Merge tag 'mlx5-updates-2023-10-10' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5-updates-2023-10-10

1) Adham Faris, Increase max supported channels number to 256

2) Leon Romanovsky, Allow IPsec soft/hard limits in bytes

3) Shay Drory, Replace global mlx5_intf_lock with
   HCA devcom component lock

4) Wei Zhang, Optimize SF creation flow

During SF creation, HCA state gets changed from INVALID to
IN_USE step by step. Accordingly, FW sends vhca event to
driver to inform about this state change asynchronously.
Each vhca event is critical because all related SW/FW
operations are triggered by it.

Currently there is only a single mlx5 general event handler
which not only handles vhca event but many other events.
This incurs huge bottleneck because all events are forced
to be handled in serial manner.

Moreover, all SFs share same table_lock which inevitably
impacts each other when they are created in parallel.

This series will solve this issue by:

1. A dedicated vhca event handler is introduced to eliminate
   the mutual impact with other mlx5 events.
2. Max FW threads work queues are employed in the vhca event
   handler to fully utilize FW capability.
3. Redesign SF active work logic to completely remove
   table_lock.

With above optimization, SF creation time is reduced by 25%,
i.e. from 80s to 60s when creating 100 SFs.

Patches summary:

Patch 1 - implement dedicated vhca event handler with max FW
          cmd threads of work queues.
Patch 2 - remove table_lock by redesigning SF active work
          logic.

* tag 'mlx5-updates-2023-10-10' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5e: Allow IPsec soft/hard limits in bytes
  net/mlx5e: Increase max supported channels number to 256
  net/mlx5e: Preparations for supporting larger number of channels
  net/mlx5e: Refactor mlx5e_rss_init() and mlx5e_rss_free() API's
  net/mlx5e: Refactor mlx5e_rss_set_rxfh() and mlx5e_rss_get_rxfh()
  net/mlx5e: Refactor rx_res_init() and rx_res_free() APIs
  net/mlx5e: Use PTR_ERR_OR_ZERO() to simplify code
  net/mlx5: Use PTR_ERR_OR_ZERO() to simplify code
  net/mlx5: fix config name in Kconfig parameter documentation
  net/mlx5: Remove unused declaration
  net/mlx5: Replace global mlx5_intf_lock with HCA devcom component lock
  net/mlx5: Refactor LAG peer device lookout bus logic to mlx5 devcom
  net/mlx5: Avoid false positive lockdep warning by adding lock_class_key
  net/mlx5: Redesign SF active work to remove table_lock
  net/mlx5: Parallelize vhca event handling
====================

Link: https://lore.kernel.org/r/20231014171908.290428-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

34 files changed, 544 insertions, 368 deletions

diff --git a/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/kconfig.rst b/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/kconfig.rst
index 0a42c3395ffa..20d3b7e87049 100644
--- a/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/kconfig.rst
+++ b/Documentation/networking/device_drivers/ethernet/mellanox/mlx5/kconfig.rst
@@ -67,7 +67,7 @@ Enabling the driver and kconfig options
 |    Enables :ref:`IPSec XFRM cryptography-offload acceleration <xfrm_device>`.
 
 
-**CONFIG_MLX5_EN_MACSEC=(y/n)**
+**CONFIG_MLX5_MACSEC=(y/n)**
 
 |    Build support for MACsec cryptography-offload acceleration in the NIC.
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/dev.c b/drivers/net/ethernet/mellanox/mlx5/core/dev.c
index 1fc03480c2ff..cf0477f53dc4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/dev.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/dev.c
@@ -38,8 +38,6 @@
 #include "devlink.h"
 #include "lag/lag.h"
 
