Merge branch 'net-mlx5e-add-pcie-congestion-event-extras'

Tariq Toukan says:

====================
net/mlx5e: Add pcie congestion event extras

This small series by Dragos covers gaps requested in the initial pcie
congestion series [1]:
- Make pcie congestion thresholds configurable via devlink.
- Add a counter for stale pcie congestion events.

[1] https://lore.kernel.org/1752130292-22249-1-git-send-email-tariqt@nvidia.com
====================

Link: https://patch.msgid.link/1757237976-531416-1-git-send-email-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

5 files changed, 238 insertions, 10 deletions

diff --git a/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/counters.rst b/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/counters.rst
index 754c81436408..cc498895f92e 100644
--- a/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/counters.rst
+++ b/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/counters.rst
@@ -1348,7 +1348,7 @@ Device Counters
        is in a congested state.
        If pci_bw_inbound_high == pci_bw_inbound_low then the device is not congested.
        If pci_bw_inbound_high > pci_bw_inbound_low then the device is congested.
-     - Tnformative
+     - Informative
 
    * - `pci_bw_inbound_low`
      - The number of times the device crossed the low inbound PCIe bandwidth
@@ -1373,3 +1373,8 @@ Device Counters
        If pci_bw_outbound_high == pci_bw_outbound_low then the device is not congested.
        If pci_bw_outbound_high > pci_bw_outbound_low then the device is congested.
      - Informative
+
+   * - `pci_bw_stale_event`
+     - The number of times the device fired a PCIe congestion event but on query
+       there was no change in state.
+     - Informative
diff --git a/Documentation/networking/devlink/mlx5.rst b/Documentation/networking/devlink/mlx5.rst
index 07b1424cbfbb..60cc9fedf1ef 100644
--- a/Documentation/networking/devlink/mlx5.rst
+++ b/Documentation/networking/devlink/mlx5.rst
@@ -146,6 +146,58 @@ parameters.
      - u32
      - driverinit
      - Control the size (in packets) of the hairpin queues.
+   * - ``pcie_cong_inbound_high``
+     - u16
+     - driverinit
+     - High threshold configuration for PCIe congestion events. The firmware
+       will send an event once device side inbound PCIe traffic went
+       above the configured high threshold for a long enough period (at least
+       200ms).
+
+       See pci_bw_inbound_high ethtool stat.
+
+       Units are 0.01 %. Accepted values are in range [0, 10000].
+       pcie_cong_inbound_low < pcie_cong_inbound_high.
+       Default value: 9000 (Corresponds to 90%).
+   * - ``pcie_cong_inbound_low``
+     - u16
+     - driverinit
+     - Low threshold configuration for PCIe congestion events. The firmware
+       will send an event once device side inbound PCIe traffic went
+       below the configured low threshold, only after having been previously in
+       a congested state.
+
+       See pci_bw_inbound_low ethtool stat.
+
+       Units are 0.01 %. Accepted values are in range [0, 10000].
+       pcie_cong_inbound_low < pcie_cong_inbound_high.
+       Default value: 7500.
+   * - ``pcie_cong_outbound_high``
+     - u16
+     - driverinit
+     - High threshold configuration for PCIe congestion events. The firmware
+       will send an event once device side outbound PCIe traffic went
+       above the configured high threshold for a long enough period (at least
+       200ms).
+
+       See pci_bw_outbound_high ethtool stat.
+
+       Units are 0.01 %. Accepted values are in range [0, 10000].
+       pcie_cong_outbound_low < pcie_cong_outbound_high.
+       Default value: 9000 (Corresponds to 90%).
+   * - ``pcie_cong_outbound_low``
+     - u16
+     - driverinit
+     - Low threshold configuration for PCIe congestion events. The firmware
+       will send an event once device side outbound PCIe traffic went
+       below the configured low threshold, only after having been previously in
+       a congested state.
+
+       See pci_bw_outbound_low ethtool stat.
+
+       Units are 0.01 %. Accepted values are in range [0, 10000].
+       pcie_cong_outbound_low < pcie_cong_outbound_high.
+       Default value: 7500.
 
