[PATCH mlx5-next 0/5] Mellanox, mlx5 roce enable devlink parameter

[PATCH mlx5-next 0/5] Mellanox, mlx5 roce enable devlink parameter

[Saeed Mahameed]

Hi All,

Currently RoCE is always enabled by default in mlx5 and there is no option
for the user to disable it.

This change introduces new generic devlink param "enable_roce". 
User should set desired param value and reload the driver to get
desired configuration.

Two options for reload:
1) reload mlx5_ib driver.
2) reload via devlink, once the devlink mlx5 reload support [1]
lands in net-next branch, after this series is applied.

mlx5 devlink reload is not part of this patchset since it depends on
changes from net-next tree.

RoCE will still be enabled by default after this change.

In case of no objection this series will be applied to mlx5-next branch
and sent later as pull request to both rdma-next and net-next branches.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git/commit/?h=net-next-mlx5&id=0915d89b86e379b26457ba50656e9257193eaa60

Thanks,
Saeed.

---

Michael Guralnik (5):
  devlink: Add new "enable_roce" generic device param
  net/mlx5: Document flow_steering_mode devlink param
  net/mlx5: Handle "enable_roce" devlink param
  IB/mlx5: Rename profile and init methods
  IB/mlx5: Load profile according to RoCE enablement state

 .../device_drivers/mellanox/mlx5.rst          | 21 ++++++++++
 .../networking/devlink-params-mlx5.txt        | 17 ++++++++
 Documentation/networking/devlink-params.txt   |  4 ++
 drivers/infiniband/hw/mlx5/ib_rep.c           |  2 +-
 drivers/infiniband/hw/mlx5/ib_rep.h           |  2 +-
 drivers/infiniband/hw/mlx5/main.c             | 39 ++++++++++---------
 .../net/ethernet/mellanox/mlx5/core/devlink.c | 22 +++++++++++
 include/linux/mlx5/driver.h                   | 11 ++++++
 include/net/devlink.h                         |  4 ++
 net/core/devlink.c                            |  5 +++
 10 files changed, 106 insertions(+), 21 deletions(-)
 create mode 100644 Documentation/networking/devlink-params-mlx5.txt

-- 
2.21.0

[PATCH mlx5-next 1/5] devlink: Add new "enable_roce" generic device param

[Saeed Mahameed]

From: Michael Guralnik <michaelgur@mellanox.com>

New device parameter to enable/disable handling of RoCE traffic in the
device.

Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 Documentation/networking/devlink-params.txt | 4 ++++
 include/net/devlink.h                       | 4 ++++
 net/core/devlink.c                          | 5 +++++
 3 files changed, 13 insertions(+)

diff --git a/Documentation/networking/devlink-params.txt b/Documentation/networking/devlink-params.txt
index ddba3e9b55b1..04e234e9acc9 100644
--- a/Documentation/networking/devlink-params.txt
+++ b/Documentation/networking/devlink-params.txt
@@ -65,3 +65,7 @@ reset_dev_on_drv_probe	[DEVICE, GENERIC]
 			  Reset only if device firmware can be found in the
 			  filesystem.
 			Type: u8
+
+enable_roce		[DEVICE, GENERIC]
+			Enable handling of RoCE traffic in the device.
+			Type: Boolean
diff --git a/include/net/devlink.h b/include/net/devlink.h
index 23e4b65ec9df..39fb4d957838 100644
--- a/include/net/devlink.h
+++ b/include/net/devlink.h
@@ -400,6 +400,7 @@ enum devlink_param_generic_id {
 	DEVLINK_PARAM_GENERIC_ID_MSIX_VEC_PER_PF_MIN,
 	DEVLINK_PARAM_GENERIC_ID_FW_LOAD_POLICY,
 	DEVLINK_PARAM_GENERIC_ID_RESET_DEV_ON_DRV_PROBE,
+	DEVLINK_PARAM_GENERIC_ID_ENABLE_ROCE,
 
 	/* add new param generic ids above here*/
 	__DEVLINK_PARAM_GENERIC_ID_MAX,
@@ -434,6 +435,9 @@ enum devlink_param_generic_id {
 	"reset_dev_on_drv_probe"
 #define DEVLINK_PARAM_GENERIC_RESET_DEV_ON_DRV_PROBE_TYPE DEVLINK_PARAM_TYPE_U8
 
