Merge branch 'mlx5-misc-fixes'

Tariq Toukan says:

====================
mlx5 misc fixes

This patchset provides bug fixes to mlx5 driver.

This is V2 of the series previously submitted as PR by Saeed:
https://lore.kernel.org/netdev/20240326144646.2078893-1-saeed@kernel.org/T/

Series generated against:
commit 237f3cf13b ("xsk: validate user input for XDP_{UMEM|COMPLETION}_FILL_RING")
====================

Link: https://lore.kernel.org/r/20240409190820.227554-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

drivers/net/ethernet/mellanox/mlx5/core/en/ptp.h

@@ -95,9 +95,15 @@ static inline void mlx5e_ptp_metadata_fifo_push(struct mlx5e_ptp_metadata_fifo *
 }
 
 static inline u8
+mlx5e_ptp_metadata_fifo_peek(struct mlx5e_ptp_metadata_fifo *fifo)
+{
+	return fifo->data[fifo->mask & fifo->cc];
+}
+
+static inline void
 mlx5e_ptp_metadata_fifo_pop(struct mlx5e_ptp_metadata_fifo *fifo)
 {
-	return fifo->data[fifo->mask & fifo->cc++];
+	fifo->cc++;
 }
 
 static inline void

drivers/net/ethernet/mellanox/mlx5/core/en/qos.c

@@ -83,24 +83,25 @@ int mlx5e_open_qos_sq(struct mlx5e_priv *priv, struct mlx5e_channels *chs,
 
 	txq_ix = mlx5e_qid_from_qos(chs, node_qid);
 
-	WARN_ON(node_qid > priv->htb_max_qos_sqs);
-	if (node_qid == priv->htb_max_qos_sqs) {
-		struct mlx5e_sq_stats *stats, **stats_list = NULL;
+	WARN_ON(node_qid >= mlx5e_htb_cur_leaf_nodes(priv->htb));
+	if (!priv->htb_qos_sq_stats) {
+		struct mlx5e_sq_stats **stats_list;
 
-		if (priv->htb_max_qos_sqs == 0) {
-			stats_list = kvcalloc(mlx5e_qos_max_leaf_nodes(priv->mdev),
-					      sizeof(*stats_list),
-					      GFP_KERNEL);
-			if (!stats_list)
-				return -ENOMEM;
-		}
-		stats = kzalloc(sizeof(*stats), GFP_KERNEL);
-		if (!stats) {
-			kvfree(stats_list);
+		stats_list = kvcalloc(mlx5e_qos_max_leaf_nodes(priv->mdev),
+				      sizeof(*stats_list), GFP_KERNEL);
+		if (!stats_list)
 			return -ENOMEM;
-		}
-		if (stats_list)
-			WRITE_ONCE(priv->htb_qos_sq_stats, stats_list);
+
+		WRITE_ONCE(priv->htb_qos_sq_stats, stats_list);
+	}
+
+	if (!priv->htb_qos_sq_stats[node_qid]) {
+		struct mlx5e_sq_stats *stats;
+
+		stats = kzalloc(sizeof(*stats), GFP_KERNEL);
+		if (!stats)
+			return -ENOMEM;
+
 		WRITE_ONCE(priv->htb_qos_sq_stats[node_qid], stats);
 		/* Order htb_max_qos_sqs increment after writing the array pointer.
 		 * Pairs with smp_load_acquire in en_stats.c.

drivers/net/ethernet/mellanox/mlx5/core/en/rqt.c

@@ -179,6 +179,13 @@ u32 mlx5e_rqt_size(struct mlx5_core_dev *mdev, unsigned int num_channels)
 	return min_t(u32, rqt_size, max_cap_rqt_size);
 }
 
