Commit 7bf481d7 authored by Parav Pandit's avatar Parav Pandit Committed by Saeed Mahameed

net/mlx5: E-Switch, let user to enable disable metadata

Currently each packet inserted in eswitch is tagged with a internal
metadata to indicate source vport. Metadata tagging is not always
needed. Metadata insertion is needed for multi-port RoCE, failover
between representors and stacked devices. In many other cases,
metadata enablement is not needed.

Metadata insertion slows down the packet processing rate of the E-switch
when it is in switchdev mode.

Below table show performance gain with metadata disabled for VXLAN
offload rules in both SMFS and DMFS steering mode on ConnectX-5 device.

----------------------------------------------
| steering | metadata | pkt size | rx pps    |
| mode     |          |          | (million) |
----------------------------------------------
| smfs     | disabled | 128Bytes | 42        |
----------------------------------------------
| smfs     | enabled  | 128Bytes | 36        |
----------------------------------------------
| dmfs     | disabled | 128Bytes | 42        |
----------------------------------------------
| dmfs     | enabled  | 128Bytes | 36        |
----------------------------------------------

Hence, allow user to disable metadata using driver specific devlink
parameter. Metadata setting of the eswitch is applicable only for the
switchdev mode.

Example to show and disable metadata before changing eswitch mode:
$ devlink dev param show pci/0000:06:00.0 name esw_port_metadata
pci/0000:06:00.0:
  name esw_port_metadata type driver-specific
    values:
      cmode runtime value true

$ devlink dev param set pci/0000:06:00.0 \
	  name esw_port_metadata value false cmode runtime