+#define DEVLINK_PARAM_GENERIC_ENABLE_ROCE_NAME "enable_roce"
+#define DEVLINK_PARAM_GENERIC_ENABLE_ROCE_TYPE DEVLINK_PARAM_TYPE_BOOL
+
 #define DEVLINK_PARAM_GENERIC(_id, _cmodes, _get, _set, _validate)	\
 {									\
 	.id = DEVLINK_PARAM_GENERIC_ID_##_id,				\
diff --git a/net/core/devlink.c b/net/core/devlink.c
index f80151eeaf51..0fbcd44aa64f 100644
--- a/net/core/devlink.c
+++ b/net/core/devlink.c
@@ -2884,6 +2884,11 @@ static const struct devlink_param devlink_param_generic[] = {
 		.name = DEVLINK_PARAM_GENERIC_RESET_DEV_ON_DRV_PROBE_NAME,
 		.type = DEVLINK_PARAM_GENERIC_RESET_DEV_ON_DRV_PROBE_TYPE,
 	},
+	{
+		.id = DEVLINK_PARAM_GENERIC_ID_ENABLE_ROCE,
+		.name = DEVLINK_PARAM_GENERIC_ENABLE_ROCE_NAME,
+		.type = DEVLINK_PARAM_GENERIC_ENABLE_ROCE_TYPE,
+	},
 };
 
 static int devlink_param_generic_verify(const struct devlink_param *param)
-- 
2.21.0

[PATCH mlx5-next 2/5] net/mlx5: Document flow_steering_mode devlink param

[Saeed Mahameed]

From: Michael Guralnik <michaelgur@mellanox.com>

Add documentation for current mlx5 supported devlink param.

Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 Documentation/networking/devlink-params-mlx5.txt | 12 ++++++++++++
 1 file changed, 12 insertions(+)
 create mode 100644 Documentation/networking/devlink-params-mlx5.txt

diff --git a/Documentation/networking/devlink-params-mlx5.txt b/Documentation/networking/devlink-params-mlx5.txt
new file mode 100644
index 000000000000..8c0b82d655dc
--- /dev/null
+++ b/Documentation/networking/devlink-params-mlx5.txt
@@ -0,0 +1,12 @@
+flow_steering_mode	[DEVICE, DRIVER-SPECIFIC]
+			Controls the flow steering mode of the driver.
+			Two modes are supported:
+			1. 'dmfs' - Device managed flow steering.
+			2. 'smfs  - Software/Driver managed flow steering.
+			In DMFS mode, the HW steering entities are created and
+			managed through the Firmware.
+			In SMFS mode, the HW steering entities are created and
+			managed though by the driver directly into Hardware
+			without firmware intervention.
+			Type: String
+			Configuration mode: runtime
-- 
2.21.0

[PATCH mlx5-next 3/5] net/mlx5: Handle "enable_roce" devlink param

[Saeed Mahameed]

From: Michael Guralnik <michaelgur@mellanox.com>

Register "enable_roce" param, default value is RoCE enabled.
Current configuration is stored on mlx5_core_dev and exposed to user
through the cmode runtime devlink param.
Changing configuration requires changing the cmode driverinit devlink
param and calling devlink reload.

Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 .../device_drivers/mellanox/mlx5.rst          | 21 ++++++++++++++++++
 .../networking/devlink-params-mlx5.txt        |  5 +++++
 .../net/ethernet/mellanox/mlx5/core/devlink.c | 22 +++++++++++++++++++
 include/linux/mlx5/driver.h                   | 11 ++++++++++
 4 files changed, 59 insertions(+)

diff --git a/Documentation/networking/device_drivers/mellanox/mlx5.rst b/Documentation/networking/device_drivers/mellanox/mlx5.rst
index d071c6b49e1f..7599dceba9f1 100644
--- a/Documentation/networking/device_drivers/mellanox/mlx5.rst
+++ b/Documentation/networking/device_drivers/mellanox/mlx5.rst
@@ -154,6 +154,27 @@ User command examples:
       values:
          cmode runtime value smfs
 
+enable_roce: RoCE enablement state
+----------------------------------
+RoCE enablement state controls driver support for RoCE traffic.
+When RoCE is disabled, there is no gid table, only raw ethernet QPs are supported and traffic on the well known UDP RoCE port is handled as raw ethernet traffic.
+
+To change RoCE enablement state a user must change the driverinit cmode value and run devlink reload.
+
+User command examples:
+
+- Disable RoCE::
+
+    $ devlink dev param set pci/0000:06:00.0 name enable_roce value false cmode driverinit
+    $ devlink dev reload pci/0000:06:00.0
+
+- Read RoCE enablement state::
+
+    $ devlink dev param show pci/0000:06:00.0 name enable_roce
+      pci/0000:06:00.0:
+      name enable_roce type generic
+      values:
+         cmode driverinit value true
 