+#define MLX5E_MAX_RQT_SIZE_ALLOWED_WITH_XOR8_HASH 256
+
+unsigned int mlx5e_rqt_max_num_channels_allowed_for_xor8(void)
+{
+	return MLX5E_MAX_RQT_SIZE_ALLOWED_WITH_XOR8_HASH / MLX5E_UNIFORM_SPREAD_RQT_FACTOR;
+}
+
 void mlx5e_rqt_destroy(struct mlx5e_rqt *rqt)
 {
 	mlx5_core_destroy_rqt(rqt->mdev, rqt->rqtn);

drivers/net/ethernet/mellanox/mlx5/core/en/rqt.h

@@ -38,6 +38,7 @@ static inline u32 mlx5e_rqt_get_rqtn(struct mlx5e_rqt *rqt)
 }
 
 u32 mlx5e_rqt_size(struct mlx5_core_dev *mdev, unsigned int num_channels);
+unsigned int mlx5e_rqt_max_num_channels_allowed_for_xor8(void);
 int mlx5e_rqt_redirect_direct(struct mlx5e_rqt *rqt, u32 rqn, u32 *vhca_id);
 int mlx5e_rqt_redirect_indir(struct mlx5e_rqt *rqt, u32 *rqns, u32 *vhca_ids,
 			     unsigned int num_rqns,

drivers/net/ethernet/mellanox/mlx5/core/en/selq.c

@@ -57,6 +57,7 @@ int mlx5e_selq_init(struct mlx5e_selq *selq, struct mutex *state_lock)
 
 void mlx5e_selq_cleanup(struct mlx5e_selq *selq)
 {
+	mutex_lock(selq->state_lock);
 	WARN_ON_ONCE(selq->is_prepared);
 
 	kvfree(selq->standby);
@@ -67,6 +68,7 @@ void mlx5e_selq_cleanup(struct mlx5e_selq *selq)
 
 	kvfree(selq->standby);
 	selq->standby = NULL;
+	mutex_unlock(selq->state_lock);
 }
 
 void mlx5e_selq_prepare_params(struct mlx5e_selq *selq, struct mlx5e_params *params)

drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c

@@ -451,6 +451,34 @@ int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
 
 	mutex_lock(&priv->state_lock);
 
+	if (mlx5e_rx_res_get_current_hash(priv->rx_res).hfunc == ETH_RSS_HASH_XOR) {
+		unsigned int xor8_max_channels = mlx5e_rqt_max_num_channels_allowed_for_xor8();
+
+		if (count > xor8_max_channels) {
+			err = -EINVAL;
+			netdev_err(priv->netdev, "%s: Requested number of channels (%d) exceeds the maximum allowed by the XOR8 RSS hfunc (%d)\n",
+				   __func__, count, xor8_max_channels);
+			goto out;
+		}
+	}
+
+	/* If RXFH is configured, changing the channels number is allowed only if
+	 * it does not require resizing the RSS table. This is because the previous
+	 * configuration may no longer be compatible with the new RSS table.
+	 */
+	if (netif_is_rxfh_configured(priv->netdev)) {
+		int cur_rqt_size = mlx5e_rqt_size(priv->mdev, cur_params->num_channels);
+		int new_rqt_size = mlx5e_rqt_size(priv->mdev, count);
+
+		if (new_rqt_size != cur_rqt_size) {
+			err = -EINVAL;
+			netdev_err(priv->netdev,
+				   "%s: RXFH is configured, block changing channels number that affects RSS table size (new: %d, current: %d)\n",
+				   __func__, new_rqt_size, cur_rqt_size);
+			goto out;
+		}
+	}
+
 	/* Don't allow changing the number of channels if HTB offload is active,
 	 * because the numeration of the QoS SQs will change, while per-queue
 	 * qdiscs are attached.
@@ -1281,17 +1309,30 @@ int mlx5e_set_rxfh(struct net_device *dev, struct ethtool_rxfh_param *rxfh,
 	struct mlx5e_priv *priv = netdev_priv(dev);
 	u32 *rss_context = &rxfh->rss_context;
 	u8 hfunc = rxfh->hfunc;
+	unsigned int count;
 	int err;
 