-/* intf dev list mutex */
-static DEFINE_MUTEX(mlx5_intf_mutex);
 static DEFINE_IDA(mlx5_adev_ida);
 
 static bool is_eth_rep_supported(struct mlx5_core_dev *dev)
@@ -337,9 +335,9 @@ static void del_adev(struct auxiliary_device *adev)
 
 void mlx5_dev_set_lightweight(struct mlx5_core_dev *dev)
 {
-	mutex_lock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
 	dev->priv.flags |= MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
-	mutex_unlock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
 }
 
 bool mlx5_dev_is_lightweight(struct mlx5_core_dev *dev)
@@ -355,7 +353,7 @@ int mlx5_attach_device(struct mlx5_core_dev *dev)
 	int ret = 0, i;
 
 	devl_assert_locked(priv_to_devlink(dev));
-	mutex_lock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
 	priv->flags &= ~MLX5_PRIV_FLAGS_DETACH;
 	for (i = 0; i < ARRAY_SIZE(mlx5_adev_devices); i++) {
 		if (!priv->adev[i]) {
@@ -400,7 +398,7 @@ int mlx5_attach_device(struct mlx5_core_dev *dev)
 			break;
 		}
 	}
-	mutex_unlock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
 	return ret;
 }
 
@@ -413,7 +411,7 @@ void mlx5_detach_device(struct mlx5_core_dev *dev, bool suspend)
 	int i;
 
 	devl_assert_locked(priv_to_devlink(dev));
-	mutex_lock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
 	for (i = ARRAY_SIZE(mlx5_adev_devices) - 1; i >= 0; i--) {
 		if (!priv->adev[i])
 			continue;
@@ -443,7 +441,7 @@ skip_suspend:
 		priv->adev[i] = NULL;
 	}
 	priv->flags |= MLX5_PRIV_FLAGS_DETACH;
-	mutex_unlock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
 }
 
 int mlx5_register_device(struct mlx5_core_dev *dev)
@@ -451,10 +449,10 @@ int mlx5_register_device(struct mlx5_core_dev *dev)
 	int ret;
 
 	devl_assert_locked(priv_to_devlink(dev));
-	mutex_lock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
 	dev->priv.flags &= ~MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
 	ret = mlx5_rescan_drivers_locked(dev);
-	mutex_unlock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
 	if (ret)
 		mlx5_unregister_device(dev);
 
@@ -464,10 +462,10 @@ int mlx5_register_device(struct mlx5_core_dev *dev)
 void mlx5_unregister_device(struct mlx5_core_dev *dev)
 {
 	devl_assert_locked(priv_to_devlink(dev));
-	mutex_lock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_lock(dev->priv.hca_devcom_comp);
 	dev->priv.flags = MLX5_PRIV_FLAGS_DISABLE_ALL_ADEV;
 	mlx5_rescan_drivers_locked(dev);
-	mutex_unlock(&mlx5_intf_mutex);
+	mlx5_devcom_comp_unlock(dev->priv.hca_devcom_comp);
 }
 
 static int add_drivers(struct mlx5_core_dev *dev)
@@ -545,7 +543,6 @@ int mlx5_rescan_drivers_locked(struct mlx5_core_dev *dev)
 {
 	struct mlx5_priv *priv = &dev->priv;
 
-	lockdep_assert_held(&mlx5_intf_mutex);
 	if (priv->flags & MLX5_PRIV_FLAGS_DETACH)
 		return 0;
 
@@ -565,85 +562,3 @@ bool mlx5_same_hw_devs(struct mlx5_core_dev *dev, struct mlx5_core_dev *peer_dev
 
 	return (fsystem_guid && psystem_guid && fsystem_guid == psystem_guid);
 }
-
-static u32 mlx5_gen_pci_id(const struct mlx5_core_dev *dev)
-{
-	return (u32)((pci_domain_nr(dev->pdev->bus) << 16) |
-		     (dev->pdev->bus->number << 8) |
-		     PCI_SLOT(dev->pdev->devfn));
-}
-
-static int _next_phys_dev(struct mlx5_core_dev *mdev,
-			  const struct mlx5_core_dev *curr)
-{
-	if (!mlx5_core_is_pf(mdev))
-		return 0;
-
-	if (mdev == curr)
-		return 0;
-
-	if (!mlx5_same_hw_devs(mdev, (struct mlx5_core_dev *)curr) &&
-	    mlx5_gen_pci_id(mdev) != mlx5_gen_pci_id(curr))
-		return 0;
-
-	return 1;
-}
-
-static void *pci_get_other_drvdata(struct device *this, struct device *other)
-{
-	if (this->driver != other->driver)
-		return NULL;
-
-	return pci_get_drvdata(to_pci_dev(other));
-}
-
-static int next_phys_dev_lag(struct device *dev, const void *data)
-{
-	struct mlx5_core_dev *mdev, *this = (struct mlx5_core_dev *)data;
-
-	mdev = pci_get_other_drvdata(this->device, dev);
-	if (!mdev)
-		return 0;
-
-	if (!mlx5_lag_is_supported(mdev))
-		return 0;
-
-	return _next_phys_dev(mdev, data);
-}
-
-static struct mlx5_core_dev *mlx5_get_next_dev(struct mlx5_core_dev *dev,
-					       int (*match)(struct device *dev, const void *data))
-{
-	struct device *next;
-
-	if (!mlx5_core_is_pf(dev))
-		return NULL;
-
-	next = bus_find_device(&pci_bus_type, NULL, dev, match);
-	if (!next)
-		return NULL;
-
-	put_device(next);
-	return pci_get_drvdata(to_pci_dev(next));
-}
-
-/* Must be called with intf_mutex held */
-struct mlx5_core_dev *mlx5_get_next_phys_dev_lag(struct mlx5_core_dev *dev)
-{
-	lockdep_assert_held(&mlx5_intf_mutex);
-	return mlx5_get_next_dev(dev, &next_phys_dev_lag);
-}
-
-void mlx5_dev_list_lock(void)
-{
-	mutex_lock(&mlx5_intf_mutex);
-}
-void mlx5_dev_list_unlock(void)
-{
-	mutex_unlock(&mlx5_intf_mutex);
-}
-
-int mlx5_dev_list_trylock(void)
-{
-	return mutex_trylock(&mlx5_intf_mutex);
-}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 2b3254e33fe5..b2a5da9739d2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -141,7 +141,7 @@ struct page_pool;
 #define MLX5E_PARAMS_DEFAULT_MIN_RX_WQES_MPW            0x2
 
 #define MLX5E_MIN_NUM_CHANNELS         0x1
-#define MLX5E_MAX_NUM_CHANNELS         (MLX5E_INDIR_RQT_SIZE / 2)
+#define MLX5E_MAX_NUM_CHANNELS         256
 #define MLX5E_TX_CQ_POLL_BUDGET        128
 #define MLX5E_TX_XSK_POLL_BUDGET       64
 #define MLX5E_SQ_RECOVER_MIN_INTERVAL  500 /* msecs */
@@ -200,7 +200,8 @@ static inline int mlx5e_get_max_num_channels(struct mlx5_core_dev *mdev)
 {
 	return is_kdump_kernel() ?
 		MLX5E_MIN_NUM_CHANNELS :
-		min_t(int, mlx5_comp_vectors_max(mdev), MLX5E_MAX_NUM_CHANNELS);
+		min3(mlx5_comp_vectors_max(mdev), (u32)MLX5E_MAX_NUM_CHANNELS,
+		     (u32)(1 << MLX5_CAP_GEN(mdev, log_max_rqt_size)));
 }
 
 /* The maximum WQE size can be retrieved by max_wqe_sz_sq in
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h b/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
index e5a44b0b9616..4d6225e0eec7 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/fs.h
@@ -150,7 +150,6 @@ struct mlx5e_flow_steering *mlx5e_fs_init(const struct mlx5e_profile *profile,
 					  struct dentry *dfs_root);
 void mlx5e_fs_cleanup(struct mlx5e_flow_steering *fs);
 struct mlx5e_vlan_table *mlx5e_fs_get_vlan(struct mlx5e_flow_steering *fs);
-void mlx5e_fs_set_tc(struct mlx5e_flow_steering *fs, struct mlx5e_tc_table *tc);
 struct mlx5e_tc_table *mlx5e_fs_get_tc(struct mlx5e_flow_steering *fs);
 struct mlx5e_l2_table *mlx5e_fs_get_l2(struct mlx5e_flow_steering *fs);
 struct mlx5_flow_namespace *mlx5e_fs_get_ns(struct mlx5e_flow_steering *fs, bool egress);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.c b/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.c
index b915fb29dd2c..7b8ff7a71003 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.c
@@ -9,7 +9,7 @@ void mlx5e_rss_params_indir_init_uniform(struct mlx5e_rss_params_indir *indir,
 {
 	unsigned int i;
 
-	for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++)
+	for (i = 0; i < indir->actual_table_size; i++)
 		indir->table[i] = i % num_channels;
 }
 
@@ -45,9 +45,9 @@ static int mlx5e_rqt_init(struct mlx5e_rqt *rqt, struct mlx5_core_dev *mdev,
 }
 
 int mlx5e_rqt_init_direct(struct mlx5e_rqt *rqt, struct mlx5_core_dev *mdev,
-			  bool indir_enabled, u32 init_rqn)
+			  bool indir_enabled, u32 init_rqn, u32 indir_table_size)
 {
-	u16 max_size = indir_enabled ? MLX5E_INDIR_RQT_SIZE : 1;
+	u16 max_size = indir_enabled ? indir_table_size : 1;
 
 	return mlx5e_rqt_init(rqt, mdev, max_size, &init_rqn, 1);
 }
@@ -68,11 +68,11 @@ static int mlx5e_calc_indir_rqns(u32 *rss_rqns, u32 *rqns, unsigned int num_rqns
 {
 	unsigned int i;
 
-	for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++) {
+	for (i = 0; i < indir->actual_table_size; i++) {
 		unsigned int ix = i;
 
 		if (hfunc == ETH_RSS_HASH_XOR)
-			ix = mlx5e_bits_invert(ix, ilog2(MLX5E_INDIR_RQT_SIZE));
+			ix = mlx5e_bits_invert(ix, ilog2(indir->actual_table_size));
 
 		ix = indir->table[ix];
 
@@ -94,7 +94,7 @@ int mlx5e_rqt_init_indir(struct mlx5e_rqt *rqt, struct mlx5_core_dev *mdev,
 	u32 *rss_rqns;
 	int err;
 
-	rss_rqns = kvmalloc_array(MLX5E_INDIR_RQT_SIZE, sizeof(*rss_rqns), GFP_KERNEL);
+	rss_rqns = kvmalloc_array(indir->actual_table_size, sizeof(*rss_rqns), GFP_KERNEL);
 	if (!rss_rqns)
 		return -ENOMEM;
 
@@ -102,13 +102,25 @@ int mlx5e_rqt_init_indir(struct mlx5e_rqt *rqt, struct mlx5_core_dev *mdev,
 	if (err)
 		goto out;
 
-	err = mlx5e_rqt_init(rqt, mdev, MLX5E_INDIR_RQT_SIZE, rss_rqns, MLX5E_INDIR_RQT_SIZE);
+	err = mlx5e_rqt_init(rqt, mdev, indir->max_table_size, rss_rqns,
+			     indir->actual_table_size);
 
 out:
 	kvfree(rss_rqns);
 	return err;
 }
 
+#define MLX5E_UNIFORM_SPREAD_RQT_FACTOR 2
+
+u32 mlx5e_rqt_size(struct mlx5_core_dev *mdev, unsigned int num_channels)
+{
+	u32 rqt_size = max_t(u32, MLX5E_INDIR_MIN_RQT_SIZE,
+			     roundup_pow_of_two(num_channels * MLX5E_UNIFORM_SPREAD_RQT_FACTOR));
+	u32 max_cap_rqt_size = 1 << MLX5_CAP_GEN(mdev, log_max_rqt_size);
+
+	return min_t(u32, rqt_size, max_cap_rqt_size);
+}
+
 void mlx5e_rqt_destroy(struct mlx5e_rqt *rqt)
 {
 	mlx5_core_destroy_rqt(rqt->mdev, rqt->rqtn);
@@ -151,10 +163,10 @@ int mlx5e_rqt_redirect_indir(struct mlx5e_rqt *rqt, u32 *rqns, unsigned int num_
 	u32 *rss_rqns;
 	int err;
 
-	if (WARN_ON(rqt->size != MLX5E_INDIR_RQT_SIZE))
+	if (WARN_ON(rqt->size != indir->max_table_size))
 		return -EINVAL;
 
-	rss_rqns = kvmalloc_array(MLX5E_INDIR_RQT_SIZE, sizeof(*rss_rqns), GFP_KERNEL);
+	rss_rqns = kvmalloc_array(indir->actual_table_size, sizeof(*rss_rqns), GFP_KERNEL);
 	if (!rss_rqns)
 		return -ENOMEM;
 
@@ -162,7 +174,7 @@ int mlx5e_rqt_redirect_indir(struct mlx5e_rqt *rqt, u32 *rqns, unsigned int num_
 	if (err)
 		goto out;
 
-	err = mlx5e_rqt_redirect(rqt, rss_rqns, MLX5E_INDIR_RQT_SIZE);
+	err = mlx5e_rqt_redirect(rqt, rss_rqns, indir->actual_table_size);
 
 out:
 	kvfree(rss_rqns);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.h b/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.h
index 60c985a12f24..77fba3ebd18d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/rqt.h
@@ -6,12 +6,14 @@
 
 #include <linux/kernel.h>
 
-#define MLX5E_INDIR_RQT_SIZE (1 << 8)
+#define MLX5E_INDIR_MIN_RQT_SIZE (BIT(8))
 
 struct mlx5_core_dev;
 
 struct mlx5e_rss_params_indir {
-	u32 table[MLX5E_INDIR_RQT_SIZE];
+	u32 *table;
+	u32 actual_table_size;
+	u32 max_table_size;
 };
 
 void mlx5e_rss_params_indir_init_uniform(struct mlx5e_rss_params_indir *indir,
@@ -24,7 +26,7 @@ struct mlx5e_rqt {
 };
 
 int mlx5e_rqt_init_direct(struct mlx5e_rqt *rqt, struct mlx5_core_dev *mdev,
-			  bool indir_enabled, u32 init_rqn);
+			  bool indir_enabled, u32 init_rqn, u32 indir_table_size);
 int mlx5e_rqt_init_indir(struct mlx5e_rqt *rqt, struct mlx5_core_dev *mdev,
 			 u32 *rqns, unsigned int num_rqns,
 			 u8 hfunc, struct mlx5e_rss_params_indir *indir);
@@ -35,6 +37,7 @@ static inline u32 mlx5e_rqt_get_rqtn(struct mlx5e_rqt *rqt)
 	return rqt->rqtn;
 }
 
+u32 mlx5e_rqt_size(struct mlx5_core_dev *mdev, unsigned int num_channels);
 int mlx5e_rqt_redirect_direct(struct mlx5e_rqt *rqt, u32 rqn);
 int mlx5e_rqt_redirect_indir(struct mlx5e_rqt *rqt, u32 *rqns, unsigned int num_rqns,
 			     u8 hfunc, struct mlx5e_rss_params_indir *indir);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/rss.c b/drivers/net/ethernet/mellanox/mlx5/core/en/rss.c
index 7f93426b88b3..c1545a2e8d6d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/rss.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/rss.c
@@ -81,14 +81,75 @@ struct mlx5e_rss {
 	refcount_t refcnt;
 };
 
-struct mlx5e_rss *mlx5e_rss_alloc(void)
+void mlx5e_rss_params_indir_modify_actual_size(struct mlx5e_rss *rss, u32 num_channels)
 {
-	return kvzalloc(sizeof(struct mlx5e_rss), GFP_KERNEL);
+	rss->indir.actual_table_size = mlx5e_rqt_size(rss->mdev, num_channels);
 }
 
-void mlx5e_rss_free(struct mlx5e_rss *rss)
+int mlx5e_rss_params_indir_init(struct mlx5e_rss_params_indir *indir, struct mlx5_core_dev *mdev,
+				u32 actual_table_size, u32 max_table_size)
 {
+	indir->table = kvmalloc_array(max_table_size, sizeof(*indir->table), GFP_KERNEL);
+	if (!indir->table)
+		return -ENOMEM;
+
+	indir->max_table_size = max_table_size;
+	indir->actual_table_size = actual_table_size;
+
+	return 0;
+}
+
+void mlx5e_rss_params_indir_cleanup(struct mlx5e_rss_params_indir *indir)
+{
+	kvfree(indir->table);
+}
+
+static int mlx5e_rss_copy(struct mlx5e_rss *to, const struct mlx5e_rss *from)
+{
+	u32 *dst_indir_table;
+
+	if (to->indir.actual_table_size != from->indir.actual_table_size ||
+	    to->indir.max_table_size != from->indir.max_table_size) {
+		mlx5e_rss_warn(to->mdev,
+			       "Failed to copy RSS due to size mismatch, src (actual %u, max %u) != dst (actual %u, max %u)\n",
+			       from->indir.actual_table_size, from->indir.max_table_size,
+			       to->indir.actual_table_size, to->indir.max_table_size);
+		return -EINVAL;
+	}
+
+	dst_indir_table = to->indir.table;
+	*to = *from;
+	to->indir.table = dst_indir_table;
+	memcpy(to->indir.table, from->indir.table,
+	       from->indir.actual_table_size * sizeof(*from->indir.table));
+	return 0;
+}
+
+static struct mlx5e_rss *mlx5e_rss_init_copy(const struct mlx5e_rss *from)
+{
+	struct mlx5e_rss *rss;
+	int err;
+
+	rss = kvzalloc(sizeof(*rss), GFP_KERNEL);
+	if (!rss)
+		return ERR_PTR(-ENOMEM);
+
+	err = mlx5e_rss_params_indir_init(&rss->indir, from->mdev, from->indir.actual_table_size,
+					  from->indir.max_table_size);
+	if (err)
+		goto err_free_rss;
+
+	err = mlx5e_rss_copy(rss, from);
+	if (err)
+		goto err_free_indir;
+
+	return rss;
+
+err_free_indir:
+	mlx5e_rss_params_indir_cleanup(&rss->indir);
+err_free_rss:
 	kvfree(rss);
+	return ERR_PTR(err);
 }
 
 static void mlx5e_rss_params_init(struct mlx5e_rss *rss)
@@ -282,28 +343,43 @@ static int mlx5e_rss_update_tirs(struct mlx5e_rss *rss)
 	return retval;
 }
 
-int mlx5e_rss_init_no_tirs(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
-			   bool inner_ft_support, u32 drop_rqn)
+static int mlx5e_rss_init_no_tirs(struct mlx5e_rss *rss)
 {
-	rss->mdev = mdev;
-	rss->inner_ft_support = inner_ft_support;
-	rss->drop_rqn = drop_rqn;
-
 	mlx5e_rss_params_init(rss);
 	refcount_set(&rss->refcnt, 1);
 
-	return mlx5e_rqt_init_direct(&rss->rqt, mdev, true, drop_rqn);
+	return mlx5e_rqt_init_direct(&rss->rqt, rss->mdev, true,
+				     rss->drop_rqn, rss->indir.max_table_size);
 }
 
-int mlx5e_rss_init(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
-		   bool inner_ft_support, u32 drop_rqn,
-		   const struct mlx5e_packet_merge_param *init_pkt_merge_param)
+struct mlx5e_rss *mlx5e_rss_init(struct mlx5_core_dev *mdev, bool inner_ft_support, u32 drop_rqn,
+				 const struct mlx5e_packet_merge_param *init_pkt_merge_param,
+				 enum mlx5e_rss_init_type type, unsigned int nch,
+				 unsigned int max_nch)
 {
+	struct mlx5e_rss *rss;
 	int err;
 
-	err = mlx5e_rss_init_no_tirs(rss, mdev, inner_ft_support, drop_rqn);
+	rss = kvzalloc(sizeof(*rss), GFP_KERNEL);
+	if (!rss)
+		return ERR_PTR(-ENOMEM);
+
+	err = mlx5e_rss_params_indir_init(&rss->indir, mdev,
+					  mlx5e_rqt_size(mdev, nch),
+					  mlx5e_rqt_size(mdev, max_nch));
+	if (err)
+		goto err_free_rss;
+
+	rss->mdev = mdev;
+	rss->inner_ft_support = inner_ft_support;
+	rss->drop_rqn = drop_rqn;
+
+	err = mlx5e_rss_init_no_tirs(rss);
 	if (err)
-		goto err_out;
+		goto err_free_indir;
+
+	if (type == MLX5E_RSS_INIT_NO_TIRS)
+		goto out;
 
 	err = mlx5e_rss_create_tirs(rss, init_pkt_merge_param, false);
 	if (err)
@@ -315,14 +391,18 @@ int mlx5e_rss_init(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
 			goto err_destroy_tirs;
 	}
 
-	return 0;
+out:
+	return rss;
 
 err_destroy_tirs:
 	mlx5e_rss_destroy_tirs(rss, false);
 err_destroy_rqt:
 	mlx5e_rqt_destroy(&rss->rqt);
-err_out:
-	return err;
+err_free_indir:
+	mlx5e_rss_params_indir_cleanup(&rss->indir);
+err_free_rss:
+	kvfree(rss);
+	return ERR_PTR(err);
 }
 
 int mlx5e_rss_cleanup(struct mlx5e_rss *rss)
@@ -336,6 +416,8 @@ int mlx5e_rss_cleanup(struct mlx5e_rss *rss)
 		mlx5e_rss_destroy_tirs(rss, true);
 
 	mlx5e_rqt_destroy(&rss->rqt);
+	mlx5e_rss_params_indir_cleanup(&rss->indir);
+	kvfree(rss);
 
 	return 0;
 }
@@ -470,11 +552,9 @@ inner_tir:
 
 int mlx5e_rss_get_rxfh(struct mlx5e_rss *rss, u32 *indir, u8 *key, u8 *hfunc)
 {
-	unsigned int i;
-
 	if (indir)
-		for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++)
-			indir[i] = rss->indir.table[i];
+		memcpy(indir, rss->indir.table,
+		       rss->indir.actual_table_size * sizeof(*rss->indir.table));
 
 	if (key)
 		memcpy(key, rss->hash.toeplitz_hash_key,
@@ -495,11 +575,9 @@ int mlx5e_rss_set_rxfh(struct mlx5e_rss *rss, const u32 *indir,
 	struct mlx5e_rss *old_rss;
 	int err = 0;
 
-	old_rss = mlx5e_rss_alloc();
-	if (!old_rss)
-		return -ENOMEM;
-
-	*old_rss = *rss;
+	old_rss = mlx5e_rss_init_copy(rss);
+	if (IS_ERR(old_rss))
+		return PTR_ERR(old_rss);
 
 	if (hfunc && *hfunc != rss->hash.hfunc) {
 		switch (*hfunc) {
@@ -523,18 +601,16 @@ int mlx5e_rss_set_rxfh(struct mlx5e_rss *rss, const u32 *indir,
 	}
 
 	if (indir) {
-		unsigned int i;
-
 		changed_indir = true;
 
-		for (i = 0; i < MLX5E_INDIR_RQT_SIZE; i++)
-			rss->indir.table[i] = indir[i];
+		memcpy(rss->indir.table, indir,
+		       rss->indir.actual_table_size * sizeof(*rss->indir.table));
 	}
 
 	if (changed_indir && rss->enabled) {
 		err = mlx5e_rss_apply(rss, rqns, num_rqns);
 		if (err) {
-			*rss = *old_rss;
+			mlx5e_rss_copy(rss, old_rss);
 			goto out;
 		}
 	}
@@ -543,7 +619,9 @@ int mlx5e_rss_set_rxfh(struct mlx5e_rss *rss, const u32 *indir,
 		mlx5e_rss_update_tirs(rss);
 
 out:
-	mlx5e_rss_free(old_rss);
+	mlx5e_rss_params_indir_cleanup(&old_rss->indir);
+	kvfree(old_rss);
+
 	return err;
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/rss.h b/drivers/net/ethernet/mellanox/mlx5/core/en/rss.h
index c6b216416344..d1d0bc350e92 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/rss.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/rss.h
@@ -8,18 +8,24 @@
 #include "tir.h"
 #include "fs.h"
 
+enum mlx5e_rss_init_type {
+	MLX5E_RSS_INIT_NO_TIRS = 0,
+	MLX5E_RSS_INIT_TIRS
+};
+
 struct mlx5e_rss_params_traffic_type
 mlx5e_rss_get_default_tt_config(enum mlx5_traffic_types tt);
 
 struct mlx5e_rss;
 
-struct mlx5e_rss *mlx5e_rss_alloc(void);
-void mlx5e_rss_free(struct mlx5e_rss *rss);
-int mlx5e_rss_init(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
-		   bool inner_ft_support, u32 drop_rqn,
-		   const struct mlx5e_packet_merge_param *init_pkt_merge_param);
-int mlx5e_rss_init_no_tirs(struct mlx5e_rss *rss, struct mlx5_core_dev *mdev,
-			   bool inner_ft_support, u32 drop_rqn);
+int mlx5e_rss_params_indir_init(struct mlx5e_rss_params_indir *indir, struct mlx5_core_dev *mdev,
+				u32 actual_table_size, u32 max_table_size);
+void mlx5e_rss_params_indir_cleanup(struct mlx5e_rss_params_indir *indir);
+void mlx5e_rss_params_indir_modify_actual_size(struct mlx5e_rss *rss, u32 num_channels);
+struct mlx5e_rss *mlx5e_rss_init(struct mlx5_core_dev *mdev, bool inner_ft_support, u32 drop_rqn,
+				 const struct mlx5e_packet_merge_param *init_pkt_merge_param,
+				 enum mlx5e_rss_init_type type, unsigned int nch,
+				 unsigned int max_nch);
 int mlx5e_rss_cleanup(struct mlx5e_rss *rss);
 
 void mlx5e_rss_refcnt_inc(struct mlx5e_rss *rss);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/rx_res.c b/drivers/net/ethernet/mellanox/mlx5/core/en/rx_res.c
index 56e6b8c7501f..b23e224e3763 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/rx_res.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/rx_res.c
@@ -18,7 +18,7 @@ struct mlx5e_rx_res {
 
 	struct mlx5e_rss *rss[MLX5E_MAX_NUM_RSS];
 	bool rss_active;
-	u32 rss_rqns[MLX5E_INDIR_RQT_SIZE];
+	u32 *rss_rqns;
 	unsigned int rss_nch;
 
 	struct {
@@ -34,41 +34,42 @@ struct mlx5e_rx_res {
 
 /* API for rx_res_rss_* */
 
+void mlx5e_rx_res_rss_update_num_channels(struct mlx5e_rx_res *res, u32 nch)
+{
+	int i;
+
+	for (i = 0; i < MLX5E_MAX_NUM_RSS; i++) {
+		if (res->rss[i])
+			mlx5e_rss_params_indir_modify_actual_size(res->rss[i], nch);
+	}
+}
+
 static int mlx5e_rx_res_rss_init_def(struct mlx5e_rx_res *res,
 				     unsigned int init_nch)
 {
 	bool inner_ft_support = res->features & MLX5E_RX_RES_FEATURE_INNER_FT;
 	struct mlx5e_rss *rss;
-	int err;
 
 	if (WARN_ON(res->rss[0]))
 		return -EINVAL;
 
-	rss = mlx5e_rss_alloc();
-	if (!rss)
-		return -ENOMEM;
-
-	err = mlx5e_rss_init(rss, res->mdev, inner_ft_support, res->drop_rqn,
-			     &res->pkt_merge_param);
-	if (err)
-		goto err_rss_free;
+	rss = mlx5e_rss_init(res->mdev, inner_ft_support, res->drop_rqn,
+			     &res->pkt_merge_param, MLX5E_RSS_INIT_TIRS, init_nch, res->max_nch);
+	if (IS_ERR(rss))
+		return PTR_ERR(rss);
 
 	mlx5e_rss_set_indir_uniform(rss, init_nch);
 
 	res->rss[0] = rss;
 
 	return 0;
-
-err_rss_free:
-	mlx5e_rss_free(rss);
-	return err;
 }
 
 int mlx5e_rx_res_rss_init(struct mlx5e_rx_res *res, u32 *rss_idx, unsigned int init_nch)
 {
 	bool inner_ft_support = res->features & MLX5E_RX_RES_FEATURE_INNER_FT;
 	struct mlx5e_rss *rss;
-	int err,