 Devlink health reporters
 ========================
diff --git a/Documentation/networking/devlink-params-mlx5.txt b/Documentation/networking/devlink-params-mlx5.txt
index 8c0b82d655dc..5071467118bd 100644
--- a/Documentation/networking/devlink-params-mlx5.txt
+++ b/Documentation/networking/devlink-params-mlx5.txt
@@ -10,3 +10,8 @@ flow_steering_mode	[DEVICE, DRIVER-SPECIFIC]
 			without firmware intervention.
 			Type: String
 			Configuration mode: runtime
+
+enable_roce		[DEVICE, GENERIC]
+			Enable handling of RoCE traffic in the device.
+			Defaultly enabled.
+			Configuration mode: driverinit
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
index 381925c90d94..b2c26388edb1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
@@ -177,12 +177,29 @@ enum mlx5_devlink_param_id {
 	MLX5_DEVLINK_PARAM_FLOW_STEERING_MODE,
 };
 
+static int mlx5_devlink_enable_roce_validate(struct devlink *devlink, u32 id,
+					     union devlink_param_value val,
+					     struct netlink_ext_ack *extack)
+{
+	struct mlx5_core_dev *dev = devlink_priv(devlink);
+	bool new_state = val.vbool;
+
+	if (new_state && !MLX5_CAP_GEN(dev, roce)) {
+		NL_SET_ERR_MSG_MOD(extack, "Device doesn't support RoCE");
+		return -EOPNOTSUPP;
+	}
+
+	return 0;
+}
+
 static const struct devlink_param mlx5_devlink_params[] = {
 	DEVLINK_PARAM_DRIVER(MLX5_DEVLINK_PARAM_FLOW_STEERING_MODE,
 			     "flow_steering_mode", DEVLINK_PARAM_TYPE_STRING,
 			     BIT(DEVLINK_PARAM_CMODE_RUNTIME),
 			     mlx5_devlink_fs_mode_get, mlx5_devlink_fs_mode_set,
 			     mlx5_devlink_fs_mode_validate),
+	DEVLINK_PARAM_GENERIC(ENABLE_ROCE, BIT(DEVLINK_PARAM_CMODE_DRIVERINIT),
+			      NULL, NULL, mlx5_devlink_enable_roce_validate),
 };
 
 static void mlx5_devlink_set_params_init_values(struct devlink *devlink)
@@ -197,6 +214,11 @@ static void mlx5_devlink_set_params_init_values(struct devlink *devlink)
 	devlink_param_driverinit_value_set(devlink,
 					   MLX5_DEVLINK_PARAM_FLOW_STEERING_MODE,
 					   value);
+
+	value.vbool = MLX5_CAP_GEN(dev, roce);
+	devlink_param_driverinit_value_set(devlink,
+					   DEVLINK_PARAM_GENERIC_ID_ENABLE_ROCE,
+					   value);
 }
 
 int mlx5_devlink_register(struct devlink *devlink, struct device *dev)
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index 7b4801e96feb..1884513aac90 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -1191,4 +1191,15 @@ enum {
 	MLX5_TRIGGERED_CMD_COMP = (u64)1 << 32,
 };
 
+static inline bool mlx5_is_roce_enabled(struct mlx5_core_dev *dev)
+{
+	struct devlink *devlink = priv_to_devlink(dev);
+	union devlink_param_value val;
+
+	devlink_param_driverinit_value_get(devlink,
+					   DEVLINK_PARAM_GENERIC_ID_ENABLE_ROCE,
+					   &val);
+	return val.vbool;
+}
+
 #endif /* MLX5_DRIVER_H */
-- 
2.21.0

[PATCH mlx5-next 4/5] IB/mlx5: Rename profile and init methods

[Saeed Mahameed]

From: Michael Guralnik <michaelgur@mellanox.com>

Rename uplink_rep_profile and its unique init and cleanup stages to
suit its upcoming use as the profile when RoCE is disabled.

Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Reviewed-by: Maor Gottlieb <maorg@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/infiniband/hw/mlx5/ib_rep.c |  2 +-
 drivers/infiniband/hw/mlx5/ib_rep.h |  2 +-
 drivers/infiniband/hw/mlx5/main.c   | 14 +++++++-------
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/ib_rep.c b/drivers/infiniband/hw/mlx5/ib_rep.c
index 74ce9249e75a..5c3d052ac30b 100644
--- a/drivers/infiniband/hw/mlx5/ib_rep.c
+++ b/drivers/infiniband/hw/mlx5/ib_rep.c
@@ -35,7 +35,7 @@ mlx5_ib_vport_rep_load(struct mlx5_core_dev *dev, struct mlx5_eswitch_rep *rep)
 	int vport_index;
 
 	if (rep->vport == MLX5_VPORT_UPLINK)
-		profile = &uplink_rep_profile;
+		profile = &raw_eth_profile;
 	else
 		return mlx5_ib_set_vport_rep(dev, rep);
 
diff --git a/drivers/infiniband/hw/mlx5/ib_rep.h b/drivers/infiniband/hw/mlx5/ib_rep.h
index de43b423bafc..3b6750cba796 100644
--- a/drivers/infiniband/hw/mlx5/ib_rep.h
+++ b/drivers/infiniband/hw/mlx5/ib_rep.h
@@ -10,7 +10,7 @@
 #include "mlx5_ib.h"
 
 #ifdef CONFIG_MLX5_ESWITCH
-extern const struct mlx5_ib_profile uplink_rep_profile;
+extern const struct mlx5_ib_profile raw_eth_profile;
 
 u8 mlx5_ib_eswitch_mode(struct mlx5_eswitch *esw);
 struct mlx5_ib_dev *mlx5_ib_get_rep_ibdev(struct mlx5_eswitch *esw,
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 8343a740c91e..d6afe33d56ac 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -6444,7 +6444,7 @@ static const struct ib_device_ops mlx5_ib_dev_port_rep_ops = {
 	.query_port = mlx5_ib_rep_query_port,
 };
 
-static int mlx5_ib_stage_rep_non_default_cb(struct mlx5_ib_dev *dev)
+static int mlx5_ib_stage_raw_eth_non_default_cb(struct mlx5_ib_dev *dev)
 {
 	ib_set_device_ops(&dev->ib_dev, &mlx5_ib_dev_port_rep_ops);
 	return 0;
@@ -6484,7 +6484,7 @@ static void mlx5_ib_stage_common_roce_cleanup(struct mlx5_ib_dev *dev)
 	mlx5_remove_netdev_notifier(dev, port_num);
 }
 
-static int mlx5_ib_stage_rep_roce_init(struct mlx5_ib_dev *dev)
+static int mlx5_ib_stage_raw_eth_roce_init(struct mlx5_ib_dev *dev)
 {
 	struct mlx5_core_dev *mdev = dev->mdev;
 	enum rdma_link_layer ll;
@@ -6500,7 +6500,7 @@ static int mlx5_ib_stage_rep_roce_init(struct mlx5_ib_dev *dev)
 	return err;
 }
 
-static void mlx5_ib_stage_rep_roce_cleanup(struct mlx5_ib_dev *dev)
+static void mlx5_ib_stage_raw_eth_roce_cleanup(struct mlx5_ib_dev *dev)
 {
 	mlx5_ib_stage_common_roce_cleanup(dev);
 }
@@ -6807,7 +6807,7 @@ static const struct mlx5_ib_profile pf_profile = {
 		     mlx5_ib_stage_delay_drop_cleanup),
 };
 
-const struct mlx5_ib_profile uplink_rep_profile = {
+const struct mlx5_ib_profile raw_eth_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_INIT,
 		     mlx5_ib_stage_init_init,
 		     mlx5_ib_stage_init_cleanup),
@@ -6818,11 +6818,11 @@ const struct mlx5_ib_profile uplink_rep_profile = {
 		     mlx5_ib_stage_caps_init,
 		     NULL),
 	STAGE_CREATE(MLX5_IB_STAGE_NON_DEFAULT_CB,
-		     mlx5_ib_stage_rep_non_default_cb,
+		     mlx5_ib_stage_raw_eth_non_default_cb,
 		     NULL),
 	STAGE_CREATE(MLX5_IB_STAGE_ROCE,
-		     mlx5_ib_stage_rep_roce_init,
-		     mlx5_ib_stage_rep_roce_cleanup),
+		     mlx5_ib_stage_raw_eth_roce_init,
+		     mlx5_ib_stage_raw_eth_roce_cleanup),
 	STAGE_CREATE(MLX5_IB_STAGE_SRQ,
 		     mlx5_init_srq_table,
 		     mlx5_cleanup_srq_table),
-- 
2.21.0

[PATCH mlx5-next 5/5] IB/mlx5: Load profile according to RoCE enablement state

[Saeed Mahameed]

From: Michael Guralnik <michaelgur@mellanox.com>

When RoCE is disabled load mlx5_ib in raw_eth profile.
Clean pf_profile roce capability checks as it will not be used without
roce capability.

