esw_functions_out, out, net_function_params);
+
+	return mlx5_esw_host_pf_from_host_params(entry);
+}
+
 static int mlx5_esw_host_functions_enabled_query(struct mlx5_eswitch *esw)
 {
+	struct mlx5_esw_pf_info host_pf_info;
 	const u32 *query_host_out;
-	void *host_params;
 
 	if (!mlx5_core_is_ecpf_esw_manager(esw->dev))
 		return 0;
@@ -1095,11 +1121,8 @@ static int mlx5_esw_host_functions_enabled_query(struct mlx5_eswitch *esw)
 	if (IS_ERR(query_host_out))
 		return PTR_ERR(query_host_out);
 
-	host_params = MLX5_ADDR_OF(query_esw_functions_out,
-				   query_host_out, net_function_params);
-	esw->esw_funcs.host_funcs_disabled =
-		MLX5_GET(host_params_context, host_params,
-			 host_pf_not_exist);
+	host_pf_info = mlx5_esw_get_host_pf_info(query_host_out);
+	esw->esw_funcs.host_funcs_disabled = host_pf_info.pf_not_exist;
 
 	kvfree(query_host_out);
 	return 0;
@@ -1523,7 +1546,7 @@ static void mlx5_eswitch_get_devlink_param(struct mlx5_eswitch *esw)
 static void
 mlx5_eswitch_update_num_of_vfs(struct mlx5_eswitch *esw, int num_vfs)
 {
-	void *host_params;
+	struct mlx5_esw_pf_info host_pf_info;
 	const u32 *out;
 
 	if (num_vfs < 0)
@@ -1538,10 +1561,8 @@ mlx5_eswitch_update_num_of_vfs(struct mlx5_eswitch *esw, int num_vfs)
 	if (IS_ERR(out))
 		return;
 
-	host_params = MLX5_ADDR_OF(query_esw_functions_out, out,
-				   net_function_params);
-	esw->esw_funcs.num_vfs = MLX5_GET(host_params_context, host_params,
-					  host_num_of_vfs);
+	host_pf_info = mlx5_esw_get_host_pf_info(out);
+	esw->esw_funcs.num_vfs = host_pf_info.num_of_vfs;
 	if (mlx5_core_ec_sriov_enabled(esw->dev))
 		esw->esw_funcs.num_ec_vfs = num_vfs;
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index 291271afa96c..cfaae59a6e7c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -71,6 +71,14 @@ struct mlx5_mapped_obj {
 	};
 };
 
+struct mlx5_esw_pf_info {
+	bool pf_not_exist;
+	bool pf_disabled;
+	u16 num_of_vfs;
+	u16 total_vfs;
+	u16 host_number;
+};
+
 #ifdef CONFIG_MLX5_ESWITCH
 
 #define ESW_OFFLOADS_DEFAULT_NUM_GROUPS 15
@@ -649,6 +657,7 @@ bool mlx5_esw_multipath_prereq(struct mlx5_core_dev *dev0,
 			       struct mlx5_core_dev *dev1);
 
 const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev);
+struct mlx5_esw_pf_info mlx5_esw_get_host_pf_info(const u32 *out);
 int mlx5_esw_host_pf_enable_hca(struct mlx5_core_dev *dev);
 int mlx5_esw_host_pf_disable_hca(struct mlx5_core_dev *dev);
 
@@ -976,6 +985,12 @@ static inline const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev)
 	return ERR_PTR(-EOPNOTSUPP);
 }
 
+static inline struct mlx5_esw_pf_info
+mlx5_esw_get_host_pf_info(const u32 *out)
+{
+	return (struct mlx5_esw_pf_info) {};
+}
+
 static inline struct mlx5_flow_handle *
 esw_add_restore_rule(struct mlx5_eswitch *esw, u32 tag)
 {
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index d95af87a4f5f..217c2fe6b690 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -3708,8 +3708,7 @@ static void esw_offloads_steering_cleanup(struct mlx5_eswitch *esw)
 
 static void esw_vfs_changed_event_handler(struct mlx5_eswitch *esw)
 {
-	bool host_pf_disabled;
-	void *host_params;
+	struct mlx5_esw_pf_info host_pf_info;
 	u16 new_num_vfs;
 	const u32 *out;
 
@@ -3717,14 +3716,10 @@ static void esw_vfs_changed_event_handler(struct mlx5_eswitch *esw)
 	if (IS_ERR(out))
 		return;
 
-	host_params = MLX5_ADDR_OF(query_esw_functions_out, out,
-				   net_function_params);
-	new_num_vfs = MLX5_GET(host_params_context, host_params,
-			       host_num_of_vfs);
-	host_pf_disabled = MLX5_GET(host_params_context, host_params,
-				    host_pf_disabled);
+	host_pf_info = mlx5_esw_get_host_pf_info(out);
+	new_num_vfs = host_pf_info.num_of_vfs;
 
-	if (new_num_vfs == esw->esw_funcs.num_vfs || host_pf_disabled)
+	if (new_num_vfs == esw->esw_funcs.num_vfs || host_pf_info.pf_disabled)
 		goto free;
 
 	mlx5_esw_reps_block(esw);
@@ -3826,8 +3821,8 @@ int mlx5_esw_funcs_changed_handler(struct notifier_block *nb,
 
 static int mlx5_esw_host_number_init(struct mlx5_eswitch *esw)
 {
+	struct mlx5_esw_pf_info host_pf_info;
 	const u32 *query_host_out;
-	void *host_params;
 
 	if (!mlx5_core_is_ecpf_esw_manager(esw->dev))
 		return 0;
@@ -3837,10 +3832,8 @@ static int mlx5_esw_host_number_init(struct mlx5_eswitch *esw)
 		return PTR_ERR(query_host_out);
 
 	/* Mark non local controller with non zero controller number. */
-	host_params = MLX5_ADDR_OF(query_esw_functions_out,
-				   query_host_out, net_function_params);
-	esw->offloads.host_number = MLX5_GET(host_params_context,
-					     host_params, host_number);
+	host_pf_info = mlx5_esw_get_host_pf_info(query_host_out);
+	esw->offloads.host_number = host_pf_info.host_number;
 	kvfree(query_host_out);
 	return 0;
 }
@@ -4980,9 +4973,8 @@ int mlx5_devlink_pf_port_fn_state_get(struct devlink_port *port,
 				      struct netlink_ext_ack *extack)
 {
 	struct mlx5_vport *vport = mlx5_devlink_port_vport_get(port);
+	struct mlx5_esw_pf_info host_pf_info;
 	const u32 *query_out;
-	void *host_params;
-	bool pf_disabled;
 
 	if (vport->vport != MLX5_VPORT_HOST_PF) {
 		NL_SET_ERR_MSG_MOD(extack, "State get is not supported for VF");
@@ -4996,13 +4988,11 @@ int mlx5_devlink_pf_port_fn_state_get(struct devlink_port *port,
 	if (IS_ERR(query_out))
 		return PTR_ERR(query_out);
 