 	mutex_lock(&priv->state_lock);
+
+	count = priv->channels.params.num_channels;
+
+	if (hfunc == ETH_RSS_HASH_XOR) {
+		unsigned int xor8_max_channels = mlx5e_rqt_max_num_channels_allowed_for_xor8();
+
+		if (count > xor8_max_channels) {
+			err = -EINVAL;
+			netdev_err(priv->netdev, "%s: Cannot set RSS hash function to XOR, current number of channels (%d) exceeds the maximum allowed for XOR8 RSS hfunc (%d)\n",
+				   __func__, count, xor8_max_channels);
+			goto unlock;
+		}
+	}
+
 	if (*rss_context && rxfh->rss_delete) {
 		err = mlx5e_rx_res_rss_destroy(priv->rx_res, *rss_context);
 		goto unlock;
 	}
 
 	if (*rss_context == ETH_RXFH_CONTEXT_ALLOC) {
-		unsigned int count = priv->channels.params.num_channels;
-
 		err = mlx5e_rx_res_rss_init(priv->rx_res, rss_context, count);
 		if (err)
 			goto unlock;

drivers/net/ethernet/mellanox/mlx5/core/en_main.c

@@ -5726,9 +5726,7 @@ void mlx5e_priv_cleanup(struct mlx5e_priv *priv)
 	kfree(priv->tx_rates);
 	kfree(priv->txq2sq);
 	destroy_workqueue(priv->wq);
-	mutex_lock(&priv->state_lock);
 	mlx5e_selq_cleanup(&priv->selq);
-	mutex_unlock(&priv->state_lock);
 	free_cpumask_var(priv->scratchpad.cpumask);
 
 	for (i = 0; i < priv->htb_max_qos_sqs; i++)

drivers/net/ethernet/mellanox/mlx5/core/en_tx.c

@@ -398,6 +398,8 @@ mlx5e_txwqe_complete(struct mlx5e_txqsq *sq, struct sk_buff *skb,
 		     (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP))) {
 		u8 metadata_index = be32_to_cpu(eseg->flow_table_metadata);
 
+		mlx5e_ptp_metadata_fifo_pop(&sq->ptpsq->metadata_freelist);
+
 		mlx5e_skb_cb_hwtstamp_init(skb);
 		mlx5e_ptp_metadata_map_put(&sq->ptpsq->metadata_map, skb,
 					   metadata_index);
@@ -496,9 +498,6 @@ mlx5e_sq_xmit_wqe(struct mlx5e_txqsq *sq, struct sk_buff *skb,
 
 err_drop:
 	stats->dropped++;
-	if (unlikely(sq->ptpsq && (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)))
-		mlx5e_ptp_metadata_fifo_push(&sq->ptpsq->metadata_freelist,
-					     be32_to_cpu(eseg->flow_table_metadata));
 	dev_kfree_skb_any(skb);
 	mlx5e_tx_flush(sq);
 }
@@ -657,7 +656,7 @@ static void mlx5e_cqe_ts_id_eseg(struct mlx5e_ptpsq *ptpsq, struct sk_buff *skb,
 {
 	if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP))
 		eseg->flow_table_metadata =
-			cpu_to_be32(mlx5e_ptp_metadata_fifo_pop(&ptpsq->metadata_freelist));
+			cpu_to_be32(mlx5e_ptp_metadata_fifo_peek(&ptpsq->metadata_freelist));
 }
 
 static void mlx5e_txwqe_build_eseg(struct mlx5e_priv *priv, struct mlx5e_txqsq *sq,

drivers/net/ethernet/mellanox/mlx5/core/eswitch.c

@@ -1868,6 +1868,7 @@ int mlx5_eswitch_init(struct mlx5_core_dev *dev)
 	if (err)
 		goto abort;
 