Signed-off-by: Michael Guralnik <michaelgur@mellanox.com>
Reviewed-by: Maor Gottlieb <maorg@mellanox.com>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c | 25 +++++++++++++------------
 1 file changed, 13 insertions(+), 12 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index d6afe33d56ac..46ea4f0b9b51 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -5145,8 +5145,7 @@ static int mlx5_port_immutable(struct ib_device *ibdev, u8 port_num,
 	immutable->pkey_tbl_len = attr.pkey_tbl_len;
 	immutable->gid_tbl_len = attr.gid_tbl_len;
 	immutable->core_cap_flags = get_core_cap_flags(ibdev, &rep);
-	if ((ll == IB_LINK_LAYER_INFINIBAND) || MLX5_CAP_GEN(dev->mdev, roce))
-		immutable->max_mad_size = IB_MGMT_MAD_SIZE;
+	immutable->max_mad_size = IB_MGMT_MAD_SIZE;
 
 	return 0;
 }
@@ -5249,11 +5248,9 @@ static int mlx5_enable_eth(struct mlx5_ib_dev *dev)
 {
 	int err;
 
-	if (MLX5_CAP_GEN(dev->mdev, roce)) {
-		err = mlx5_nic_vport_enable_roce(dev->mdev);
-		if (err)
-			return err;
-	}
+	err = mlx5_nic_vport_enable_roce(dev->mdev);
+	if (err)
+		return err;
 
 	err = mlx5_eth_lag_init(dev);
 	if (err)
@@ -5262,8 +5259,7 @@ static int mlx5_enable_eth(struct mlx5_ib_dev *dev)
 	return 0;
 
 err_disable_roce:
-	if (MLX5_CAP_GEN(dev->mdev, roce))
-		mlx5_nic_vport_disable_roce(dev->mdev);
+	mlx5_nic_vport_disable_roce(dev->mdev);
 
 	return err;
 }
@@ -5271,8 +5267,7 @@ static int mlx5_enable_eth(struct mlx5_ib_dev *dev)
 static void mlx5_disable_eth(struct mlx5_ib_dev *dev)
 {
 	mlx5_eth_lag_cleanup(dev);
-	if (MLX5_CAP_GEN(dev->mdev, roce))
-		mlx5_nic_vport_disable_roce(dev->mdev);
+	mlx5_nic_vport_disable_roce(dev->mdev);
 }
 
 struct mlx5_ib_counter {
@@ -6898,6 +6893,7 @@ static void *mlx5_ib_add_slave_port(struct mlx5_core_dev *mdev)
 
 static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
 {
+	const struct mlx5_ib_profile *profile;
 	enum rdma_link_layer ll;
 	struct mlx5_ib_dev *dev;
 	int port_type_cap;
@@ -6933,7 +6929,12 @@ static void *mlx5_ib_add(struct mlx5_core_dev *mdev)
 	dev->mdev = mdev;
 	dev->num_ports = num_ports;
 
-	return __mlx5_ib_add(dev, &pf_profile);
+	if (ll == IB_LINK_LAYER_ETHERNET && !mlx5_is_roce_enabled(mdev))
+		profile = &raw_eth_profile;
+	else
+		profile = &pf_profile;
+
+	return __mlx5_ib_add(dev, profile);
 }
 
 static void mlx5_ib_remove(struct mlx5_core_dev *mdev, void *context)
-- 
2.21.0

Re: [PATCH mlx5-next 0/5] Mellanox, mlx5 roce enable devlink parameter

[Saeed Mahameed]

On Fri, 2019-11-08 at 23:45 +0000, Saeed Mahameed wrote:
> Hi All,
> 
> Currently RoCE is always enabled by default in mlx5 and there is no
> option
> for the user to disable it.
> 
> This change introduces new generic devlink param "enable_roce". 
> User should set desired param value and reload the driver to get
> desired configuration.
> 
> Two options for reload:
> 1) reload mlx5_ib driver.
> 2) reload via devlink, once the devlink mlx5 reload support [1]
> lands in net-next branch, after this series is applied.
> 
> mlx5 devlink reload is not part of this patchset since it depends on
> changes from net-next tree.
> 
> RoCE will still be enabled by default after this change.
> 
> In case of no objection this series will be applied to mlx5-next
> branch
> and sent later as pull request to both rdma-next and net-next
> branches.
> 

Series applied to mlx5-next.

Thanks.