-	host_params = MLX5_ADDR_OF(query_esw_functions_out, query_out,
-				   net_function_params);
-	pf_disabled = MLX5_GET(host_params_context, host_params,
-			       host_pf_disabled);
+	host_pf_info = mlx5_esw_get_host_pf_info(query_out);
 
-	*opstate = pf_disabled ? DEVLINK_PORT_FN_OPSTATE_DETACHED :
-				 DEVLINK_PORT_FN_OPSTATE_ATTACHED;
+	*opstate = host_pf_info.pf_disabled ?
+			DEVLINK_PORT_FN_OPSTATE_DETACHED :
+			DEVLINK_PORT_FN_OPSTATE_ATTACHED;
 
 	kvfree(query_out);
 	return 0;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
index 6eb6026eadd6..79f76c456d72 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
@@ -273,8 +273,8 @@ void mlx5_sriov_detach(struct mlx5_core_dev *dev)
 
 static u16 mlx5_get_max_vfs(struct mlx5_core_dev *dev)
 {
+	struct mlx5_esw_pf_info host_pf_info;
 	u16 host_total_vfs;
-	void *host_params;
 	const u32 *out;
 
 	if (mlx5_core_is_ecpf_esw_manager(dev)) {
@@ -285,10 +285,8 @@ static u16 mlx5_get_max_vfs(struct mlx5_core_dev *dev)
 		 */
 		if (IS_ERR(out))
 			goto done;
-		host_params = MLX5_ADDR_OF(query_esw_functions_out, out,
-					   net_function_params);
-		host_total_vfs = MLX5_GET(host_params_context, host_params,
-					  host_total_vfs);
+		host_pf_info = mlx5_esw_get_host_pf_info(out);
+		host_total_vfs = host_pf_info.total_vfs;
 		kvfree(out);
 		return host_total_vfs;
 	}
-- 
2.44.0

[PATCH net-next V2 2/8] net/mlx5: Use v1 response layout for query_esw_functions

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

Use the v1 response layout for the query_esw_functions command when
supported by the device. When query_host_net_function_v1 capability is
set, use MLX5_QUERY_ESW_FUNC_OP_MOD_LAYOUT_V1 to retrieve parameters
for multiple network functions, allocating the output buffer according
to query_host_net_function_num_max. Validate that firmware does not
return more entries than the allocated buffer.

The v1 layout reports vhca_state instead of the legacy host_pf_disabled
bit. PFs transition through ALLOCATED, ACTIVE, and IN_USE states (they
do not use TEARDOWN_REQUEST as SFs do). When the ECPF calls disable_hca,
firmware resets the PF and moves it to ALLOCATED. When the ECPF calls
enable_hca, the PF moves to ACTIVE, and once the PF driver enables it,
it reaches IN_USE. The PF is only fully operational in IN_USE, so
pf_disabled is derived as vhca_state != IN_USE, equivalent to the legacy
host_pf_disabled bit.

The mlx5_esw_get_host_pf_info() helper abstracts parsing the command
output in both legacy and new formats, so callers do not need to handle
the different layouts.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/eswitch.c | 88 +++++++++++++++++--
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |  5 +-
 .../mellanox/mlx5/core/eswitch_offloads.c     |  6 +-
 .../net/ethernet/mellanox/mlx5/core/sriov.c   |  2 +-
 4 files changed, 89 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index 861e79ddb489..8b62dde7eb70 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -1063,11 +1063,28 @@ static int eswitch_vport_event(struct notifier_block *nb,
  */
 const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev)
 {
-	int outlen = MLX5_ST_SZ_BYTES(query_esw_functions_out);
+	bool net_func_v1 = MLX5_CAP_GEN(dev, query_host_net_function_v1);
 	u32 in[MLX5_ST_SZ_DW(query_esw_functions_in)] = {};
+	int alloc_entries;
+	int outlen;
 	u32 *out;
 	int err;
 
+	if (net_func_v1) {
+		alloc_entries = MLX5_CAP_GEN(dev,
+					     query_host_net_function_num_max);
+		alloc_entries = max(alloc_entries, 1);
+		MLX5_SET(query_esw_functions_in, in, op_mod,
+			 MLX5_QUERY_ESW_FUNC_OP_MOD_LAYOUT_V1);
+		outlen = MLX5_BYTE_OFF(query_esw_functions_out,
+				       net_function_params) +
+			 alloc_entries * MLX5_UN_SZ_BYTES(net_function_params);
+		outlen = max_t(int, outlen,
+			       MLX5_ST_SZ_BYTES(query_esw_functions_out));
+	} else {
+		outlen = MLX5_ST_SZ_BYTES(query_esw_functions_out);
+	}
+
 	out = kvzalloc(outlen, GFP_KERNEL);
 	if (!out)
 		return ERR_PTR(-ENOMEM);
@@ -1076,9 +1093,25 @@ const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev)
 		 MLX5_CMD_OP_QUERY_ESW_FUNCTIONS);
 
 	err = mlx5_cmd_exec(dev, in, sizeof(in), out, outlen);
-	if (!err)
-		return out;
+	if (err)
+		goto free;
+
+	if (net_func_v1) {
+		int num_entries;
+
+		num_entries = MLX5_GET(query_esw_functions_out, out,
+				       net_function_num);
+		if (num_entries > alloc_entries) {
+			mlx5_core_warn(dev, "Got %d entries, max expected %d\n",
+				       num_entries, alloc_entries);
+			err = -EINVAL;
+			goto free;
+		}
+	}
+
+	return out;
 
+free:
 	kvfree(out);
 	return ERR_PTR(err);
 }
@@ -1100,12 +1133,55 @@ mlx5_esw_host_pf_from_host_params(const void *entry)
 	};
 }
 
-struct mlx5_esw_pf_info mlx5_esw_get_host_pf_info(const u32 *out)
+static struct mlx5_esw_pf_info
+mlx5_esw_host_pf_from_net_func_params(const u8 *entry, int num_entries)
+{
+	int i;
+
+	for (i = 0; i < num_entries; i++) {
+		int pf_type, state;
+
+		pf_type = MLX5_GET(network_function_params, entry, pci_pf_type);
+		if (pf_type != MLX5_PCI_PF_TYPE_EXTERNAL_HOST_PF) {
+			entry += MLX5_UN_SZ_BYTES(net_function_params);
+			continue;
+		}
+
+		state = MLX5_GET(network_function_params, entry, vhca_state);
+
+		return (struct mlx5_esw_pf_info) {
+			.pf_disabled = state != MLX5_VHCA_STATE_IN_USE,
+			.num_of_vfs = MLX5_GET(network_function_params,
+					       entry, pci_num_vfs),
+			.total_vfs = MLX5_GET(network_function_params,
+					      entry, pci_total_vfs),
+			.host_number = MLX5_GET(network_function_params,
+						entry, host_number),
+		};
+	}
+
+	/* No external host PF entry found */
+	return (struct mlx5_esw_pf_info) {
+		.pf_not_exist = true,
+		.pf_disabled = true,
+	};
+}
+
+struct mlx5_esw_pf_info
+mlx5_esw_get_host_pf_info(struct mlx5_core_dev *dev, const u32 *out)
 {
 	const void *entry;
 
 	entry = MLX5_ADDR_OF(query_esw_functions_out, out, net_function_params);
 