+	dev->priv.eswitch = esw;
 	err = esw_offloads_init(esw);
 	if (err)
 		goto reps_err;
@@ -1892,11 +1893,6 @@ int mlx5_eswitch_init(struct mlx5_core_dev *dev)
 		esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_BASIC;
 	else
 		esw->offloads.encap = DEVLINK_ESWITCH_ENCAP_MODE_NONE;
-	if (MLX5_ESWITCH_MANAGER(dev) &&
-	    mlx5_esw_vport_match_metadata_supported(esw))
-		esw->flags |= MLX5_ESWITCH_VPORT_MATCH_METADATA;
-
-	dev->priv.eswitch = esw;
 	BLOCKING_INIT_NOTIFIER_HEAD(&esw->n_head);
 
 	esw_info(dev,
@@ -1908,6 +1904,7 @@ int mlx5_eswitch_init(struct mlx5_core_dev *dev)
 
 reps_err:
 	mlx5_esw_vports_cleanup(esw);
+	dev->priv.eswitch = NULL;
 abort:
 	if (esw->work_queue)
 		destroy_workqueue(esw->work_queue);
@@ -1926,7 +1923,6 @@ void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
 
 	esw_info(esw->dev, "cleanup\n");
 
-	esw->dev->priv.eswitch = NULL;
 	destroy_workqueue(esw->work_queue);
 	WARN_ON(refcount_read(&esw->qos.refcnt));
 	mutex_destroy(&esw->state_lock);
@@ -1937,6 +1933,7 @@ void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
 	mutex_destroy(&esw->offloads.encap_tbl_lock);
 	mutex_destroy(&esw->offloads.decap_tbl_lock);
 	esw_offloads_cleanup(esw);
+	esw->dev->priv.eswitch = NULL;
 	mlx5_esw_vports_cleanup(esw);
 	debugfs_remove_recursive(esw->debugfs_root);
 	devl_params_unregister(priv_to_devlink(esw->dev), mlx5_eswitch_params,

drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c

@@ -43,6 +43,7 @@
 #include "rdma.h"
 #include "en.h"
 #include "fs_core.h"
+#include "lib/mlx5.h"
 #include "lib/devcom.h"
 #include "lib/eq.h"
 #include "lib/fs_chains.h"
@@ -2476,6 +2477,10 @@ int esw_offloads_init(struct mlx5_eswitch *esw)
 	if (err)
 		return err;
 
+	if (MLX5_ESWITCH_MANAGER(esw->dev) &&
+	    mlx5_esw_vport_match_metadata_supported(esw))
+		esw->flags |= MLX5_ESWITCH_VPORT_MATCH_METADATA;
+
 	err = devl_params_register(priv_to_devlink(esw->dev),
 				   esw_devlink_params,
 				   ARRAY_SIZE(esw_devlink_params));
@@ -3707,6 +3712,12 @@ int mlx5_devlink_eswitch_mode_set(struct devlink *devlink, u16 mode,
 	if (esw_mode_from_devlink(mode, &mlx5_mode))
 		return -EINVAL;
 
+	if (mode == DEVLINK_ESWITCH_MODE_SWITCHDEV && mlx5_get_sd(esw->dev)) {
+		NL_SET_ERR_MSG_MOD(extack,
+				   "Can't change E-Switch mode to switchdev when multi-PF netdev (Socket Direct) is configured.");
+		return -EPERM;
+	}
+
 	mlx5_lag_disable_change(esw->dev);
 	err = mlx5_esw_try_lock(esw);
 	if (err < 0) {

drivers/net/ethernet/mellanox/mlx5/core/fs_core.c

@@ -1664,6 +1664,16 @@ static int create_auto_flow_group(struct mlx5_flow_table *ft,
 	return err;
 }
 