    * - ``cqe_compress_type``
      - string
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
index 326d438e75b5..a0b68321355a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
@@ -651,6 +651,105 @@ static void mlx5_devlink_eth_params_unregister(struct devlink *devlink)
 			       ARRAY_SIZE(mlx5_devlink_eth_params));
 }
 
+#define MLX5_PCIE_CONG_THRESH_MAX	10000
+#define MLX5_PCIE_CONG_THRESH_DEF_LOW	7500
+#define MLX5_PCIE_CONG_THRESH_DEF_HIGH	9000
+
+static int
+mlx5_devlink_pcie_cong_thresh_validate(struct devlink *devl, u32 id,
+				       union devlink_param_value val,
+				       struct netlink_ext_ack *extack)
+{
+	if (val.vu16 > MLX5_PCIE_CONG_THRESH_MAX) {
+		NL_SET_ERR_MSG_FMT_MOD(extack, "Value %u > max supported (%u)",
+				       val.vu16, MLX5_PCIE_CONG_THRESH_MAX);
+
+		return -EINVAL;
+	}
+
+	switch (id) {
+	case MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_LOW:
+	case MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_HIGH:
+	case MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_LOW:
+	case MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_HIGH:
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	return 0;
+}
+
+static void mlx5_devlink_pcie_cong_init_values(struct devlink *devlink)
+{
+	union devlink_param_value value;
+	u32 id;
+
+	value.vu16 = MLX5_PCIE_CONG_THRESH_DEF_LOW;
+	id = MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_LOW;
+	devl_param_driverinit_value_set(devlink, id, value);
+
+	value.vu16 = MLX5_PCIE_CONG_THRESH_DEF_HIGH;
+	id = MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_HIGH;
+	devl_param_driverinit_value_set(devlink, id, value);
+
+	value.vu16 = MLX5_PCIE_CONG_THRESH_DEF_LOW;
+	id = MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_LOW;
+	devl_param_driverinit_value_set(devlink, id, value);
+
+	value.vu16 = MLX5_PCIE_CONG_THRESH_DEF_HIGH;
+	id = MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_HIGH;
+	devl_param_driverinit_value_set(devlink, id, value);
+}
+
+static const struct devlink_param mlx5_devlink_pcie_cong_params[] = {
+	DEVLINK_PARAM_DRIVER(MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_LOW,
+			     "pcie_cong_inbound_low", DEVLINK_PARAM_TYPE_U16,
+			     BIT(DEVLINK_PARAM_CMODE_DRIVERINIT), NULL, NULL,
+			     mlx5_devlink_pcie_cong_thresh_validate),
+	DEVLINK_PARAM_DRIVER(MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_HIGH,
+			     "pcie_cong_inbound_high", DEVLINK_PARAM_TYPE_U16,
+			     BIT(DEVLINK_PARAM_CMODE_DRIVERINIT), NULL, NULL,
+			     mlx5_devlink_pcie_cong_thresh_validate),
+	DEVLINK_PARAM_DRIVER(MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_LOW,
+			     "pcie_cong_outbound_low", DEVLINK_PARAM_TYPE_U16,
+			     BIT(DEVLINK_PARAM_CMODE_DRIVERINIT), NULL, NULL,
+			     mlx5_devlink_pcie_cong_thresh_validate),
+	DEVLINK_PARAM_DRIVER(MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_HIGH,
+			     "pcie_cong_outbound_high", DEVLINK_PARAM_TYPE_U16,
+			     BIT(DEVLINK_PARAM_CMODE_DRIVERINIT), NULL, NULL,
+			     mlx5_devlink_pcie_cong_thresh_validate),
+};
+
+static int mlx5_devlink_pcie_cong_params_register(struct devlink *devlink)
+{
+	struct mlx5_core_dev *dev = devlink_priv(devlink);
+	int err;
+
+	if (!mlx5_pcie_cong_event_supported(dev))
+		return 0;
+
+	err = devl_params_register(devlink, mlx5_devlink_pcie_cong_params,
+				   ARRAY_SIZE(mlx5_devlink_pcie_cong_params));
+	if (err)
+		return err;
+
+	mlx5_devlink_pcie_cong_init_values(devlink);
+
+	return 0;
+}
+
+static void mlx5_devlink_pcie_cong_params_unregister(struct devlink *devlink)
+{
+	struct mlx5_core_dev *dev = devlink_priv(devlink);
+
+	if (!mlx5_pcie_cong_event_supported(dev))
+		return;
+
+	devl_params_unregister(devlink, mlx5_devlink_pcie_cong_params,
+			       ARRAY_SIZE(mlx5_devlink_pcie_cong_params));
+}
+
 static int mlx5_devlink_enable_rdma_validate(struct devlink *devlink, u32 id,
 					     union devlink_param_value val,
 					     struct netlink_ext_ack *extack)
@@ -896,6 +995,10 @@ int mlx5_devlink_params_register(struct devlink *devlink)
 	if (err)
 		goto max_uc_list_err;
 
+	err = mlx5_devlink_pcie_cong_params_register(devlink);
+	if (err)
+		goto pcie_cong_err;
+
 	err = mlx5_nv_param_register_dl_params(devlink);
 	if (err)
 		goto nv_param_err;
@@ -903,6 +1006,8 @@ int mlx5_devlink_params_register(struct devlink *devlink)
 	return 0;
 
 nv_param_err:
+	mlx5_devlink_pcie_cong_params_unregister(devlink);
+pcie_cong_err:
 	mlx5_devlink_max_uc_list_params_unregister(devlink);
 max_uc_list_err:
 	mlx5_devlink_auxdev_params_unregister(devlink);
@@ -915,6 +1020,7 @@ auxdev_reg_err:
 void mlx5_devlink_params_unregister(struct devlink *devlink)
 {
 	mlx5_nv_param_unregister_dl_params(devlink);
+	mlx5_devlink_pcie_cong_params_unregister(devlink);
 	mlx5_devlink_max_uc_list_params_unregister(devlink);
 	mlx5_devlink_auxdev_params_unregister(devlink);
 	devl_params_unregister(devlink, mlx5_devlink_params,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/devlink.h b/drivers/net/ethernet/mellanox/mlx5/core/devlink.h
index 74bcdfa70361..c9555119a661 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/devlink.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/devlink.h
@@ -22,6 +22,10 @@ enum mlx5_devlink_param_id {
 	MLX5_DEVLINK_PARAM_ID_ESW_MULTIPORT,
 	MLX5_DEVLINK_PARAM_ID_HAIRPIN_NUM_QUEUES,
 	MLX5_DEVLINK_PARAM_ID_HAIRPIN_QUEUE_SIZE,
+	MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_LOW,
+	MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_HIGH,
+	MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_LOW,
+	MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_HIGH,
 	MLX5_DEVLINK_PARAM_ID_CQE_COMPRESSION_TYPE
 };
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/pcie_cong_event.c b/drivers/net/ethernet/mellanox/mlx5/core/en/pcie_cong_event.c
index 0ed017569a19..2eb666a46f39 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/pcie_cong_event.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/pcie_cong_event.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
 // Copyright (c) 2025, NVIDIA CORPORATION & AFFILIATES.
 
+#include "../devlink.h"
 #include "en.h"
 #include "pcie_cong_event.h"
 
@@ -23,6 +24,7 @@ struct mlx5e_pcie_cong_stats {
 	u32 pci_bw_inbound_low;
 	u32 pci_bw_outbound_high;
 	u32 pci_bw_outbound_low;
+	u32 pci_bw_stale_event;
 };
 
 struct mlx5e_pcie_cong_event {
@@ -41,13 +43,6 @@ struct mlx5e_pcie_cong_event {
 	struct mlx5e_pcie_cong_stats stats;
 };
 
-/* In units of 0.01 % */
-static const struct mlx5e_pcie_cong_thresh default_thresh_config = {
-	.inbound_high = 9000,
-	.inbound_low = 7500,
-	.outbound_high = 9000,
-	.outbound_low = 7500,
-};
 
 static const struct counter_desc mlx5e_pcie_cong_stats_desc[] = {
 	{ MLX5E_DECLARE_STAT(struct mlx5e_pcie_cong_stats,
@@ -58,6 +53,8 @@ static const struct counter_desc mlx5e_pcie_cong_stats_desc[] = {
 			     pci_bw_outbound_high) },
 	{ MLX5E_DECLARE_STAT(struct mlx5e_pcie_cong_stats,
 			     pci_bw_outbound_low) },
+	{ MLX5E_DECLARE_STAT(struct mlx5e_pcie_cong_stats,
+			     pci_bw_stale_event) },
 };
 
 #define NUM_PCIE_CONG_COUNTERS ARRAY_SIZE(mlx5e_pcie_cong_stats_desc)
@@ -218,8 +215,10 @@ static void mlx5e_pcie_cong_event_work(struct work_struct *work)
 	}
 
 	changes = cong_event->state ^ new_cong_state;
-	if (!changes)
+	if (!changes) {
+		cong_event->stats.pci_bw_stale_event++;
 		return;
+	}
 
 	cong_event->state = new_cong_state;
 
@@ -249,8 +248,60 @@ static int mlx5e_pcie_cong_event_handler(struct notifier_block *nb,
 	return NOTIFY_OK;
 }
 
+static int
+mlx5e_pcie_cong_get_thresh_config(struct mlx5_core_dev *dev,
+				  struct mlx5e_pcie_cong_thresh *config)
+{
+	u32 ids[4] = {
+		MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_LOW,
+		MLX5_DEVLINK_PARAM_ID_PCIE_CONG_IN_HIGH,
+		MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_LOW,
+		MLX5_DEVLINK_PARAM_ID_PCIE_CONG_OUT_HIGH,
+	};
+	struct devlink *devlink = priv_to_devlink(dev);
+	union devlink_param_value val[4];
+
+	for (int i = 0; i < 4; i++) {
+		u32 id = ids[i];
+		int err;
+
+		err = devl_param_driverinit_value_get(devlink, id, &val[i]);
+		if (err)
+			return err;
+	}
+
+	config->inbound_low = val[0].vu16;
+	config->inbound_high = val[1].vu16;
+	config->outbound_low = val[2].vu16;
+	config->outbound_high = val[3].vu16;
+
+	return 0;
+}
+
+static int
+mlx5e_thresh_config_validate(struct mlx5_core_dev *mdev,
+			     const struct mlx5e_pcie_cong_thresh *config)
+{
+	int err = 0;
+
+	if (config->inbound_low >= config->inbound_high) {
+		err = -EINVAL;
+		mlx5_core_err(mdev, "PCIe inbound congestion threshold configuration invalid: low (%u) >= high (%u).\n",
+			      config->inbound_low, config->inbound_high);
+	}
+
+	if (config->outbound_low >= config->outbound_high) {
+		err = -EINVAL;
+		mlx5_core_err(mdev, "PCIe outbound congestion threshold configuration invalid: low (%u) >= high (%u).\n",
+			      config->outbound_low, config->outbound_high);
+	}
+
+	return err;
+}
+
 int mlx5e_pcie_cong_event_init(struct mlx5e_priv *priv)
 {
+	struct mlx5e_pcie_cong_thresh thresh_config = {};
 	struct mlx5e_pcie_cong_event *cong_event;
 	struct mlx5_core_dev *mdev = priv->mdev;
 	int err;
@@ -258,6 +309,16 @@ int mlx5e_pcie_cong_event_init(struct mlx5e_priv *priv)
 	if (!mlx5_pcie_cong_event_supported(mdev))
 		return 0;
 
+	err = mlx5e_pcie_cong_get_thresh_config(mdev, &thresh_config);
+	if (WARN_ON(err))
+		return err;
+
+	err = mlx5e_thresh_config_validate(mdev, &thresh_config);
+	if (err) {
+		mlx5_core_err(mdev, "PCIe congestion event feature disabled\n");
+		return err;
+	}
+
 	cong_event = kvzalloc_node(sizeof(*cong_event), GFP_KERNEL,
 				   mdev->priv.numa_node);
 	if (!cong_event)
@@ -269,7 +330,7 @@ int mlx5e_pcie_cong_event_init(struct mlx5e_priv *priv)
 
 	cong_event->priv = priv;
 
-	err = mlx5_cmd_pcie_cong_event_set(mdev, &default_thresh_config,
+	err = mlx5_cmd_pcie_cong_event_set(mdev, &thresh_config,
 					   &cong_event->obj_id);
 	if (err) {
 		mlx5_core_warn(mdev, "Error creating a PCIe congestion event object\n");