+	if (MLX5_CAP_GEN(dev, query_host_net_function_v1)) {
+		int num_entries = MLX5_GET(query_esw_functions_out, out,
+					   net_function_num);
+
+		return mlx5_esw_host_pf_from_net_func_params(entry,
+							     num_entries);
+	}
+
 	return mlx5_esw_host_pf_from_host_params(entry);
 }
 
@@ -1121,7 +1197,7 @@ static int mlx5_esw_host_functions_enabled_query(struct mlx5_eswitch *esw)
 	if (IS_ERR(query_host_out))
 		return PTR_ERR(query_host_out);
 
-	host_pf_info = mlx5_esw_get_host_pf_info(query_host_out);
+	host_pf_info = mlx5_esw_get_host_pf_info(esw->dev, query_host_out);
 	esw->esw_funcs.host_funcs_disabled = host_pf_info.pf_not_exist;
 
 	kvfree(query_host_out);
@@ -1561,7 +1637,7 @@ mlx5_eswitch_update_num_of_vfs(struct mlx5_eswitch *esw, int num_vfs)
 	if (IS_ERR(out))
 		return;
 
-	host_pf_info = mlx5_esw_get_host_pf_info(out);
+	host_pf_info = mlx5_esw_get_host_pf_info(esw->dev, out);
 	esw->esw_funcs.num_vfs = host_pf_info.num_of_vfs;
 	if (mlx5_core_ec_sriov_enabled(esw->dev))
 		esw->esw_funcs.num_ec_vfs = num_vfs;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index cfaae59a6e7c..a5f832ed2251 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -657,7 +657,8 @@ bool mlx5_esw_multipath_prereq(struct mlx5_core_dev *dev0,
 			       struct mlx5_core_dev *dev1);
 
 const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev);
-struct mlx5_esw_pf_info mlx5_esw_get_host_pf_info(const u32 *out);
+struct mlx5_esw_pf_info mlx5_esw_get_host_pf_info(struct mlx5_core_dev *dev,
+						  const u32 *out);
 int mlx5_esw_host_pf_enable_hca(struct mlx5_core_dev *dev);
 int mlx5_esw_host_pf_disable_hca(struct mlx5_core_dev *dev);
 
@@ -986,7 +987,7 @@ static inline const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev)
 }
 
 static inline struct mlx5_esw_pf_info
-mlx5_esw_get_host_pf_info(const u32 *out)
+mlx5_esw_get_host_pf_info(struct mlx5_core_dev *dev, const u32 *out)
 {
 	return (struct mlx5_esw_pf_info) {};
 }
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index 217c2fe6b690..acbc37b05308 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -3716,7 +3716,7 @@ static void esw_vfs_changed_event_handler(struct mlx5_eswitch *esw)
 	if (IS_ERR(out))
 		return;
 
-	host_pf_info = mlx5_esw_get_host_pf_info(out);
+	host_pf_info = mlx5_esw_get_host_pf_info(esw->dev, out);
 	new_num_vfs = host_pf_info.num_of_vfs;
 
 	if (new_num_vfs == esw->esw_funcs.num_vfs || host_pf_info.pf_disabled)
@@ -3832,7 +3832,7 @@ static int mlx5_esw_host_number_init(struct mlx5_eswitch *esw)
 		return PTR_ERR(query_host_out);
 
 	/* Mark non local controller with non zero controller number. */
-	host_pf_info = mlx5_esw_get_host_pf_info(query_host_out);
+	host_pf_info = mlx5_esw_get_host_pf_info(esw->dev, query_host_out);
 	esw->offloads.host_number = host_pf_info.host_number;
 	kvfree(query_host_out);
 	return 0;
@@ -4988,7 +4988,7 @@ int mlx5_devlink_pf_port_fn_state_get(struct devlink_port *port,
 	if (IS_ERR(query_out))
 		return PTR_ERR(query_out);
 
-	host_pf_info = mlx5_esw_get_host_pf_info(query_out);
+	host_pf_info = mlx5_esw_get_host_pf_info(vport->dev, query_out);
 
 	*opstate = host_pf_info.pf_disabled ?
 			DEVLINK_PORT_FN_OPSTATE_DETACHED :
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
index 79f76c456d72..0770b5d99c5d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/sriov.c
@@ -285,7 +285,7 @@ static u16 mlx5_get_max_vfs(struct mlx5_core_dev *dev)
 		 */
 		if (IS_ERR(out))
 			goto done;
-		host_pf_info = mlx5_esw_get_host_pf_info(out);
+		host_pf_info = mlx5_esw_get_host_pf_info(dev, out);
 		host_total_vfs = host_pf_info.total_vfs;
 		kvfree(out);
 		return host_total_vfs;
-- 
2.44.0

[PATCH net-next V2 3/8] net/mlx5: Use mlx5_eswitch_is_vf_vport() for IPsec VF checks

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

IPsec eswitch offload operations and the enabled_ipsec_vf_count counter
are intended for VF vports only. Replace the MLX5_VPORT_HOST_PF checks
with mlx5_eswitch_is_vf_vport() to properly identify VF vports, as
preparation for adding another type of PF vports.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c | 2 +-
 drivers/net/ethernet/mellanox/mlx5/core/eswitch.c   | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
index 4811b60ea430..b830ccd91e62 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
@@ -209,7 +209,7 @@ static int esw_ipsec_vf_offload_set_bytype(struct mlx5_eswitch *esw, struct mlx5
 	struct mlx5_core_dev *dev = esw->dev;
 	int err;
 
-	if (vport->vport == MLX5_VPORT_HOST_PF)
+	if (!mlx5_eswitch_is_vf_vport(esw, vport->vport))
 		return -EOPNOTSUPP;
 
 	if (type == MLX5_ESW_VPORT_IPSEC_CRYPTO_OFFLOAD) {
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index 8b62dde7eb70..9a7de7c9a667 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -958,7 +958,7 @@ int mlx5_esw_vport_enable(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
 	/* Sync with current vport context */
 	vport->enabled_events = enabled_events;
 	vport->enabled = true;
-	if (vport->vport != MLX5_VPORT_HOST_PF &&
+	if (mlx5_eswitch_is_vf_vport(esw, vport_num) &&
 	    (vport->info.ipsec_crypto_enabled || vport->info.ipsec_packet_enabled))
 		esw->enabled_ipsec_vf_count++;
 
@@ -1020,7 +1020,7 @@ void mlx5_esw_vport_disable(struct mlx5_eswitch *esw, struct mlx5_vport *vport)
 		mlx5_esw_vport_vhca_id_unmap(esw, vport);
 	}
 
-	if (vport->vport != MLX5_VPORT_HOST_PF &&
+	if (mlx5_eswitch_is_vf_vport(esw, vport_num) &&
 	    (vport->info.ipsec_crypto_enabled || vport->info.ipsec_packet_enabled))
 		esw->enabled_ipsec_vf_count--;
 
-- 
2.44.0

Re: [PATCH net-next V2 3/8] net/mlx5: Use mlx5_eswitch_is_vf_vport() for IPsec VF checks

[Moshe Shemesh]

On 5/18/2026 10:13 AM, Tariq Toukan wrote:
> From: Moshe Shemesh <moshe@nvidia.com>
> 
> IPsec eswitch offload operations and the enabled_ipsec_vf_count counter
> are intended for VF vports only. Replace the MLX5_VPORT_HOST_PF checks
> with mlx5_eswitch_is_vf_vport() to properly identify VF vports, as
> preparation for adding another type of PF vports.
> 
> Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>   drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c | 2 +-
>   drivers/net/ethernet/mellanox/mlx5/core/eswitch.c   | 4 ++--
>   2 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
> index 4811b60ea430..b830ccd91e62 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
> @@ -209,7 +209,7 @@ static int esw_ipsec_vf_offload_set_bytype(struct mlx5_eswitch *esw, struct mlx5
>   	struct mlx5_core_dev *dev = esw->dev;
>   	int err;
>   
> -	if (vport->vport == MLX5_VPORT_HOST_PF)
> +	if (!mlx5_eswitch_is_vf_vport(esw, vport->vport))
>   		return -EOPNOTSUPP;
>   
>   	if (type == MLX5_ESW_VPORT_IPSEC_CRYPTO_OFFLOAD) {
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
> index 8b62dde7eb70..9a7de7c9a667 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
> @@ -958,7 +958,7 @@ int mlx5_esw_vport_enable(struct mlx5_eswitch *esw, struct mlx5_vport *vport,
>   	/* Sync with current vport context */
>   	vport->enabled_events = enabled_events;
>   	vport->enabled = true;
> -	if (vport->vport != MLX5_VPORT_HOST_PF &&
> +	if (mlx5_eswitch_is_vf_vport(esw, vport_num) &&
>   	    (vport->info.ipsec_crypto_enabled || vport->info.ipsec_packet_enabled))
>   		esw->enabled_ipsec_vf_count++;
>   

Sashiko.dev says:
" This isn't a regression introduced by this patch, but is it possible 
for enabled_ipsec_vf_count to leak in the error path?"

As it says, pre-existing bug, not related to this patch change.
I will have a separate patch to fix that.

> @@ -1020,7 +1020,7 @@ void mlx5_esw_vport_disable(struct mlx5_eswitch *esw, struct mlx5_vport *vport)
>   		mlx5_esw_vport_vhca_id_unmap(esw, vport);
>   	}
>   
> -	if (vport->vport != MLX5_VPORT_HOST_PF &&
> +	if (mlx5_eswitch_is_vf_vport(esw, vport_num) &&
>   	    (vport->info.ipsec_crypto_enabled || vport->info.ipsec_packet_enabled))
>   		esw->enabled_ipsec_vf_count--;
>   

Sashiko.dev says:
" This isn't a regression introduced by this patch, but does this logic
handle the case where both crypto and packet offloads are enabled on a
single VF via devlink? "
	
Same here.

[PATCH net-next V2 4/8] net/mlx5: Switch vport HCA cap helpers to kvzalloc

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

mlx5_vport_set_other_func_cap() and mlx5_vport_get_vhca_id() allocate
command buffers that embed the HCA capability union, exceeding 4KiB.
Use kvzalloc/kvfree so the allocation can fall back to vmalloc when
contiguous memory is scarce.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/vport.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/vport.c b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
index 4effe37fd455..f8e6b1ab7c5c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/vport.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
@@ -1336,7 +1336,7 @@ int mlx5_vport_get_vhca_id(struct mlx5_core_dev *dev, u16 vport, u16 *vhca_id)
 	if (mlx5_esw_vport_vhca_id(dev->priv.eswitch, vport, vhca_id))
 		return 0;
 
-	query_ctx = kzalloc(query_out_sz, GFP_KERNEL);
+	query_ctx = kvzalloc(query_out_sz, GFP_KERNEL);
 	if (!query_ctx)
 		return -ENOMEM;
 
@@ -1348,7 +1348,7 @@ int mlx5_vport_get_vhca_id(struct mlx5_core_dev *dev, u16 vport, u16 *vhca_id)
 	*vhca_id = MLX5_GET(cmd_hca_cap, hca_caps, vhca_id);
 
 out_free:
-	kfree(query_ctx);
+	kvfree(query_ctx);
 	return err;
 }
 EXPORT_SYMBOL_GPL(mlx5_vport_get_vhca_id);
@@ -1363,7 +1363,7 @@ int mlx5_vport_set_other_func_cap(struct mlx5_core_dev *dev, const void *hca_cap
 	void *set_ctx;
 	int ret;
 
-	set_ctx = kzalloc(set_sz, GFP_KERNEL);
+	set_ctx = kvzalloc(set_sz, GFP_KERNEL);
 	if (!set_ctx)
 		return -ENOMEM;
 
@@ -1392,6 +1392,6 @@ int mlx5_vport_set_other_func_cap(struct mlx5_core_dev *dev, const void *hca_cap
 	MLX5_SET(set_hca_cap_in, set_ctx, function_id, function_id);
 	ret = mlx5_cmd_exec_in(dev, set_hca_cap, set_ctx);
 
-	kfree(set_ctx);
+	kvfree(set_ctx);
 	return ret;
 }
-- 
2.44.0

Re: [PATCH net-next V2 4/8] net/mlx5: Switch vport HCA cap helpers to kvzalloc

[Moshe Shemesh]

On 5/18/2026 10:13 AM, Tariq Toukan wrote:
> From: Moshe Shemesh <moshe@nvidia.com>
> 
> mlx5_vport_set_other_func_cap() and mlx5_vport_get_vhca_id() allocate
> command buffers that embed the HCA capability union, exceeding 4KiB.
> Use kvzalloc/kvfree so the allocation can fall back to vmalloc when
> contiguous memory is scarce.
> 
> Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>   drivers/net/ethernet/mellanox/mlx5/core/vport.c | 8 ++++----
>   1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/vport.c b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
> index 4effe37fd455..f8e6b1ab7c5c 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/vport.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
> @@ -1336,7 +1336,7 @@ int mlx5_vport_get_vhca_id(struct mlx5_core_dev *dev, u16 vport, u16 *vhca_id)
>   	if (mlx5_esw_vport_vhca_id(dev->priv.eswitch, vport, vhca_id))
>   		return 0;
>   
> -	query_ctx = kzalloc(query_out_sz, GFP_KERNEL);
> +	query_ctx = kvzalloc(query_out_sz, GFP_KERNEL);
>   	if (!query_ctx)
>   		return -ENOMEM;

Sashiko.dev says:
"
The commit message notes that these structures exceed 4KiB and are 
vulnerable to allocation failures under memory fragmentation. Are there 
other allocations of these same structures that were missed?
For instance, mlx5_core_get_caps_mode() in main.c allocates 
query_hca_cap_out with kzalloc(). Since this is on the initialization 
path for every mlx5 device (including dynamically created SR-IOV VFs), 
could this fail under fragmentation and prevent VF creation?
Other locations that still appear to use kzalloc() for these >4KiB 
structures include devlink configuration paths like 
mlx5_devlink_port_fn_migratable_set() and 
mlx5_devlink_port_fn_roce_set() in eswitch_offloads.c and eswitch.c.
Should these be converted to kvzalloc() as well to prevent similar 
allocation failures?
"

As it says, there are more places in the driver, this patch handles it 
for the vport. I will send cleanup patches for others.

>   
> @@ -1348,7 +1348,7 @@ int mlx5_vport_get_vhca_id(struct mlx5_core_dev *dev, u16 vport, u16 *vhca_id)
>   	*vhca_id = MLX5_GET(cmd_hca_cap, hca_caps, vhca_id);
>   
>   out_free:
> -	kfree(query_ctx);
> +	kvfree(query_ctx);
>   	return err;
>   }
>   EXPORT_SYMBOL_GPL(mlx5_vport_get_vhca_id);
> @@ -1363,7 +1363,7 @@ int mlx5_vport_set_other_func_cap(struct mlx5_core_dev *dev, const void *hca_cap
>   	void *set_ctx;
>   	int ret;
>   
> -	set_ctx = kzalloc(set_sz, GFP_KERNEL);
> +	set_ctx = kvzalloc(set_sz, GFP_KERNEL);
>   	if (!set_ctx)
>   		return -ENOMEM;
>   
> @@ -1392,6 +1392,6 @@ int mlx5_vport_set_other_func_cap(struct mlx5_core_dev *dev, const void *hca_cap
>   	MLX5_SET(set_hca_cap_in, set_ctx, function_id, function_id);
>   	ret = mlx5_cmd_exec_in(dev, set_hca_cap, set_ctx);
>   
> -	kfree(set_ctx);
> +	kvfree(set_ctx);
>   	return ret;
>   }

[PATCH net-next V2 5/8] net/mlx5: Add mlx5_vport_set_other_func_general_cap macro

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

Add mlx5_vport_set_other_func_general_cap() convenience macro, symmetric
to the existing mlx5_vport_get_other_func_general_cap(), and use it in
mlx5_devlink_port_fn_roce_set().

No functional change in this patch.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c | 4 ++--
 drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h        | 4 ++++
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
index acbc37b05308..b06b10d443bd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c
@@ -4951,8 +4951,8 @@ int mlx5_devlink_port_fn_roce_set(struct devlink_port *port, bool enable,
 	hca_caps = MLX5_ADDR_OF(query_hca_cap_out, query_ctx, capability);
 	MLX5_SET(cmd_hca_cap, hca_caps, roce, enable);
 
-	err = mlx5_vport_set_other_func_cap(esw->dev, hca_caps, vport_num,
-					    MLX5_SET_HCA_CAP_OP_MOD_GENERAL_DEVICE);
+	err = mlx5_vport_set_other_func_general_cap(esw->dev, hca_caps,
+						    vport_num);
 	if (err) {
 		NL_SET_ERR_MSG_MOD(extack, "Failed setting HCA roce cap");
 		goto out_free;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
index d70907f499a9..2eba141bd521 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
@@ -457,6 +457,10 @@ int mlx5_vport_set_other_func_cap(struct mlx5_core_dev *dev, const void *hca_cap
 #define mlx5_vport_get_other_func_general_cap(dev, vport, out)		\
 	mlx5_vport_get_other_func_cap(dev, vport, out, MLX5_CAP_GENERAL)
 
+#define mlx5_vport_set_other_func_general_cap(dev, hca_cap, vport)	\
+	mlx5_vport_set_other_func_cap(dev, hca_cap, vport,		\
+				      MLX5_SET_HCA_CAP_OP_MOD_GENERAL_DEVICE)
+
 static inline u32 mlx5_sriov_get_vf_total_msix(struct pci_dev *pdev)
 {
 	struct mlx5_core_dev *dev = pci_get_drvdata(pdev);
-- 
2.44.0

[PATCH net-next V2 6/8] net/mlx5: Refactor mlx5_set_msix_vec_count() SET_HCA_CAP

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

Use mlx5_vport_set_other_func_general_cap() instead of open-coding the
SET_HCA_CAP command. This removes redundant buffer allocation and
ensures consistent use of vport-based function addressing.

mlx5_vport_set_other_func_general_cap() supports both function_id and
vhca_id based addressing, so this also enables SET_HCA_CAP for vhca_id
indexed functions which was not supported before.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/pci_irq.c | 27 +++++--------------
 1 file changed, 7 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
index e051b9a939ee..0f5b8bc7861e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
@@ -87,9 +87,8 @@ int mlx5_set_msix_vec_count(struct mlx5_core_dev *dev, int function_id,
 			    int msix_vec_count)
 {
 	int query_sz = MLX5_ST_SZ_BYTES(query_hca_cap_out);
-	int set_sz = MLX5_ST_SZ_BYTES(set_hca_cap_in);
-	void *hca_cap = NULL, *query_cap = NULL, *cap;
 	int num_vf_msix, min_msix, max_msix;
+	void *query_cap, *hca_caps;
 	bool ec_vf_function;
 	int vport;
 	int ret;
@@ -111,11 +110,8 @@ int mlx5_set_msix_vec_count(struct mlx5_core_dev *dev, int function_id,
 		return -EOVERFLOW;
 
 	query_cap = kvzalloc(query_sz, GFP_KERNEL);
-	hca_cap = kvzalloc(set_sz, GFP_KERNEL);
-	if (!hca_cap || !query_cap) {
-		ret = -ENOMEM;
-		goto out;
-	}
+	if (!query_cap)
+		return -ENOMEM;
 
 	ec_vf_function = mlx5_core_ec_sriov_enabled(dev);
 	vport = mlx5_core_func_to_vport(dev, function_id, ec_vf_function);
@@ -123,21 +119,12 @@ int mlx5_set_msix_vec_count(struct mlx5_core_dev *dev, int function_id,
 	if (ret)
 		goto out;
 
-	cap = MLX5_ADDR_OF(set_hca_cap_in, hca_cap, capability);
-	memcpy(cap, MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability),
-	       MLX5_UN_SZ_BYTES(hca_cap_union));
-	MLX5_SET(cmd_hca_cap, cap, dynamic_msix_table_size, msix_vec_count);
-
-	MLX5_SET(set_hca_cap_in, hca_cap, opcode, MLX5_CMD_OP_SET_HCA_CAP);
-	MLX5_SET(set_hca_cap_in, hca_cap, other_function, 1);
-	MLX5_SET(set_hca_cap_in, hca_cap, ec_vf_function, ec_vf_function);
-	MLX5_SET(set_hca_cap_in, hca_cap, function_id, function_id);
+	hca_caps = MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability);
+	MLX5_SET(cmd_hca_cap, hca_caps, dynamic_msix_table_size,
+		 msix_vec_count);
 
-	MLX5_SET(set_hca_cap_in, hca_cap, op_mod,
-		 MLX5_SET_HCA_CAP_OP_MOD_GENERAL_DEVICE << 1);
-	ret = mlx5_cmd_exec_in(dev, set_hca_cap, hca_cap);
+	ret = mlx5_vport_set_other_func_general_cap(dev, hca_caps, vport);
 out:
-	kvfree(hca_cap);
 	kvfree(query_cap);
 	return ret;
 }
-- 
2.44.0

[PATCH net-next V2 7/8] net/mlx5: Use vport helper for IPsec eswitch set caps

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

Use mlx5_vport_set_other_func_cap() and
mlx5_vport_set_other_func_general_cap() in the IPsec eswitch functions
instead of open-coding the SET_HCA_CAP command. This removes redundant
buffer allocation and boilerplate, and also enables vhca_id based
addressing when supported.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 .../ethernet/mellanox/mlx5/core/esw/ipsec.c   | 81 ++++++-------------
 1 file changed, 23 insertions(+), 58 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c b/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
index b830ccd91e62..2b5765ab60d1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/esw/ipsec.c
@@ -81,38 +81,25 @@ int mlx5_esw_ipsec_vf_offload_get(struct mlx5_core_dev *dev, struct mlx5_vport *
 static int esw_ipsec_vf_set_generic(struct mlx5_core_dev *dev, u16 vport_num, bool ipsec_ofld)
 {
 	int query_sz = MLX5_ST_SZ_BYTES(query_hca_cap_out);
-	int set_sz = MLX5_ST_SZ_BYTES(set_hca_cap_in);
-	void *hca_cap, *query_cap, *cap;
+	void *query_cap, *hca_caps;
 	int ret;
 
 	if (!MLX5_CAP_GEN(dev, vhca_resource_manager))
 		return -EOPNOTSUPP;
 
 	query_cap = kvzalloc(query_sz, GFP_KERNEL);
-	hca_cap = kvzalloc(set_sz, GFP_KERNEL);
-	if (!hca_cap || !query_cap) {
-		ret = -ENOMEM;
-		goto free;
-	}
+	if (!query_cap)
+		return -ENOMEM;
 
 	ret = mlx5_vport_get_other_func_general_cap(dev, vport_num, query_cap);
 	if (ret)
 		goto free;
 
-	cap = MLX5_ADDR_OF(set_hca_cap_in, hca_cap, capability);
-	memcpy(cap, MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability),
-	       MLX5_UN_SZ_BYTES(hca_cap_union));
-	MLX5_SET(cmd_hca_cap, cap, ipsec_offload, ipsec_ofld);
+	hca_caps = MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability);
+	MLX5_SET(cmd_hca_cap, hca_caps, ipsec_offload, ipsec_ofld);
 
-	MLX5_SET(set_hca_cap_in, hca_cap, opcode, MLX5_CMD_OP_SET_HCA_CAP);
-	MLX5_SET(set_hca_cap_in, hca_cap, other_function, 1);
-	MLX5_SET(set_hca_cap_in, hca_cap, function_id, vport_num);
-
-	MLX5_SET(set_hca_cap_in, hca_cap, op_mod,
-		 MLX5_SET_HCA_CAP_OP_MOD_GENERAL_DEVICE << 1);
-	ret = mlx5_cmd_exec_in(dev, set_hca_cap, hca_cap);
+	ret = mlx5_vport_set_other_func_general_cap(dev, hca_caps, vport_num);
 free:
-	kvfree(hca_cap);
 	kvfree(query_cap);
 	return ret;
 }
@@ -121,49 +108,37 @@ static int esw_ipsec_vf_set_bytype(struct mlx5_core_dev *dev, struct mlx5_vport
 				   bool enable, enum esw_vport_ipsec_offload type)
 {
 	int query_sz = MLX5_ST_SZ_BYTES(query_hca_cap_out);
-	int set_sz = MLX5_ST_SZ_BYTES(set_hca_cap_in);
-	void *hca_cap, *query_cap, *cap;
+	void *query_cap, *hca_caps;
 	int ret;
 
 	if (!MLX5_CAP_GEN(dev, vhca_resource_manager))
 		return -EOPNOTSUPP;
 
 	query_cap = kvzalloc(query_sz, GFP_KERNEL);
-	hca_cap = kvzalloc(set_sz, GFP_KERNEL);
-	if (!hca_cap || !query_cap) {
-		ret = -ENOMEM;
-		goto free;
-	}
+	if (!query_cap)
+		return -ENOMEM;
 
 	ret = mlx5_vport_get_other_func_cap(dev, vport->vport, query_cap, MLX5_CAP_IPSEC);
 	if (ret)
 		goto free;
 
-	cap = MLX5_ADDR_OF(set_hca_cap_in, hca_cap, capability);
-	memcpy(cap, MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability),
-	       MLX5_UN_SZ_BYTES(hca_cap_union));
+	hca_caps = MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability);
 
 	switch (type) {
 	case MLX5_ESW_VPORT_IPSEC_CRYPTO_OFFLOAD:
-		MLX5_SET(ipsec_cap, cap, ipsec_crypto_offload, enable);
+		MLX5_SET(ipsec_cap, hca_caps, ipsec_crypto_offload, enable);
 		break;
 	case MLX5_ESW_VPORT_IPSEC_PACKET_OFFLOAD:
-		MLX5_SET(ipsec_cap, cap, ipsec_full_offload, enable);
+		MLX5_SET(ipsec_cap, hca_caps, ipsec_full_offload, enable);
 		break;
 	default:
 		ret = -EOPNOTSUPP;
 		goto free;
 	}
 
-	MLX5_SET(set_hca_cap_in, hca_cap, opcode, MLX5_CMD_OP_SET_HCA_CAP);
-	MLX5_SET(set_hca_cap_in, hca_cap, other_function, 1);
-	MLX5_SET(set_hca_cap_in, hca_cap, function_id, vport->vport);
-
-	MLX5_SET(set_hca_cap_in, hca_cap, op_mod,
-		 MLX5_SET_HCA_CAP_OP_MOD_IPSEC << 1);
-	ret = mlx5_cmd_exec_in(dev, set_hca_cap, hca_cap);
+	ret = mlx5_vport_set_other_func_cap(dev, hca_caps, vport->vport,
+					    MLX5_SET_HCA_CAP_OP_MOD_IPSEC);
 free:
-	kvfree(hca_cap);
 	kvfree(query_cap);
 	return ret;
 }
@@ -171,34 +146,24 @@ static int esw_ipsec_vf_set_bytype(struct mlx5_core_dev *dev, struct mlx5_vport
 static int esw_ipsec_vf_crypto_aux_caps_set(struct mlx5_core_dev *dev, u16 vport_num, bool enable)
 {
 	int query_sz = MLX5_ST_SZ_BYTES(query_hca_cap_out);
-	int set_sz = MLX5_ST_SZ_BYTES(set_hca_cap_in);
-	struct mlx5_eswitch *esw = dev->priv.eswitch;
-	void *hca_cap, *query_cap, *cap;
+	void *query_cap, *hca_caps;
 	int ret;
 
 	query_cap = kvzalloc(query_sz, GFP_KERNEL);
-	hca_cap = kvzalloc(set_sz, GFP_KERNEL);
-	if (!hca_cap || !query_cap) {
-		ret = -ENOMEM;
-		goto free;
-	}
+	if (!query_cap)
+		return -ENOMEM;
 
 	ret = mlx5_vport_get_other_func_cap(dev, vport_num, query_cap, MLX5_CAP_ETHERNET_OFFLOADS);
 	if (ret)
 		goto free;
 
-	cap = MLX5_ADDR_OF(set_hca_cap_in, hca_cap, capability);
-	memcpy(cap, MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability),
-	       MLX5_UN_SZ_BYTES(hca_cap_union));
-	MLX5_SET(per_protocol_networking_offload_caps, cap, insert_trailer, enable);
-	MLX5_SET(set_hca_cap_in, hca_cap, opcode, MLX5_CMD_OP_SET_HCA_CAP);
-	MLX5_SET(set_hca_cap_in, hca_cap, other_function, 1);
-	MLX5_SET(set_hca_cap_in, hca_cap, function_id, vport_num);
-	MLX5_SET(set_hca_cap_in, hca_cap, op_mod,
-		 MLX5_SET_HCA_CAP_OP_MOD_ETHERNET_OFFLOADS << 1);
-	ret = mlx5_cmd_exec_in(esw->dev, set_hca_cap, hca_cap);
+	hca_caps = MLX5_ADDR_OF(query_hca_cap_out, query_cap, capability);
+	MLX5_SET(per_protocol_networking_offload_caps, hca_caps,
+		 insert_trailer, enable);
+
+	ret = mlx5_vport_set_other_func_cap(dev, hca_caps, vport_num,
+					    MLX5_SET_HCA_CAP_OP_MOD_ETHERNET_OFFLOADS);
 free:
-	kvfree(hca_cap);
 	kvfree(query_cap);
 	return ret;
 }
-- 
2.44.0

[PATCH net-next V2 8/8] net/mlx5: Generalize enable/disable HCA for any PF vport

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

Refactor the host-PF-specific mlx5_cmd_host_pf_enable/disable_hca()
into generic mlx5_cmd_pf_enable/disable_hca() that accept a vport
number. The new functions use vhca_id as function_id when supported.

Similarly, refactor the eswitch layer into generic static helpers
mlx5_esw_pf_enable/disable_hca() with thin wrappers for the host PF
case, in preparation for enable_hca on satellite PF vports.

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/ecpf.c    | 24 +++++++++++-----
 .../net/ethernet/mellanox/mlx5/core/ecpf.h    |  4 +--
 .../net/ethernet/mellanox/mlx5/core/eswitch.c | 28 +++++++++++++------
 .../ethernet/mellanox/mlx5/core/mlx5_core.h   |  2 ++
 .../net/ethernet/mellanox/mlx5/core/vport.c   |  4 +--
 5 files changed, 42 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/ecpf.c b/drivers/net/ethernet/mellanox/mlx5/core/ecpf.c
index 15cb27aea2c9..350c47d3643b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/ecpf.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/ecpf.c
@@ -18,25 +18,35 @@ static bool mlx5_ecpf_esw_admins_host_pf(const struct mlx5_core_dev *dev)
 	return mlx5_core_is_ecpf_esw_manager(dev);
 }
 
-int mlx5_cmd_host_pf_enable_hca(struct mlx5_core_dev *dev)
+int mlx5_cmd_pf_enable_hca(struct mlx5_core_dev *dev, u16 vport_num)
 {
 	u32 out[MLX5_ST_SZ_DW(enable_hca_out)] = {};
 	u32 in[MLX5_ST_SZ_DW(enable_hca_in)]   = {};
+	u16 vhca_id;
 
 	MLX5_SET(enable_hca_in, in, opcode, MLX5_CMD_OP_ENABLE_HCA);
-	MLX5_SET(enable_hca_in, in, function_id, 0);
-	MLX5_SET(enable_hca_in, in, embedded_cpu_function, 0);
-	return mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
+	if (mlx5_vport_use_vhca_id_as_func_id(dev, vport_num, &vhca_id)) {
+		MLX5_SET(enable_hca_in, in, function_id, vhca_id);
+		MLX5_SET(enable_hca_in, in, function_id_type, 1);
+	} else {
+		MLX5_SET(enable_hca_in, in, function_id, vport_num);
+	}
+	return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
 }
 
-int mlx5_cmd_host_pf_disable_hca(struct mlx5_core_dev *dev)
+int mlx5_cmd_pf_disable_hca(struct mlx5_core_dev *dev, u16 vport_num)
 {
 	u32 out[MLX5_ST_SZ_DW(disable_hca_out)] = {};
 	u32 in[MLX5_ST_SZ_DW(disable_hca_in)]   = {};
+	u16 vhca_id;
 
 	MLX5_SET(disable_hca_in, in, opcode, MLX5_CMD_OP_DISABLE_HCA);
-	MLX5_SET(disable_hca_in, in, function_id, 0);
-	MLX5_SET(disable_hca_in, in, embedded_cpu_function, 0);
+	if (mlx5_vport_use_vhca_id_as_func_id(dev, vport_num, &vhca_id)) {
+		MLX5_SET(disable_hca_in, in, function_id, vhca_id);
+		MLX5_SET(disable_hca_in, in, function_id_type, 1);
+	} else {
+		MLX5_SET(disable_hca_in, in, function_id, vport_num);
+	}
 	return mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/ecpf.h b/drivers/net/ethernet/mellanox/mlx5/core/ecpf.h
index 40b6ad76dca6..d9f9a53b019b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/ecpf.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/ecpf.h
@@ -17,8 +17,8 @@ bool mlx5_read_embedded_cpu(struct mlx5_core_dev *dev);
 int mlx5_ec_init(struct mlx5_core_dev *dev);
 void mlx5_ec_cleanup(struct mlx5_core_dev *dev);
 
-int mlx5_cmd_host_pf_enable_hca(struct mlx5_core_dev *dev);
-int mlx5_cmd_host_pf_disable_hca(struct mlx5_core_dev *dev);
+int mlx5_cmd_pf_enable_hca(struct mlx5_core_dev *dev, u16 vport_num);
+int mlx5_cmd_pf_disable_hca(struct mlx5_core_dev *dev, u16 vport_num);
 
 #else  /* CONFIG_MLX5_ESWITCH */
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
index 9a7de7c9a667..206911817a04 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.c
@@ -1452,7 +1452,7 @@ static int mlx5_eswitch_load_ec_vf_vports(struct mlx5_eswitch *esw, u16 num_ec_v
 	return err;
 }
 
-int mlx5_esw_host_pf_enable_hca(struct mlx5_core_dev *dev)
+static int mlx5_esw_pf_enable_hca(struct mlx5_core_dev *dev, u16 vport_num)
 {
 	struct mlx5_eswitch *esw = dev->priv.eswitch;
 	struct mlx5_vport *vport;
@@ -1461,15 +1461,15 @@ int mlx5_esw_host_pf_enable_hca(struct mlx5_core_dev *dev)
 	if (!mlx5_core_is_ecpf(dev) || !mlx5_esw_allowed(esw))
 		return 0;
 
-	vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_HOST_PF);
+	vport = mlx5_eswitch_get_vport(esw, vport_num);
 	if (IS_ERR(vport))
 		return PTR_ERR(vport);
 
-	/* Once vport and representor are ready, take out the external host PF
-	 * out of initializing state. Enabling HCA clears the iser->initializing
-	 * bit and host PF driver loading can progress.
+	/* Once vport and representor are ready, take the PF out of
+	 * initializing state. Enabling HCA clears the iser->initializing
+	 * bit and PF driver loading can progress.
 	 */
-	err = mlx5_cmd_host_pf_enable_hca(dev);
+	err = mlx5_cmd_pf_enable_hca(dev, vport_num);
 	if (err)
 		return err;
 
@@ -1478,7 +1478,7 @@ int mlx5_esw_host_pf_enable_hca(struct mlx5_core_dev *dev)
 	return 0;
 }
 
-int mlx5_esw_host_pf_disable_hca(struct mlx5_core_dev *dev)
+static int mlx5_esw_pf_disable_hca(struct mlx5_core_dev *dev, u16 vport_num)
 {
 	struct mlx5_eswitch *esw = dev->priv.eswitch;
 	struct mlx5_vport *vport;
@@ -1487,11 +1487,11 @@ int mlx5_esw_host_pf_disable_hca(struct mlx5_core_dev *dev)
 	if (!mlx5_core_is_ecpf(dev) || !mlx5_esw_allowed(esw))
 		return 0;
 
-	vport = mlx5_eswitch_get_vport(esw, MLX5_VPORT_HOST_PF);
+	vport = mlx5_eswitch_get_vport(esw, vport_num);
 	if (IS_ERR(vport))
 		return PTR_ERR(vport);
 
-	err = mlx5_cmd_host_pf_disable_hca(dev);
+	err = mlx5_cmd_pf_disable_hca(dev, vport_num);
 	if (err)
 		return err;
 
@@ -1500,6 +1500,16 @@ int mlx5_esw_host_pf_disable_hca(struct mlx5_core_dev *dev)
 	return 0;
 }
 
+int mlx5_esw_host_pf_enable_hca(struct mlx5_core_dev *dev)
+{
+	return mlx5_esw_pf_enable_hca(dev, MLX5_VPORT_HOST_PF);
+}
+
+int mlx5_esw_host_pf_disable_hca(struct mlx5_core_dev *dev)
+{
+	return mlx5_esw_pf_disable_hca(dev, MLX5_VPORT_HOST_PF);
+}
+
 /* mlx5_eswitch_enable_pf_vf_vports() enables vports of PF, ECPF and VFs
  * whichever are present on the eswitch.
  */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
index 2eba141bd521..51637e58a48b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h
@@ -452,6 +452,8 @@ void mlx5_unload_one_light(struct mlx5_core_dev *dev);
 
 void mlx5_query_nic_sw_system_image_guid(struct mlx5_core_dev *mdev, u8 *buf,
 					 u8 *len);
+bool mlx5_vport_use_vhca_id_as_func_id(struct mlx5_core_dev *dev,
+				       u16 vport_num, u16 *vhca_id);
 int mlx5_vport_set_other_func_cap(struct mlx5_core_dev *dev, const void *hca_cap, u16 vport,
 				  u16 opmod);
 #define mlx5_vport_get_other_func_general_cap(dev, vport, out)		\
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/vport.c b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
index f8e6b1ab7c5c..e0848f4e88dd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/vport.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/vport.c
@@ -1283,8 +1283,8 @@ void mlx5_query_nic_sw_system_image_guid(struct mlx5_core_dev *mdev, u8 *buf,
 		buf[(*len)++] = MLX5_CAP_GEN_2(mdev, load_balance_id);
 }
 
-static bool mlx5_vport_use_vhca_id_as_func_id(struct mlx5_core_dev *dev,
-					      u16 vport_num, u16 *vhca_id)
+bool mlx5_vport_use_vhca_id_as_func_id(struct mlx5_core_dev *dev,
+				       u16 vport_num, u16 *vhca_id)
 {
 	if (!MLX5_CAP_GEN_2(dev, function_id_type_vhca_id))
 		return false;
-- 
2.44.0

Re: [PATCH net-next V2 0/8] net/mlx5: Prepare eswitch infrastructure for satellite PF support

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Paolo Abeni <pabeni@redhat.com>:

On Mon, 18 May 2026 10:13:48 +0300 you wrote:
> Hi,
> 
> See detailed description by Moshe below.
> 
> Regards,
> Tariq
> 
> [...]

Here is the summary with links:
  - [net-next,V2,1/8] net/mlx5: Use helper to parse host PF info
    https://git.kernel.org/netdev/net-next/c/920fdc4d21e3
  - [net-next,V2,2/8] net/mlx5: Use v1 response layout for query_esw_functions
    https://git.kernel.org/netdev/net-next/c/a4f75c4238b0
  - [net-next,V2,3/8] net/mlx5: Use mlx5_eswitch_is_vf_vport() for IPsec VF checks
    https://git.kernel.org/netdev/net-next/c/3fefa7e7c14b
  - [net-next,V2,4/8] net/mlx5: Switch vport HCA cap helpers to kvzalloc
    https://git.kernel.org/netdev/net-next/c/62af408fd772
  - [net-next,V2,5/8] net/mlx5: Add mlx5_vport_set_other_func_general_cap macro
    https://git.kernel.org/netdev/net-next/c/9244a323125c
  - [net-next,V2,6/8] net/mlx5: Refactor mlx5_set_msix_vec_count() SET_HCA_CAP
    https://git.kernel.org/netdev/net-next/c/fa2852a28c5b
  - [net-next,V2,7/8] net/mlx5: Use vport helper for IPsec eswitch set caps
    https://git.kernel.org/netdev/net-next/c/d7ec361003fa
  - [net-next,V2,8/8] net/mlx5: Generalize enable/disable HCA for any PF vport
    https://git.kernel.org/netdev/net-next/c/4a3b5efee2e5

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html