$ devlink dev eswitch set pci/0000:06:00.0 mode switchdev
Signed-off-by: default avatarParav Pandit <parav@nvidia.com>
Reviewed-by: default avatarRoi Dayan <roid@nvidia.com>
Reviewed-by: default avatarVu Pham <vuhuong@nvidia.com>
Signed-off-by: default avatarSaeed Mahameed <saeedm@nvidia.com>
---
changelog:
v1->v2:
 - added performance numbers in commit log
 - updated commit log and documentation for switchdev mode
 - added explicit note on when user can disable metadata in
   documentation
parent 8ef7adc6
...@@ -183,6 +183,40 @@ User command examples: ...@@ -183,6 +183,40 @@ User command examples:
values: values:
cmode driverinit value true cmode driverinit value true
esw_port_metadata: Eswitch port metadata state
----------------------------------------------
When applicable, disabling Eswitch metadata can increase packet rate
up to 20% depending on the use case and packet sizes.
Eswitch port metadata state controls whether to internally tag packets with
metadata. Metadata tagging must be enabled for multi-port RoCE, failover
between representors and stacked devices.
By default metadata is enabled on the supported devices in E-switch.
Metadata is applicable only for E-switch in switchdev mode and
users may disable it when NONE of the below use cases will be in use:
1. HCA is in Dual/multi-port RoCE mode.
2. VF/SF representor bonding (Usually used for Live migration)
3. Stacked devices
When metadata is disabled, the above use cases will fail to initialize if
users try to enable them.
- Show eswitch port metadata::
$ devlink dev param show pci/0000:06:00.0 name esw_port_metadata
pci/0000:06:00.0:
name esw_port_metadata type driver-specific
values:
cmode runtime value true
- Disable eswitch port metadata::
$ devlink dev param set pci/0000:06:00.0 name esw_port_metadata value false cmode runtime
- Change eswitch mode to switchdev mode where after choosing the metadata value::
$ devlink dev eswitch set pci/0000:06:00.0 mode switchdev
mlx5 subfunction mlx5 subfunction
================ ================
mlx5 supports subfunction management using devlink port (see :ref:`Documentation/networking/devlink/devlink-port.rst <devlink_port>`) interface. mlx5 supports subfunction management using devlink port (see :ref:`Documentation/networking/devlink/devlink-port.rst <devlink_port>`) interface.
......
...@@ -456,6 +456,50 @@ static int mlx5_devlink_large_group_num_validate(struct devlink *devlink, u32 id ...@@ -456,6 +456,50 @@ static int mlx5_devlink_large_group_num_validate(struct devlink *devlink, u32 id
return 0; return 0;
} }
static int mlx5_devlink_esw_port_metadata_set(struct devlink *devlink, u32 id,
struct devlink_param_gset_ctx *ctx)
{
struct mlx5_core_dev *dev = devlink_priv(devlink);
if (!MLX5_ESWITCH_MANAGER(dev))
return -EOPNOTSUPP;
return mlx5_esw_offloads_vport_metadata_set(dev->priv.eswitch, ctx->val.vbool);
}
static int mlx5_devlink_esw_port_metadata_get(struct devlink *devlink, u32 id,
struct devlink_param_gset_ctx *ctx)
{
struct mlx5_core_dev *dev = devlink_priv(devlink);
if (!MLX5_ESWITCH_MANAGER(dev))
return -EOPNOTSUPP;
ctx->val.vbool = mlx5_eswitch_vport_match_metadata_enabled(dev->priv.eswitch);
return 0;
}
static int mlx5_devlink_esw_port_metadata_validate(struct devlink *devlink, u32 id,
union devlink_param_value val,
struct netlink_ext_ack *extack)
{
struct mlx5_core_dev *dev = devlink_priv(devlink);
u8 esw_mode;
if (!MLX5_ESWITCH_MANAGER(dev)) {
NL_SET_ERR_MSG_MOD(extack, "E-Switch is unsupported");
return -EOPNOTSUPP;
}
esw_mode = mlx5_eswitch_mode(dev);
if (esw_mode == MLX5_ESWITCH_OFFLOADS) {
NL_SET_ERR_MSG_MOD(extack,
"E-Switch must either disabled or non switchdev mode");
return -EBUSY;
}
return 0;
}
#endif #endif
static int mlx5_devlink_enable_remote_dev_reset_set(struct devlink *devlink, u32 id, static int mlx5_devlink_enable_remote_dev_reset_set(struct devlink *devlink, u32 id,
...@@ -490,6 +534,12 @@ static const struct devlink_param mlx5_devlink_params[] = { ...@@ -490,6 +534,12 @@ static const struct devlink_param mlx5_devlink_params[] = {
BIT(DEVLINK_PARAM_CMODE_DRIVERINIT), BIT(DEVLINK_PARAM_CMODE_DRIVERINIT),
NULL, NULL, NULL, NULL,
mlx5_devlink_large_group_num_validate), mlx5_devlink_large_group_num_validate),
DEVLINK_PARAM_DRIVER(MLX5_DEVLINK_PARAM_ID_ESW_PORT_METADATA,
"esw_port_metadata", DEVLINK_PARAM_TYPE_BOOL,
BIT(DEVLINK_PARAM_CMODE_RUNTIME),
mlx5_devlink_esw_port_metadata_get,
mlx5_devlink_esw_port_metadata_set,
mlx5_devlink_esw_port_metadata_validate),
#endif #endif
DEVLINK_PARAM_GENERIC(ENABLE_REMOTE_DEV_RESET, BIT(DEVLINK_PARAM_CMODE_RUNTIME), DEVLINK_PARAM_GENERIC(ENABLE_REMOTE_DEV_RESET, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
mlx5_devlink_enable_remote_dev_reset_get, mlx5_devlink_enable_remote_dev_reset_get,
...@@ -519,6 +569,18 @@ static void mlx5_devlink_set_params_init_values(struct devlink *devlink) ...@@ -519,6 +569,18 @@ static void mlx5_devlink_set_params_init_values(struct devlink *devlink)