+static bool mlx5_pkt_reformat_cmp(struct mlx5_pkt_reformat *p1,
+				  struct mlx5_pkt_reformat *p2)
+{
+	return p1->owner == p2->owner &&
+		(p1->owner == MLX5_FLOW_RESOURCE_OWNER_FW ?
+		 p1->id == p2->id :
+		 mlx5_fs_dr_action_get_pkt_reformat_id(p1) ==
+		 mlx5_fs_dr_action_get_pkt_reformat_id(p2));
+}
+
 static bool mlx5_flow_dests_cmp(struct mlx5_flow_destination *d1,
 				struct mlx5_flow_destination *d2)
 {
@@ -1675,8 +1685,8 @@ static bool mlx5_flow_dests_cmp(struct mlx5_flow_destination *d1,
 		     ((d1->vport.flags & MLX5_FLOW_DEST_VPORT_VHCA_ID) ?
 		      (d1->vport.vhca_id == d2->vport.vhca_id) : true) &&
 		     ((d1->vport.flags & MLX5_FLOW_DEST_VPORT_REFORMAT_ID) ?
-		      (d1->vport.pkt_reformat->id ==
-		       d2->vport.pkt_reformat->id) : true)) ||
+		      mlx5_pkt_reformat_cmp(d1->vport.pkt_reformat,
+					    d2->vport.pkt_reformat) : true)) ||
 		    (d1->type == MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE &&
 		     d1->ft == d2->ft) ||
 		    (d1->type == MLX5_FLOW_DESTINATION_TYPE_TIR &&
@@ -1808,8 +1818,9 @@ static struct mlx5_flow_handle *add_rule_fg(struct mlx5_flow_group *fg,
 	}
 	trace_mlx5_fs_set_fte(fte, false);
 
+	/* Link newly added rules into the tree. */
 	for (i = 0; i < handle->num_rules; i++) {
-		if (refcount_read(&handle->rule[i]->node.refcount) == 1) {
+		if (!handle->rule[i]->node.parent) {
 			tree_add_node(&handle->rule[i]->node, &fte->node);
 			trace_mlx5_fs_add_rule(handle->rule[i]);
 		}

drivers/net/ethernet/mellanox/mlx5/core/main.c

@@ -1480,6 +1480,14 @@ int mlx5_init_one_devl_locked(struct mlx5_core_dev *dev)
 	if (err)
 		goto err_register;
 
+	err = mlx5_crdump_enable(dev);
+	if (err)
+		mlx5_core_err(dev, "mlx5_crdump_enable failed with error code %d\n", err);
+
+	err = mlx5_hwmon_dev_register(dev);
+	if (err)
+		mlx5_core_err(dev, "mlx5_hwmon_dev_register failed with error code %d\n", err);
+
 	mutex_unlock(&dev->intf_state_mutex);
 	return 0;
 
@@ -1505,7 +1513,10 @@ int mlx5_init_one(struct mlx5_core_dev *dev)
 	int err;
 
 	devl_lock(devlink);
+	devl_register(devlink);
 	err = mlx5_init_one_devl_locked(dev);
+	if (err)
+		devl_unregister(devlink);
 	devl_unlock(devlink);
 	return err;
 }
@@ -1517,6 +1528,8 @@ void mlx5_uninit_one(struct mlx5_core_dev *dev)
 	devl_lock(devlink);
 	mutex_lock(&dev->intf_state_mutex);
 
+	mlx5_hwmon_dev_unregister(dev);
+	mlx5_crdump_disable(dev);
 	mlx5_unregister_device(dev);
 
 	if (!test_bit(MLX5_INTERFACE_STATE_UP, &dev->intf_state)) {
@@ -1534,6 +1547,7 @@ void mlx5_uninit_one(struct mlx5_core_dev *dev)
 	mlx5_function_teardown(dev, true);
 out:
 	mutex_unlock(&dev->intf_state_mutex);
+	devl_unregister(devlink);
 	devl_unlock(devlink);
 }
 
@@ -1680,16 +1694,20 @@ int mlx5_init_one_light(struct mlx5_core_dev *dev)
 	}
 
 	devl_lock(devlink);
+	devl_register(devlink);
+
 	err = mlx5_devlink_params_register(priv_to_devlink(dev));
-	devl_unlock(devlink);
 	if (err) {
 		mlx5_core_warn(dev, "mlx5_devlink_param_reg err = %d\n", err);
 		goto query_hca_caps_err;
 	}
 
+	devl_unlock(devlink);
 	return 0;
 
 query_hca_caps_err:
+	devl_unregister(devlink);
+	devl_unlock(devlink);
 	mlx5_function_disable(dev, true);
 out:
 	dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
@@ -1702,6 +1720,7 @@ void mlx5_uninit_one_light(struct mlx5_core_dev *dev)
 
 	devl_lock(devlink);
 	mlx5_devlink_params_unregister(priv_to_devlink(dev));
+	devl_unregister(devlink);
 	devl_unlock(devlink);
 	if (dev->state != MLX5_DEVICE_STATE_UP)
 		return;
@@ -1943,16 +1962,7 @@ static int probe_one(struct pci_dev *pdev, const struct pci_device_id *id)
 		goto err_init_one;
 	}
 
-	err = mlx5_crdump_enable(dev);
-	if (err)
-		dev_err(&pdev->dev, "mlx5_crdump_enable failed with error code %d\n", err);
-
-	err = mlx5_hwmon_dev_register(dev);
-	if (err)
-		mlx5_core_err(dev, "mlx5_hwmon_dev_register failed with error code %d\n", err);
-
 	pci_save_state(pdev);
-	devlink_register(devlink);
 	return 0;
 
 err_init_one:
@@ -1973,16 +1983,9 @@ static void remove_one(struct pci_dev *pdev)
 	struct devlink *devlink = priv_to_devlink(dev);
 
 	set_bit(MLX5_BREAK_FW_WAIT, &dev->intf_state);
-	/* mlx5_drain_fw_reset() and mlx5_drain_health_wq() are using
-	 * devlink notify APIs.
-	 * Hence, we must drain them before unregistering the devlink.
-	 */
 	mlx5_drain_fw_reset(dev);
 	mlx5_drain_health_wq(dev);
-	devlink_unregister(devlink);
 	mlx5_sriov_disable(pdev, false);
-	mlx5_hwmon_dev_unregister(dev);
-	mlx5_crdump_disable(dev);
 	mlx5_uninit_one(dev);
 	mlx5_pci_close(dev);
 	mlx5_mdev_uninit(dev);

drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c

@@ -19,6 +19,7 @@
 #define MLX5_IRQ_CTRL_SF_MAX 8
 /* min num of vectors for SFs to be enabled */
 #define MLX5_IRQ_VEC_COMP_BASE_SF 2
+#define MLX5_IRQ_VEC_COMP_BASE 1
 
 #define MLX5_EQ_SHARE_IRQ_MAX_COMP (8)
 #define MLX5_EQ_SHARE_IRQ_MAX_CTRL (UINT_MAX)
@@ -246,6 +247,7 @@ static void irq_set_name(struct mlx5_irq_pool *pool, char *name, int vecidx)
 		return;
 	}
 
+	vecidx -= MLX5_IRQ_VEC_COMP_BASE;
 	snprintf(name, MLX5_MAX_IRQ_NAME, "mlx5_comp%d", vecidx);
 }
 
@@ -585,7 +587,7 @@ struct mlx5_irq *mlx5_irq_request_vector(struct mlx5_core_dev *dev, u16 cpu,
 	struct mlx5_irq_table *table = mlx5_irq_table_get(dev);
 	struct mlx5_irq_pool *pool = table->pcif_pool;
 	struct irq_affinity_desc af_desc;
-	int offset = 1;
+	int offset = MLX5_IRQ_VEC_COMP_BASE;
 
 	if (!pool->xa_num_irqs.max)
 		offset = 0;

drivers/net/ethernet/mellanox/mlx5/core/sf/dev/driver.c

@@ -101,7 +101,6 @@ static void mlx5_sf_dev_remove(struct auxiliary_device *adev)
 	devlink = priv_to_devlink(mdev);
 	set_bit(MLX5_BREAK_FW_WAIT, &mdev->intf_state);
 	mlx5_drain_health_wq(mdev);
-	devlink_unregister(devlink);
 	if (mlx5_dev_is_lightweight(mdev))
 		mlx5_uninit_one_light(mdev);
 	else