devlink_param_driverinit_value_set(devlink, devlink_param_driverinit_value_set(devlink,
MLX5_DEVLINK_PARAM_ID_ESW_LARGE_GROUP_NUM, MLX5_DEVLINK_PARAM_ID_ESW_LARGE_GROUP_NUM,
value); value);
if (MLX5_ESWITCH_MANAGER(dev)) {
if (mlx5_esw_vport_match_metadata_supported(dev->priv.eswitch)) {
dev->priv.eswitch->flags |= MLX5_ESWITCH_VPORT_MATCH_METADATA;
value.vbool = true;
} else {
value.vbool = false;
}
devlink_param_driverinit_value_set(devlink,
MLX5_DEVLINK_PARAM_ID_ESW_PORT_METADATA,
value);
}
#endif #endif
} }
......
...@@ -10,6 +10,7 @@ enum mlx5_devlink_param_id { ...@@ -10,6 +10,7 @@ enum mlx5_devlink_param_id {
MLX5_DEVLINK_PARAM_ID_BASE = DEVLINK_PARAM_GENERIC_ID_MAX, MLX5_DEVLINK_PARAM_ID_BASE = DEVLINK_PARAM_GENERIC_ID_MAX,
MLX5_DEVLINK_PARAM_ID_FLOW_STEERING_MODE, MLX5_DEVLINK_PARAM_ID_FLOW_STEERING_MODE,
MLX5_DEVLINK_PARAM_ID_ESW_LARGE_GROUP_NUM, MLX5_DEVLINK_PARAM_ID_ESW_LARGE_GROUP_NUM,
MLX5_DEVLINK_PARAM_ID_ESW_PORT_METADATA,
}; };
struct mlx5_trap_ctx { struct mlx5_trap_ctx {
......
...@@ -314,6 +314,8 @@ int esw_offloads_enable(struct mlx5_eswitch *esw); ...@@ -314,6 +314,8 @@ int esw_offloads_enable(struct mlx5_eswitch *esw);
void esw_offloads_cleanup_reps(struct mlx5_eswitch *esw); void esw_offloads_cleanup_reps(struct mlx5_eswitch *esw);
int esw_offloads_init_reps(struct mlx5_eswitch *esw); int esw_offloads_init_reps(struct mlx5_eswitch *esw);
bool mlx5_esw_vport_match_metadata_supported(const struct mlx5_eswitch *esw);
int mlx5_esw_offloads_vport_metadata_set(struct mlx5_eswitch *esw, bool enable);
u32 mlx5_esw_match_metadata_alloc(struct mlx5_eswitch *esw); u32 mlx5_esw_match_metadata_alloc(struct mlx5_eswitch *esw);
void mlx5_esw_match_metadata_free(struct mlx5_eswitch *esw, u32 metadata); void mlx5_esw_match_metadata_free(struct mlx5_eswitch *esw, u32 metadata);
......
...@@ -2351,8 +2351,7 @@ static void esw_offloads_devcom_cleanup(struct mlx5_eswitch *esw) ...@@ -2351,8 +2351,7 @@ static void esw_offloads_devcom_cleanup(struct mlx5_eswitch *esw)
mlx5_devcom_unregister_component(devcom, MLX5_DEVCOM_ESW_OFFLOADS); mlx5_devcom_unregister_component(devcom, MLX5_DEVCOM_ESW_OFFLOADS);
} }
static bool bool mlx5_esw_vport_match_metadata_supported(const struct mlx5_eswitch *esw)
esw_check_vport_match_metadata_supported(const struct mlx5_eswitch *esw)
{ {
if (!MLX5_CAP_ESW(esw->dev, esw_uplink_ingress_acl)) if (!MLX5_CAP_ESW(esw->dev, esw_uplink_ingress_acl))
return false; return false;
...@@ -2452,6 +2451,28 @@ static int esw_offloads_metadata_init(struct mlx5_eswitch *esw) ...@@ -2452,6 +2451,28 @@ static int esw_offloads_metadata_init(struct mlx5_eswitch *esw)
return err; return err;
} }
int mlx5_esw_offloads_vport_metadata_set(struct mlx5_eswitch *esw, bool enable)
{
int err = 0;
down_write(&esw->mode_lock);
if (esw->mode != MLX5_ESWITCH_NONE) {
err = -EBUSY;
goto done;
}
if (!mlx5_esw_vport_match_metadata_supported(esw)) {
err = -EOPNOTSUPP;
goto done;
}
if (enable)
esw->flags |= MLX5_ESWITCH_VPORT_MATCH_METADATA;
else
esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA;
done:
up_write(&esw->mode_lock);
return err;
}
int int
esw_vport_create_offloads_acl_tables(struct mlx5_eswitch *esw, esw_vport_create_offloads_acl_tables(struct mlx5_eswitch *esw,
struct mlx5_vport *vport) struct mlx5_vport *vport)
...@@ -2673,9 +2694,6 @@ int esw_offloads_enable(struct mlx5_eswitch *esw) ...@@ -2673,9 +2694,6 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)
if (err) if (err)
goto err_metadata; goto err_metadata;
if (esw_check_vport_match_metadata_supported(esw))
esw->flags |= MLX5_ESWITCH_VPORT_MATCH_METADATA;
err = esw_offloads_metadata_init(esw); err = esw_offloads_metadata_init(esw);
if (err) if (err)
goto err_metadata; goto err_metadata;
...@@ -2725,7 +2743,6 @@ int esw_offloads_enable(struct mlx5_eswitch *esw) ...@@ -2725,7 +2743,6 @@ int esw_offloads_enable(struct mlx5_eswitch *esw)
err_vport_metadata: err_vport_metadata:
esw_offloads_metadata_uninit(esw); esw_offloads_metadata_uninit(esw);
err_metadata: err_metadata:
esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA;
mlx5_rdma_disable_roce(esw->dev); mlx5_rdma_disable_roce(esw->dev);
mutex_destroy(&esw->offloads.termtbl_mutex); mutex_destroy(&esw->offloads.termtbl_mutex);
return err; return err;
...@@ -2761,7 +2778,6 @@ void esw_offloads_disable(struct mlx5_eswitch *esw) ...@@ -2761,7 +2778,6 @@ void esw_offloads_disable(struct mlx5_eswitch *esw)
esw_offloads_steering_cleanup(esw); esw_offloads_steering_cleanup(esw);
mapping_destroy(esw->offloads.reg_c0_obj_pool); mapping_destroy(esw->offloads.reg_c0_obj_pool);
esw_offloads_metadata_uninit(esw); esw_offloads_metadata_uninit(esw);
esw->flags &= ~MLX5_ESWITCH_VPORT_MATCH_METADATA;
mlx5_rdma_disable_roce(esw->dev); mlx5_rdma_disable_roce(esw->dev);
mutex_destroy(&esw->offloads.termtbl_mutex); mutex_destroy(&esw->offloads.termtbl_mutex);
esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE; esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment