[PATCH net-next 0/9] net: ethtool: add dedicated RXFH driver callbacks

[PATCH net-next 0/9] net: ethtool: add dedicated RXFH driver callbacks

[Jakub Kicinski]

Andrew asked me to plumb the RXFH header fields configuration
thru to netlink. Before we do that we need to clean up the driver
facing API a little bit. Right now RXFH configuration shares the
callbacks with n-tuple filters. The future of n-tuple filters
is uncertain within netlink. Separate the two for clarity both
of the core code and the driver facing API.

This series adds the new callbacks and converts the initial
handful of drivers. There is 31 more driver patches to come,
then we can stop calling rxnfc in the core for rxfh.

Jakub Kicinski (9):
  net: ethtool: copy the rxfh flow handling
  net: ethtool: remove the duplicated handling from rxfh and rxnfc
  net: ethtool: require drivers to opt into the per-RSS ctx RXFH
  net: ethtool: add dedicated callbacks for getting and setting rxfh
    fields
  eth: remove empty RXFH handling from drivers
  eth: fbnic: migrate to new RXFH callbacks
  net: drv: vmxnet3: migrate to new RXFH callbacks
  net: drv: virtio: migrate to new RXFH callbacks
  net: drv: hyperv: migrate to new RXFH callbacks

 include/linux/ethtool.h                       |  23 ++++
 drivers/net/ethernet/google/gve/gve_ethtool.c |   6 -
 drivers/net/ethernet/marvell/mvneta.c         |   2 -
 .../ethernet/mellanox/mlx5/core/en_ethtool.c  |   1 +
 .../net/ethernet/meta/fbnic/fbnic_ethtool.c   | 111 ++++++++--------
 drivers/net/ethernet/sfc/ethtool.c            |   1 +
 drivers/net/hyperv/netvsc_drv.c               |  30 ++---
 drivers/net/virtio_net.c                      |  47 +++----
 drivers/net/vmxnet3/vmxnet3_ethtool.c         |  74 ++++-------
 net/ethtool/ioctl.c                           | 121 +++++++++++++++---
 10 files changed, 239 insertions(+), 177 deletions(-)

-- 
2.49.0

[PATCH net-next 1/9] net: ethtool: copy the rxfh flow handling

[Jakub Kicinski]

RX Flow Hash configuration uses the same argument structure
as flow filters. This is probably why ethtool IOCTL handles
them together. The more checks we add the more convoluted
this code is getting (as some of the checks apply only
to flow filters and others only to the hashing).

Copy the code to separate the handling. This is an exact
copy, the next change will remove unnecessary handling.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: andrew@lunn.ch
CC: ecree.xilinx@gmail.com
---
 net/ethtool/ioctl.c | 93 ++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 92 insertions(+), 1 deletion(-)

diff --git a/net/ethtool/ioctl.c b/net/ethtool/ioctl.c
index 39ec920f5de7..5c15eff53c80 100644
--- a/net/ethtool/ioctl.c
+++ b/net/ethtool/ioctl.c
@@ -1060,6 +1060,93 @@ static int ethtool_check_flow_types(struct net_device *dev, u32 input_xfrm)
 	return 0;
 }
 
+static noinline_for_stack int
+ethtool_set_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
+{
+	const struct ethtool_ops *ops = dev->ethtool_ops;
+	struct ethtool_rxnfc info;
+	size_t info_size = sizeof(info);
+	int rc;
+
+	if (!ops->set_rxnfc)
+		return -EOPNOTSUPP;
+
+	rc = ethtool_rxnfc_copy_struct(cmd, &info, &info_size, useraddr);
+	if (rc)
+		return rc;
+
+	if (cmd == ETHTOOL_SRXCLSRLINS && info.fs.flow_type & FLOW_RSS) {
+		/* Nonzero ring with RSS only makes sense
+		 * if NIC adds them together
+		 */
+		if (!ops->cap_rss_rxnfc_adds &&
+		    ethtool_get_flow_spec_ring(info.fs.ring_cookie))
+			return -EINVAL;
+
+		if (!xa_load(&dev->ethtool->rss_ctx, info.rss_context))
+			return -EINVAL;
+	}
+
+	if (cmd == ETHTOOL_SRXFH && ops->get_rxfh) {
+		struct ethtool_rxfh_param rxfh = {};
+
+		rc = ops->get_rxfh(dev, &rxfh);
+		if (rc)
+			return rc;
+
+		rc = ethtool_check_xfrm_rxfh(rxfh.input_xfrm, info.data);
+		if (rc)
+			return rc;
+	}
+
+	rc = ops->set_rxnfc(dev, &info);
+	if (rc)
+		return rc;
+
+	if (cmd == ETHTOOL_SRXCLSRLINS &&
+	    ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, NULL))
+		return -EFAULT;
+
+	return 0;
+}
+
+static noinline_for_stack int
+ethtool_get_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
+{
+	struct ethtool_rxnfc info;
+	size_t info_size = sizeof(info);
+	const struct ethtool_ops *ops = dev->ethtool_ops;
+	int ret;
+	void *rule_buf = NULL;
+
+	if (!ops->get_rxnfc)
+		return -EOPNOTSUPP;
+
+	ret = ethtool_rxnfc_copy_struct(cmd, &info, &info_size, useraddr);
+	if (ret)
+		return ret;
+
+	if (info.cmd == ETHTOOL_GRXCLSRLALL) {
+		if (info.rule_cnt > 0) {
+			if (info.rule_cnt <= KMALLOC_MAX_SIZE / sizeof(u32))
+				rule_buf = kcalloc(info.rule_cnt, sizeof(u32),
+						   GFP_USER);
+			if (!rule_buf)
+				return -ENOMEM;
+		}
+	}
+
+	ret = ops->get_rxnfc(dev, &info, rule_buf);
+	if (ret < 0)
+		goto err_out;
+
+	ret = ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, rule_buf);
+err_out:
+	kfree(rule_buf);
+
+	return ret;
+}
+
 static noinline_for_stack int ethtool_set_rxnfc(struct net_device *dev,
 						u32 cmd, void __user *useraddr)
 {
@@ -3338,13 +3425,17 @@ __dev_ethtool(struct net *net, struct ifreq *ifr, void __user *useraddr,
 				       dev->ethtool_ops->set_priv_flags);
 		break;
 	case ETHTOOL_GRXFH:
+		rc = ethtool_get_rxfh_fields(dev, ethcmd, useraddr);
+		break;
+	case ETHTOOL_SRXFH:
+		rc = ethtool_set_rxfh_fields(dev, ethcmd, useraddr);
+		break;
 	case ETHTOOL_GRXRINGS:
 	case ETHTOOL_GRXCLSRLCNT:
 	case ETHTOOL_GRXCLSRULE:
 	case ETHTOOL_GRXCLSRLALL:
 		rc = ethtool_get_rxnfc(dev, ethcmd, useraddr);
 		break;
-	case ETHTOOL_SRXFH:
 	case ETHTOOL_SRXCLSRLDEL:
 	case ETHTOOL_SRXCLSRLINS:
 		rc = ethtool_set_rxnfc(dev, ethcmd, useraddr);
-- 
2.49.0

Re: [PATCH net-next 1/9] net: ethtool: copy the rxfh flow handling

[Joe Damato]

On Wed, Jun 11, 2025 at 07:59:41AM -0700, Jakub Kicinski wrote:
> RX Flow Hash configuration uses the same argument structure
> as flow filters. This is probably why ethtool IOCTL handles
> them together. The more checks we add the more convoluted
> this code is getting (as some of the checks apply only
> to flow filters and others only to the hashing).
> 
> Copy the code to separate the handling. This is an exact
> copy, the next change will remove unnecessary handling.
> 
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: andrew@lunn.ch
> CC: ecree.xilinx@gmail.com
> ---
>  net/ethtool/ioctl.c | 93 ++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 92 insertions(+), 1 deletion(-)
>

Reviewed-by: Joe Damato <joe@dama.to>

[PATCH net-next 2/9] net: ethtool: remove the duplicated handling from rxfh and rxnfc

[Jakub Kicinski]

Now that the handles have been separated - remove the RX Flow Hash
handling from rxnfc functions and vice versa.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: andrew@lunn.ch
CC: ecree.xilinx@gmail.com
---
 net/ethtool/ioctl.c | 57 ++++-----------------------------------------
 1 file changed, 5 insertions(+), 52 deletions(-)

diff --git a/net/ethtool/ioctl.c b/net/ethtool/ioctl.c
index 5c15eff53c80..330ca99800ce 100644
--- a/net/ethtool/ioctl.c
+++ b/net/ethtool/ioctl.c
@@ -1075,19 +1075,7 @@ ethtool_set_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 	if (rc)
 		return rc;
 
-	if (cmd == ETHTOOL_SRXCLSRLINS && info.fs.flow_type & FLOW_RSS) {
-		/* Nonzero ring with RSS only makes sense
-		 * if NIC adds them together
-		 */
-		if (!ops->cap_rss_rxnfc_adds &&
-		    ethtool_get_flow_spec_ring(info.fs.ring_cookie))
-			return -EINVAL;
-
-		if (!xa_load(&dev->ethtool->rss_ctx, info.rss_context))
-			return -EINVAL;
-	}
-
-	if (cmd == ETHTOOL_SRXFH && ops->get_rxfh) {
+	if (ops->get_rxfh) {
 		struct ethtool_rxfh_param rxfh = {};
 
 		rc = ops->get_rxfh(dev, &rxfh);
@@ -1099,15 +1087,7 @@ ethtool_set_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 			return rc;
 	}
 
-	rc = ops->set_rxnfc(dev, &info);
-	if (rc)
-		return rc;
-
-	if (cmd == ETHTOOL_SRXCLSRLINS &&
-	    ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, NULL))
-		return -EFAULT;
-
-	return 0;
+	return ops->set_rxnfc(dev, &info);
 }
 
 static noinline_for_stack int
@@ -1117,7 +1097,6 @@ ethtool_get_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 	size_t info_size = sizeof(info);
 	const struct ethtool_ops *ops = dev->ethtool_ops;
 	int ret;
-	void *rule_buf = NULL;
 
 	if (!ops->get_rxnfc)
 		return -EOPNOTSUPP;
@@ -1126,25 +1105,11 @@ ethtool_get_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 	if (ret)
 		return ret;
 
-	if (info.cmd == ETHTOOL_GRXCLSRLALL) {
-		if (info.rule_cnt > 0) {
-			if (info.rule_cnt <= KMALLOC_MAX_SIZE / sizeof(u32))
-				rule_buf = kcalloc(info.rule_cnt, sizeof(u32),
-						   GFP_USER);
-			if (!rule_buf)
-				return -ENOMEM;
-		}
-	}
-
-	ret = ops->get_rxnfc(dev, &info, rule_buf);
+	ret = ops->get_rxnfc(dev, &info, NULL);
 	if (ret < 0)
-		goto err_out;
+		return ret;
 
-	ret = ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, rule_buf);
-err_out:
-	kfree(rule_buf);
-
-	return ret;
+	return ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, NULL);
 }
 
 static noinline_for_stack int ethtool_set_rxnfc(struct net_device *dev,
@@ -1174,18 +1139,6 @@ static noinline_for_stack int ethtool_set_rxnfc(struct net_device *dev,
 			return -EINVAL;
 	}
 
-	if (cmd == ETHTOOL_SRXFH && ops->get_rxfh) {
-		struct ethtool_rxfh_param rxfh = {};
-
-		rc = ops->get_rxfh(dev, &rxfh);
-		if (rc)
-			return rc;
-
-		rc = ethtool_check_xfrm_rxfh(rxfh.input_xfrm, info.data);
-		if (rc)
-			return rc;
-	}
-
 	rc = ops->set_rxnfc(dev, &info);
 	if (rc)
 		return rc;
-- 
2.49.0

Re: [PATCH net-next 2/9] net: ethtool: remove the duplicated handling from rxfh and rxnfc

[Joe Damato]

On Wed, Jun 11, 2025 at 07:59:42AM -0700, Jakub Kicinski wrote:
> Now that the handles have been separated - remove the RX Flow Hash
> handling from rxnfc functions and vice versa.
> 
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: andrew@lunn.ch
> CC: ecree.xilinx@gmail.com
> ---
>  net/ethtool/ioctl.c | 57 ++++-----------------------------------------
>  1 file changed, 5 insertions(+), 52 deletions(-)

Reviewed-by: Joe Damato <joe@dama.to>

[PATCH net-next 3/9] net: ethtool: require drivers to opt into the per-RSS ctx RXFH

[Jakub Kicinski]

RX Flow Hashing supports using different configuration for different
RSS contexts. Only two drivers seem to support it. Make sure we
uniformly error out for drivers which don't.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: saeedm@nvidia.com
CC: tariqt@nvidia.com
CC: leon@kernel.org
CC: ecree.xilinx@gmail.com
CC: linux-rdma@vger.kernel.org
CC: linux-net-drivers@amd.com
---
 include/linux/ethtool.h                              | 3 +++
 drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c | 1 +
 drivers/net/ethernet/sfc/ethtool.c                   | 1 +
 net/ethtool/ioctl.c                                  | 8 ++++++++
 4 files changed, 13 insertions(+)

diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
index 5e0dd333ad1f..fc1c2379e7ff 100644
--- a/include/linux/ethtool.h
+++ b/include/linux/ethtool.h
@@ -855,6 +855,8 @@ struct kernel_ethtool_ts_info {
  * @cap_rss_ctx_supported: indicates if the driver supports RSS
  *	contexts via legacy API, drivers implementing @create_rxfh_context
  *	do not have to set this bit.
+ * @rxfh_per_ctx_fields: device supports selecting different header fields
+ *	for Rx hash calculation and RSS for each additional context.
  * @rxfh_per_ctx_key: device supports setting different RSS key for each
  *	additional context. Netlink API should report hfunc, key, and input_xfrm
  *	for every context, not just context 0.
@@ -1084,6 +1086,7 @@ struct ethtool_ops {
 	u32     supported_input_xfrm:8;
 	u32     cap_link_lanes_supported:1;
 	u32     cap_rss_ctx_supported:1;
+	u32	rxfh_per_ctx_fields:1;
 	u32	rxfh_per_ctx_key:1;
 	u32	cap_rss_rxnfc_adds:1;
 	u32	rxfh_indir_space;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
index ea078c9f5d15..90c760057bb6 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
@@ -2619,6 +2619,7 @@ static void mlx5e_get_ts_stats(struct net_device *netdev,
 const struct ethtool_ops mlx5e_ethtool_ops = {
 	.cap_link_lanes_supported = true,
 	.cap_rss_ctx_supported	= true,
+	.rxfh_per_ctx_fields	= true,
 	.rxfh_per_ctx_key	= true,
 	.supported_coalesce_params = ETHTOOL_COALESCE_USECS |
 				     ETHTOOL_COALESCE_MAX_FRAMES |
diff --git a/drivers/net/ethernet/sfc/ethtool.c b/drivers/net/ethernet/sfc/ethtool.c
index 83d715544f7f..afbedca63b29 100644
--- a/drivers/net/ethernet/sfc/ethtool.c
+++ b/drivers/net/ethernet/sfc/ethtool.c
@@ -262,6 +262,7 @@ const struct ethtool_ops efx_ethtool_ops = {
 	.set_rxnfc		= efx_ethtool_set_rxnfc,
 	.get_rxfh_indir_size	= efx_ethtool_get_rxfh_indir_size,
 	.get_rxfh_key_size	= efx_ethtool_get_rxfh_key_size,
+	.rxfh_per_ctx_fields	= true,
 	.rxfh_per_ctx_key	= true,
 	.cap_rss_rxnfc_adds	= true,
 	.rxfh_priv_size		= sizeof(struct efx_rss_context_priv),
diff --git a/net/ethtool/ioctl.c b/net/ethtool/ioctl.c
index 330ca99800ce..bd9fd95bb82f 100644
--- a/net/ethtool/ioctl.c
+++ b/net/ethtool/ioctl.c
@@ -1075,6 +1075,10 @@ ethtool_set_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 	if (rc)
 		return rc;
 
+	if (info.flow_type & FLOW_RSS && info.rss_context &&
+	    !ops->rxfh_per_ctx_fields)
+		return -EINVAL;
+
 	if (ops->get_rxfh) {
 		struct ethtool_rxfh_param rxfh = {};
 
@@ -1105,6 +1109,10 @@ ethtool_get_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 	if (ret)
 		return ret;
 
+	if (info.flow_type & FLOW_RSS && info.rss_context &&
+	    !ops->rxfh_per_ctx_fields)
+		return -EINVAL;
+
 	ret = ops->get_rxnfc(dev, &info, NULL);
 	if (ret < 0)
 		return ret;
-- 
2.49.0

Re: [PATCH net-next 3/9] net: ethtool: require drivers to opt into the per-RSS ctx RXFH

[Joe Damato]

On Wed, Jun 11, 2025 at 07:59:43AM -0700, Jakub Kicinski wrote:
> RX Flow Hashing supports using different configuration for different
> RSS contexts. Only two drivers seem to support it. Make sure we
> uniformly error out for drivers which don't.
> 
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: saeedm@nvidia.com
> CC: tariqt@nvidia.com
> CC: leon@kernel.org
> CC: ecree.xilinx@gmail.com
> CC: linux-rdma@vger.kernel.org
> CC: linux-net-drivers@amd.com
> ---
>  include/linux/ethtool.h                              | 3 +++
>  drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c | 1 +
>  drivers/net/ethernet/sfc/ethtool.c                   | 1 +
>  net/ethtool/ioctl.c                                  | 8 ++++++++
>  4 files changed, 13 insertions(+)

Reviewed-by: Joe Damato <joe@dama.to>

[PATCH net-next 4/9] net: ethtool: add dedicated callbacks for getting and setting rxfh fields

[Jakub Kicinski]

We mux multiple calls to the drivers via the .get_nfc and .set_nfc
callbacks. This is slightly inconvenient to the drivers as they
have to de-mux them back. It will also be awkward for netlink code
to construct struct ethtool_rxnfc when it wants to get info about
RX Flow Hash, from the RSS module.

Add dedicated driver callbacks. Create struct ethtool_rxfh_fields
which contains only data relevant to RXFH. Maintain the names of
the fields to avoid having to heavily modify the drivers.

For now support both callbacks, once all drivers are converted
ethtool_*et_rxfh_fields() will stop using the rxnfc callbacks.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: andrew@lunn.ch
CC: ecree.xilinx@gmail.com
---
 include/linux/ethtool.h | 20 +++++++++++++++
 net/ethtool/ioctl.c     | 55 +++++++++++++++++++++++++++++++++--------
 2 files changed, 65 insertions(+), 10 deletions(-)

diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
index fc1c2379e7ff..b2e71e641f62 100644
--- a/include/linux/ethtool.h
+++ b/include/linux/ethtool.h
@@ -825,6 +825,19 @@ struct ethtool_rxfh_param {
 	u8	input_xfrm;
 };
 
+/**
+ * struct ethtool_rxfh_fields - Rx Flow Hashing (RXFH) header field config
+ * @data: which header fields are used for hashing, bitmask of RXH_* defines
+ * @flow_type: L2-L4 network traffic flow type
+ * @rss_context: RSS context, will only be used if rxfh_per_ctx_fields is
+ *	set in struct ethtool_ops
+ */
+struct ethtool_rxfh_fields {
+	u32 data;
+	u32 flow_type;
+	u32 rss_context;
+};
+
 /**
  * struct kernel_ethtool_ts_info - kernel copy of struct ethtool_ts_info
  * @cmd: command number = %ETHTOOL_GET_TS_INFO
@@ -970,6 +983,8 @@ struct kernel_ethtool_ts_info {
  *	will remain unchanged.
  *	Returns a negative error code or zero. An error code must be returned
  *	if at least one unsupported change was requested.
+ * @get_rxfh_fields: Get header fields used for flow hashing.
+ * @set_rxfh_fields: Set header fields used for flow hashing.
  * @create_rxfh_context: Create a new RSS context with the specified RX flow
  *	hash indirection table, hash key, and hash function.
  *	The &struct ethtool_rxfh_context for this context is passed in @ctx;
@@ -1156,6 +1171,11 @@ struct ethtool_ops {
 	int	(*get_rxfh)(struct net_device *, struct ethtool_rxfh_param *);
 	int	(*set_rxfh)(struct net_device *, struct ethtool_rxfh_param *,
 			    struct netlink_ext_ack *extack);
+	int	(*get_rxfh_fields)(struct net_device *,
+				   struct ethtool_rxfh_fields *);
+	int	(*set_rxfh_fields)(struct net_device *,
+				   const struct ethtool_rxfh_fields *,
+				   struct netlink_ext_ack *extack);
 	int	(*create_rxfh_context)(struct net_device *,
 				       struct ethtool_rxfh_context *ctx,
 				       const struct ethtool_rxfh_param *rxfh,
diff --git a/net/ethtool/ioctl.c b/net/ethtool/ioctl.c
index bd9fd95bb82f..f4d4d60275f8 100644
--- a/net/ethtool/ioctl.c
+++ b/net/ethtool/ioctl.c
@@ -1048,9 +1048,20 @@ static int ethtool_check_flow_types(struct net_device *dev, u32 input_xfrm)
 			continue;
 
 		info.flow_type = i;
-		err = ops->get_rxnfc(dev, &info, NULL);
-		if (err)
-			continue;
+
+		if (ops->get_rxfh_fields) {
+			struct ethtool_rxfh_fields fields = {
+				.flow_type	= info.flow_type,
+			};
+
+			if (ops->get_rxfh_fields(dev, &fields))
+				continue;
+
+			info.data = fields.data;
+		} else {
+			if (ops->get_rxnfc(dev, &info, NULL))
+				continue;
+		}
 
 		err = ethtool_check_xfrm_rxfh(input_xfrm, info.data);
 		if (err)
@@ -1064,11 +1075,12 @@ static noinline_for_stack int
 ethtool_set_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 {
 	const struct ethtool_ops *ops = dev->ethtool_ops;
+	struct ethtool_rxfh_fields fields = {};
 	struct ethtool_rxnfc info;
 	size_t info_size = sizeof(info);
 	int rc;
 
-	if (!ops->set_rxnfc)
+	if (!ops->set_rxnfc && !ops->set_rxfh_fields)
 		return -EOPNOTSUPP;
 
 	rc = ethtool_rxnfc_copy_struct(cmd, &info, &info_size, useraddr);
@@ -1091,7 +1103,15 @@ ethtool_set_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 			return rc;
 	}
 
-	return ops->set_rxnfc(dev, &info);
+	if (!ops->set_rxfh_fields)
+		return ops->set_rxnfc(dev, &info);
+
+	fields.data = info.data;
+	fields.flow_type = info.flow_type & ~FLOW_RSS;
+	if (info.flow_type & FLOW_RSS)
+		fields.rss_context = info.rss_context;
+
+	return ops->set_rxfh_fields(dev, &fields, NULL);
 }
 
 static noinline_for_stack int
@@ -1102,7 +1122,7 @@ ethtool_get_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 	const struct ethtool_ops *ops = dev->ethtool_ops;
 	int ret;
 
-	if (!ops->get_rxnfc)
+	if (!ops->get_rxnfc && !ops->get_rxfh_fields)
 		return -EOPNOTSUPP;
 
 	ret = ethtool_rxnfc_copy_struct(cmd, &info, &info_size, useraddr);
@@ -1113,9 +1133,24 @@ ethtool_get_rxfh_fields(struct net_device *dev, u32 cmd, void __user *useraddr)
 	    !ops->rxfh_per_ctx_fields)
 		return -EINVAL;
 
-	ret = ops->get_rxnfc(dev, &info, NULL);
-	if (ret < 0)
-		return ret;
+	if (ops->get_rxfh_fields) {
+		struct ethtool_rxfh_fields fields = {
+			.flow_type	= info.flow_type & ~FLOW_RSS,
+		};
+
+		if (info.flow_type & FLOW_RSS)
+			fields.rss_context = info.rss_context;
+
+		ret = ops->get_rxfh_fields(dev, &fields);
+		if (ret < 0)
+			return ret;
+
+		info.data = fields.data;
+	} else {
+		ret = ops->get_rxnfc(dev, &info, NULL);
+		if (ret < 0)
+			return ret;
+	}
 
 	return ethtool_rxnfc_copy_to_user(useraddr, &info, info_size, NULL);
 }
@@ -1492,7 +1527,7 @@ static noinline_for_stack int ethtool_set_rxfh(struct net_device *dev,
 	u8 *rss_config;
 	int ret;
 
-	if (!ops->get_rxnfc || !ops->set_rxfh)
+	if ((!ops->get_rxnfc && !ops->get_rxfh_fields) || !ops->set_rxfh)
 		return -EOPNOTSUPP;
 
 	if (ops->get_rxfh_indir_size)
-- 
2.49.0

Re: [PATCH net-next 4/9] net: ethtool: add dedicated callbacks for getting and setting rxfh fields

[Joe Damato]

On Wed, Jun 11, 2025 at 07:59:44AM -0700, Jakub Kicinski wrote:
> We mux multiple calls to the drivers via the .get_nfc and .set_nfc
> callbacks. This is slightly inconvenient to the drivers as they
> have to de-mux them back. It will also be awkward for netlink code
> to construct struct ethtool_rxnfc when it wants to get info about
> RX Flow Hash, from the RSS module.
> 
> Add dedicated driver callbacks. Create struct ethtool_rxfh_fields
> which contains only data relevant to RXFH. Maintain the names of
> the fields to avoid having to heavily modify the drivers.
> 
> For now support both callbacks, once all drivers are converted
> ethtool_*et_rxfh_fields() will stop using the rxnfc callbacks.
> 
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: andrew@lunn.ch
> CC: ecree.xilinx@gmail.com
> ---

[...]

> diff --git a/net/ethtool/ioctl.c b/net/ethtool/ioctl.c
> index bd9fd95bb82f..f4d4d60275f8 100644
> --- a/net/ethtool/ioctl.c
> +++ b/net/ethtool/ioctl.c

[...]

> @@ -1492,7 +1527,7 @@ static noinline_for_stack int ethtool_set_rxfh(struct net_device *dev,
>  	u8 *rss_config;
>  	int ret;
>  
> -	if (!ops->get_rxnfc || !ops->set_rxfh)
> +	if ((!ops->get_rxnfc && !ops->get_rxfh_fields) || !ops->set_rxfh)
>  		return -EOPNOTSUPP;

I realize I am late to the thread, but is this part above correct? It seems
like ethtool_set_rxfh calls ops->get_rxnfc but not ops->get_rxfh_fields,
unless I missed something in an earlier patch?

Re: [PATCH net-next 4/9] net: ethtool: add dedicated callbacks for getting and setting rxfh fields

[Jakub Kicinski]

On Fri, 13 Jun 2025 08:36:15 +0300 Joe Damato wrote:
> > @@ -1492,7 +1527,7 @@ static noinline_for_stack int ethtool_set_rxfh(struct net_device *dev,
> >  	u8 *rss_config;
> >  	int ret;
> >  
> > -	if (!ops->get_rxnfc || !ops->set_rxfh)
> > +	if ((!ops->get_rxnfc && !ops->get_rxfh_fields) || !ops->set_rxfh)
> >  		return -EOPNOTSUPP;  
> 
> I realize I am late to the thread, but is this part above correct? It seems
> like ethtool_set_rxfh calls ops->get_rxnfc but not ops->get_rxfh_fields,
> unless I missed something in an earlier patch?

It calls ethtool_check_flow_types(), but you're right it also calls
->get_rxnfc, so it should have been an ||. I'll correct in the "closing
patch" that removes the calls to get_rxnfc once all drivers are done.
Thanks!

[PATCH net-next 5/9] eth: remove empty RXFH handling from drivers

[Jakub Kicinski]

We're migrating RXFH config to new callbacks.
Remove RXFH handling from drivers where it does nothing.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: jeroendb@google.com
CC: hramamurthy@google.com
CC: marcin.s.wojtas@gmail.com
CC: willemb@google.com
CC: pkaligineedi@google.com
CC: joshwash@google.com
CC: ziweixiao@google.com
---
 drivers/net/ethernet/google/gve/gve_ethtool.c | 6 ------
 drivers/net/ethernet/marvell/mvneta.c         | 2 --
 2 files changed, 8 deletions(-)

diff --git a/drivers/net/ethernet/google/gve/gve_ethtool.c b/drivers/net/ethernet/google/gve/gve_ethtool.c
index 3c1da0cf3f61..a6d0089ecd7b 100644
--- a/drivers/net/ethernet/google/gve/gve_ethtool.c
+++ b/drivers/net/ethernet/google/gve/gve_ethtool.c
@@ -798,9 +798,6 @@ static int gve_set_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *cmd)
 	case ETHTOOL_SRXCLSRLDEL:
 		err = gve_del_flow_rule(priv, cmd);
 		break;
-	case ETHTOOL_SRXFH:
-		err = -EOPNOTSUPP;
-		break;
 	default:
 		err = -EOPNOTSUPP;
 		break;
@@ -835,9 +832,6 @@ static int gve_get_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *cmd, u
 	case ETHTOOL_GRXCLSRLALL:
 		err = gve_get_flow_rule_ids(priv, cmd, (u32 *)rule_locs);
 		break;
-	case ETHTOOL_GRXFH:
-		err = -EOPNOTSUPP;
-		break;
 	default:
 		err = -EOPNOTSUPP;
 		break;
diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
index 147571fdada3..feab392ab2ee 100644
--- a/drivers/net/ethernet/marvell/mvneta.c
+++ b/drivers/net/ethernet/marvell/mvneta.c
@@ -5014,8 +5014,6 @@ static int mvneta_ethtool_get_rxnfc(struct net_device *dev,
 	case ETHTOOL_GRXRINGS:
 		info->data =  rxq_number;
 		return 0;
-	case ETHTOOL_GRXFH:
-		return -EOPNOTSUPP;
 	default:
 		return -EOPNOTSUPP;
 	}
-- 
2.49.0

Re: [PATCH net-next 5/9] eth: remove empty RXFH handling from drivers

[Ziwei Xiao]

On Wed, Jun 11, 2025 at 7:59 AM Jakub Kicinski <kuba@kernel.org> wrote:
>
> We're migrating RXFH config to new callbacks.
> Remove RXFH handling from drivers where it does nothing.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: jeroendb@google.com
> CC: hramamurthy@google.com
> CC: marcin.s.wojtas@gmail.com
> CC: willemb@google.com
> CC: pkaligineedi@google.com
> CC: joshwash@google.com
> CC: ziweixiao@google.com
> ---

Reviewed-by: Ziwei Xiao <ziweixiao@google.com>

[PATCH net-next 6/9] eth: fbnic: migrate to new RXFH callbacks

[Jakub Kicinski]

Add support for the new rxfh_fields callbacks, instead of de-muxing
the rxnfc calls. The code is moved as we try to declare the functions
in the order ing which they appear in the ops struct.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: alexanderduyck@fb.com
CC: kernel-team@meta.com
CC: mohsin.bashr@gmail.com
---
 .../net/ethernet/meta/fbnic/fbnic_ethtool.c   | 111 +++++++++---------
 1 file changed, 56 insertions(+), 55 deletions(-)

diff --git a/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c b/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
index 5c7556c8c4c5..1439d0c88fa6 100644
--- a/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
+++ b/drivers/net/ethernet/meta/fbnic/fbnic_ethtool.c
@@ -531,20 +531,6 @@ static int fbnic_get_rss_hash_idx(u32 flow_type)
 	return -1;
 }
 
-static int
-fbnic_get_rss_hash_opts(struct fbnic_net *fbn, struct ethtool_rxnfc *cmd)
-{
-	int hash_opt_idx = fbnic_get_rss_hash_idx(cmd->flow_type);
-
-	if (hash_opt_idx < 0)
-		return -EINVAL;
-
-	/* Report options from rss_en table in fbn */
-	cmd->data = fbn->rss_flow_hash[hash_opt_idx];
-
-	return 0;
-}
-
 static int fbnic_get_cls_rule_all(struct fbnic_net *fbn,
 				  struct ethtool_rxnfc *cmd,
 				  u32 *rule_locs)
@@ -779,9 +765,6 @@ static int fbnic_get_rxnfc(struct net_device *netdev,
 		cmd->data = fbn->num_rx_queues;
 		ret = 0;
 		break;
-	case ETHTOOL_GRXFH:
-		ret = fbnic_get_rss_hash_opts(fbn, cmd);
-		break;
 	case ETHTOOL_GRXCLSRULE:
 		ret = fbnic_get_cls_rule(fbn, cmd);
 		break;
@@ -803,41 +786,6 @@ static int fbnic_get_rxnfc(struct net_device *netdev,
 	return ret;
 }
 
-#define FBNIC_L2_HASH_OPTIONS \
-	(RXH_L2DA | RXH_DISCARD)
-#define FBNIC_L3_HASH_OPTIONS \
-	(FBNIC_L2_HASH_OPTIONS | RXH_IP_SRC | RXH_IP_DST)
-#define FBNIC_L4_HASH_OPTIONS \
-	(FBNIC_L3_HASH_OPTIONS | RXH_L4_B_0_1 | RXH_L4_B_2_3)
-
-static int
-fbnic_set_rss_hash_opts(struct fbnic_net *fbn, const struct ethtool_rxnfc *cmd)
-{
-	int hash_opt_idx;
-
-	/* Verify the type requested is correct */
-	hash_opt_idx = fbnic_get_rss_hash_idx(cmd->flow_type);
-	if (hash_opt_idx < 0)
-		return -EINVAL;
-
-	/* Verify the fields asked for can actually be assigned based on type */
-	if (cmd->data & ~FBNIC_L4_HASH_OPTIONS ||
-	    (hash_opt_idx > FBNIC_L4_HASH_OPT &&
-	     cmd->data & ~FBNIC_L3_HASH_OPTIONS) ||
-	    (hash_opt_idx > FBNIC_IP_HASH_OPT &&
-	     cmd->data & ~FBNIC_L2_HASH_OPTIONS))
-		return -EINVAL;
-
-	fbn->rss_flow_hash[hash_opt_idx] = cmd->data;
-
-	if (netif_running(fbn->netdev)) {
-		fbnic_rss_reinit(fbn->fbd, fbn);
-		fbnic_write_rules(fbn->fbd);
-	}
-
-	return 0;
-}
-
 static int fbnic_cls_rule_any_loc(struct fbnic_dev *fbd)
 {
 	int i;
@@ -1244,9 +1192,6 @@ static int fbnic_set_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *cmd)
 	int ret = -EOPNOTSUPP;
 
 	switch (cmd->cmd) {
-	case ETHTOOL_SRXFH:
-		ret = fbnic_set_rss_hash_opts(fbn, cmd);
-		break;
 	case ETHTOOL_SRXCLSRLINS:
 		ret = fbnic_set_cls_rule_ins(fbn, cmd);
 		break;
@@ -1346,6 +1291,60 @@ fbnic_set_rxfh(struct net_device *netdev, struct ethtool_rxfh_param *rxfh,
 	return 0;
 }
 
+static int
+fbnic_get_rss_hash_opts(struct net_device *netdev,
+			struct ethtool_rxfh_fields *cmd)
+{
+	int hash_opt_idx = fbnic_get_rss_hash_idx(cmd->flow_type);
+	struct fbnic_net *fbn = netdev_priv(netdev);
+
+	if (hash_opt_idx < 0)
+		return -EINVAL;
+
+	/* Report options from rss_en table in fbn */
+	cmd->data = fbn->rss_flow_hash[hash_opt_idx];
+
+	return 0;
+}
+
+#define FBNIC_L2_HASH_OPTIONS \
+	(RXH_L2DA | RXH_DISCARD)
+#define FBNIC_L3_HASH_OPTIONS \
+	(FBNIC_L2_HASH_OPTIONS | RXH_IP_SRC | RXH_IP_DST)
+#define FBNIC_L4_HASH_OPTIONS \
+	(FBNIC_L3_HASH_OPTIONS | RXH_L4_B_0_1 | RXH_L4_B_2_3)
+
+static int
+fbnic_set_rss_hash_opts(struct net_device *netdev,
+			const struct ethtool_rxfh_fields *cmd,
+			struct netlink_ext_ack *extack)
+{
+	struct fbnic_net *fbn = netdev_priv(netdev);
+	int hash_opt_idx;
+
+	/* Verify the type requested is correct */
+	hash_opt_idx = fbnic_get_rss_hash_idx(cmd->flow_type);
+	if (hash_opt_idx < 0)
+		return -EINVAL;
+
+	/* Verify the fields asked for can actually be assigned based on type */
+	if (cmd->data & ~FBNIC_L4_HASH_OPTIONS ||
+	    (hash_opt_idx > FBNIC_L4_HASH_OPT &&
+	     cmd->data & ~FBNIC_L3_HASH_OPTIONS) ||
+	    (hash_opt_idx > FBNIC_IP_HASH_OPT &&
+	     cmd->data & ~FBNIC_L2_HASH_OPTIONS))
+		return -EINVAL;
+
+	fbn->rss_flow_hash[hash_opt_idx] = cmd->data;
+
+	if (netif_running(fbn->netdev)) {
+		fbnic_rss_reinit(fbn->fbd, fbn);
+		fbnic_write_rules(fbn->fbd);
+	}
+
+	return 0;
+}
+
 static int
 fbnic_modify_rxfh_context(struct net_device *netdev,
 			  struct ethtool_rxfh_context *ctx,
@@ -1633,6 +1632,8 @@ static const struct ethtool_ops fbnic_ethtool_ops = {
 	.get_rxfh_indir_size	= fbnic_get_rxfh_indir_size,
 	.get_rxfh		= fbnic_get_rxfh,
 	.set_rxfh		= fbnic_set_rxfh,
+	.get_rxfh_fields	= fbnic_get_rss_hash_opts,
+	.set_rxfh_fields	= fbnic_set_rss_hash_opts,
 	.create_rxfh_context	= fbnic_create_rxfh_context,
 	.modify_rxfh_context	= fbnic_modify_rxfh_context,
 	.remove_rxfh_context	= fbnic_remove_rxfh_context,
-- 
2.49.0

[PATCH net-next 7/9] net: drv: vmxnet3: migrate to new RXFH callbacks

[Jakub Kicinski]

Add support for the new rxfh_fields callbacks, instead of de-muxing
the rxnfc calls. This driver does not support flow filtering so
the set_rxnfc callback is completely removed.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: ronak.doshi@broadcom.com
CC: bcm-kernel-feedback-list@broadcom.com
---
 drivers/net/vmxnet3/vmxnet3_ethtool.c | 74 +++++++++------------------
 1 file changed, 25 insertions(+), 49 deletions(-)

diff --git a/drivers/net/vmxnet3/vmxnet3_ethtool.c b/drivers/net/vmxnet3/vmxnet3_ethtool.c
index 471f91c4204a..cc4d7573839d 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethtool.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethtool.c
@@ -833,11 +833,19 @@ vmxnet3_set_ringparam(struct net_device *netdev,
 }
 
 static int
-vmxnet3_get_rss_hash_opts(struct vmxnet3_adapter *adapter,
-			  struct ethtool_rxnfc *info)
+vmxnet3_get_rss_hash_opts(struct net_device *netdev,
+			  struct ethtool_rxfh_fields *info)
 {
+	struct vmxnet3_adapter *adapter = netdev_priv(netdev);
 	enum Vmxnet3_RSSField rss_fields;
 
+	if (!VMXNET3_VERSION_GE_4(adapter))
+		return -EOPNOTSUPP;
+#ifdef VMXNET3_RSS
+	if (!adapter->rss)
+		return -EOPNOTSUPP;
+#endif
+
 	if (netif_running(adapter->netdev)) {
 		unsigned long flags;
 
@@ -900,10 +908,20 @@ vmxnet3_get_rss_hash_opts(struct vmxnet3_adapter *adapter,
 
 static int
 vmxnet3_set_rss_hash_opt(struct net_device *netdev,
-			 struct vmxnet3_adapter *adapter,
-			 struct ethtool_rxnfc *nfc)
+			 const struct ethtool_rxfh_fields *nfc,
+			 struct netlink_ext_ack *extack)
 {
-	enum Vmxnet3_RSSField rss_fields = adapter->rss_fields;
+	struct vmxnet3_adapter *adapter = netdev_priv(netdev);
+	enum Vmxnet3_RSSField rss_fields;
+
+	if (!VMXNET3_VERSION_GE_4(adapter))
+		return -EOPNOTSUPP;
+#ifdef VMXNET3_RSS
+	if (!adapter->rss)
+		return -EOPNOTSUPP;
+#endif
+
+	rss_fields = adapter->rss_fields;
 
 	/* RSS does not support anything other than hashing
 	 * to queues on src and dst IPs and ports
@@ -1074,19 +1092,6 @@ vmxnet3_get_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *info,
 	case ETHTOOL_GRXRINGS:
 		info->data = adapter->num_rx_queues;
 		break;
-	case ETHTOOL_GRXFH:
-		if (!VMXNET3_VERSION_GE_4(adapter)) {
-			err = -EOPNOTSUPP;
-			break;
-		}
-#ifdef VMXNET3_RSS
-		if (!adapter->rss) {
-			err = -EOPNOTSUPP;
-			break;
-		}
-#endif
-		err = vmxnet3_get_rss_hash_opts(adapter, info);
-		break;
 	default:
 		err = -EOPNOTSUPP;
 		break;
@@ -1095,36 +1100,6 @@ vmxnet3_get_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *info,
 	return err;
 }
 
-static int
-vmxnet3_set_rxnfc(struct net_device *netdev, struct ethtool_rxnfc *info)
-{
-	struct vmxnet3_adapter *adapter = netdev_priv(netdev);
-	int err = 0;
-
-	if (!VMXNET3_VERSION_GE_4(adapter)) {
-		err = -EOPNOTSUPP;
-		goto done;
-	}
-#ifdef VMXNET3_RSS
-	if (!adapter->rss) {
-		err = -EOPNOTSUPP;
-		goto done;
-	}
-#endif
-
-	switch (info->cmd) {
-	case ETHTOOL_SRXFH:
-		err = vmxnet3_set_rss_hash_opt(netdev, adapter, info);
-		break;
-	default:
-		err = -EOPNOTSUPP;
-		break;
-	}
-
-done:
-	return err;
-}
-
 #ifdef VMXNET3_RSS
 static u32
 vmxnet3_get_rss_indir_size(struct net_device *netdev)
@@ -1361,12 +1336,13 @@ static const struct ethtool_ops vmxnet3_ethtool_ops = {
 	.get_ringparam     = vmxnet3_get_ringparam,
 	.set_ringparam     = vmxnet3_set_ringparam,
 	.get_rxnfc         = vmxnet3_get_rxnfc,
-	.set_rxnfc         = vmxnet3_set_rxnfc,
 #ifdef VMXNET3_RSS
 	.get_rxfh_indir_size = vmxnet3_get_rss_indir_size,
 	.get_rxfh          = vmxnet3_get_rss,
 	.set_rxfh          = vmxnet3_set_rss,
 #endif
+	.get_rxfh_fields   = vmxnet3_get_rss_hash_opts,
+	.set_rxfh_fields   = vmxnet3_set_rss_hash_opt,
 	.get_link_ksettings = vmxnet3_get_link_ksettings,
 	.get_channels      = vmxnet3_get_channels,
 };
-- 
2.49.0

[PATCH net-next 8/9] net: drv: virtio: migrate to new RXFH callbacks

[Jakub Kicinski]

Add support for the new rxfh_fields callbacks, instead of de-muxing
the rxnfc calls. This driver does not support flow filtering so
the set_rxnfc callback is completely removed.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: mst@redhat.com
CC: jasowang@redhat.com
CC: xuanzhuo@linux.alibaba.com
CC: eperezma@redhat.com
CC: virtualization@lists.linux.dev
---
 drivers/net/virtio_net.c | 47 +++++++++++++++-------------------------
 1 file changed, 18 insertions(+), 29 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index e53ba600605a..07e41dce4203 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -4193,8 +4193,11 @@ static void virtnet_init_default_rss(struct virtnet_info *vi)
 	netdev_rss_key_fill(vi->rss_hash_key_data, vi->rss_key_size);
 }
 
-static void virtnet_get_hashflow(const struct virtnet_info *vi, struct ethtool_rxnfc *info)
+static int virtnet_get_hashflow(struct net_device *dev,
+				struct ethtool_rxfh_fields *info)
 {
+	struct virtnet_info *vi = netdev_priv(dev);
+
 	info->data = 0;
 	switch (info->flow_type) {
 	case TCP_V4_FLOW:
@@ -4243,17 +4246,22 @@ static void virtnet_get_hashflow(const struct virtnet_info *vi, struct ethtool_r
 		info->data = 0;
 		break;
 	}
+
+	return 0;
 }
 
-static bool virtnet_set_hashflow(struct virtnet_info *vi, struct ethtool_rxnfc *info)
+static int virtnet_set_hashflow(struct net_device *dev,
+				const struct ethtool_rxfh_fields *info,
+				struct netlink_ext_ack *extack)
 {
+	struct virtnet_info *vi = netdev_priv(dev);
 	u32 new_hashtypes = vi->rss_hash_types_saved;
 	bool is_disable = info->data & RXH_DISCARD;
 	bool is_l4 = info->data == (RXH_IP_SRC | RXH_IP_DST | RXH_L4_B_0_1 | RXH_L4_B_2_3);
 
 	/* supports only 'sd', 'sdfn' and 'r' */
 	if (!((info->data == (RXH_IP_SRC | RXH_IP_DST)) | is_l4 | is_disable))
-		return false;
+		return -EINVAL;
 
 	switch (info->flow_type) {
 	case TCP_V4_FLOW:
@@ -4292,21 +4300,22 @@ static bool virtnet_set_hashflow(struct virtnet_info *vi, struct ethtool_rxnfc *
 		break;
 	default:
 		/* unsupported flow */
-		return false;
+		return -EINVAL;
 	}
 
 	/* if unsupported hashtype was set */
 	if (new_hashtypes != (new_hashtypes & vi->rss_hash_types_supported))
-		return false;
+		return -EINVAL;
 
 	if (new_hashtypes != vi->rss_hash_types_saved) {
 		vi->rss_hash_types_saved = new_hashtypes;
 		vi->rss_hdr->hash_types = cpu_to_le32(vi->rss_hash_types_saved);
 		if (vi->dev->features & NETIF_F_RXHASH)
-			return virtnet_commit_rss_command(vi);
+			if (!virtnet_commit_rss_command(vi))
+				return -EINVAL;
 	}
 
-	return true;
+	return 0;
 }
 
 static void virtnet_get_drvinfo(struct net_device *dev,
@@ -5539,27 +5548,6 @@ static int virtnet_get_rxnfc(struct net_device *dev, struct ethtool_rxnfc *info,
 	switch (info->cmd) {
 	case ETHTOOL_GRXRINGS:
 		info->data = vi->curr_queue_pairs;
-		break;
-	case ETHTOOL_GRXFH:
-		virtnet_get_hashflow(vi, info);
-		break;
-	default:
-		rc = -EOPNOTSUPP;
-	}
-
-	return rc;
-}
-
-static int virtnet_set_rxnfc(struct net_device *dev, struct ethtool_rxnfc *info)
-{
-	struct virtnet_info *vi = netdev_priv(dev);
-	int rc = 0;
-
-	switch (info->cmd) {
-	case ETHTOOL_SRXFH:
-		if (!virtnet_set_hashflow(vi, info))
-			rc = -EINVAL;
-
 		break;
 	default:
 		rc = -EOPNOTSUPP;
@@ -5591,8 +5579,9 @@ static const struct ethtool_ops virtnet_ethtool_ops = {
 	.get_rxfh_indir_size = virtnet_get_rxfh_indir_size,
 	.get_rxfh = virtnet_get_rxfh,
 	.set_rxfh = virtnet_set_rxfh,
+	.get_rxfh_fields = virtnet_get_hashflow,
+	.set_rxfh_fields = virtnet_set_hashflow,
 	.get_rxnfc = virtnet_get_rxnfc,
-	.set_rxnfc = virtnet_set_rxnfc,
 };
 
 static void virtnet_get_queue_stats_rx(struct net_device *dev, int i,
-- 
2.49.0

Re: [PATCH net-next 8/9] net: drv: virtio: migrate to new RXFH callbacks

[Jason Wang]

On Wed, Jun 11, 2025 at 11:00 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> Add support for the new rxfh_fields callbacks, instead of de-muxing
> the rxnfc calls. This driver does not support flow filtering so
> the set_rxnfc callback is completely removed.
>
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: mst@redhat.com
> CC: jasowang@redhat.com
> CC: xuanzhuo@linux.alibaba.com
> CC: eperezma@redhat.com
> CC: virtualization@lists.linux.dev
> ---

Acked-by: Jason Wang <jasowang@redhat.com>

Thanks

[PATCH net-next 9/9] net: drv: hyperv: migrate to new RXFH callbacks

[Jakub Kicinski]

Add support for the new rxfh_fields callbacks, instead of de-muxing
the rxnfc calls. This driver does not support flow filtering so
the set_rxnfc callback is completely removed.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: kys@microsoft.com
CC: haiyangz@microsoft.com
CC: wei.liu@kernel.org
CC: decui@microsoft.com
CC: linux-hyperv@vger.kernel.org
---
 drivers/net/hyperv/netvsc_drv.c | 30 +++++++++++-------------------
 1 file changed, 11 insertions(+), 19 deletions(-)

diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index c41a025c66f0..42d98e99566e 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -1580,9 +1580,10 @@ static void netvsc_get_strings(struct net_device *dev, u32 stringset, u8 *data)
 }
 
 static int
-netvsc_get_rss_hash_opts(struct net_device_context *ndc,
-			 struct ethtool_rxnfc *info)
+netvsc_get_rxfh_fields(struct net_device *ndev,
+		       struct ethtool_rxfh_fields *info)
 {
+	struct net_device_context *ndc = netdev_priv(ndev);
 	const u32 l4_flag = RXH_L4_B_0_1 | RXH_L4_B_2_3;
 
 	info->data = RXH_IP_SRC | RXH_IP_DST;
@@ -1637,16 +1638,17 @@ netvsc_get_rxnfc(struct net_device *dev, struct ethtool_rxnfc *info,
 	case ETHTOOL_GRXRINGS:
 		info->data = nvdev->num_chn;
 		return 0;
-
-	case ETHTOOL_GRXFH:
-		return netvsc_get_rss_hash_opts(ndc, info);
 	}
 	return -EOPNOTSUPP;
 }
 
-static int netvsc_set_rss_hash_opts(struct net_device_context *ndc,
-				    struct ethtool_rxnfc *info)
+static int
+netvsc_set_rxfh_fields(struct net_device *dev,
+		       const struct ethtool_rxfh_fields *info,
+		       struct netlink_ext_ack *extack)
 {
+	struct net_device_context *ndc = netdev_priv(dev);
+
 	if (info->data == (RXH_IP_SRC | RXH_IP_DST |
 			   RXH_L4_B_0_1 | RXH_L4_B_2_3)) {
 		switch (info->flow_type) {
@@ -1701,17 +1703,6 @@ static int netvsc_set_rss_hash_opts(struct net_device_context *ndc,
 	return -EOPNOTSUPP;
 }
 
-static int
-netvsc_set_rxnfc(struct net_device *ndev, struct ethtool_rxnfc *info)
-{
-	struct net_device_context *ndc = netdev_priv(ndev);
-
-	if (info->cmd == ETHTOOL_SRXFH)
-		return netvsc_set_rss_hash_opts(ndc, info);
-
-	return -EOPNOTSUPP;
-}
-
 static u32 netvsc_get_rxfh_key_size(struct net_device *dev)
 {
 	return NETVSC_HASH_KEYLEN;
@@ -1979,11 +1970,12 @@ static const struct ethtool_ops ethtool_ops = {
 	.set_channels   = netvsc_set_channels,
 	.get_ts_info	= ethtool_op_get_ts_info,
 	.get_rxnfc	= netvsc_get_rxnfc,
-	.set_rxnfc	= netvsc_set_rxnfc,
 	.get_rxfh_key_size = netvsc_get_rxfh_key_size,
 	.get_rxfh_indir_size = netvsc_rss_indir_size,
 	.get_rxfh	= netvsc_get_rxfh,
 	.set_rxfh	= netvsc_set_rxfh,
+	.get_rxfh_fields = netvsc_get_rxfh_fields,
+	.set_rxfh_fields = netvsc_set_rxfh_fields,
 	.get_link_ksettings = netvsc_get_link_ksettings,
 	.set_link_ksettings = netvsc_set_link_ksettings,
 	.get_ringparam	= netvsc_get_ringparam,
-- 
2.49.0

Re: [PATCH net-next 0/9] net: ethtool: add dedicated RXFH driver callbacks

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Wed, 11 Jun 2025 07:59:40 -0700 you wrote:
> Andrew asked me to plumb the RXFH header fields configuration
> thru to netlink. Before we do that we need to clean up the driver
> facing API a little bit. Right now RXFH configuration shares the
> callbacks with n-tuple filters. The future of n-tuple filters
> is uncertain within netlink. Separate the two for clarity both
> of the core code and the driver facing API.
> 
> [...]

Here is the summary with links:
  - [net-next,1/9] net: ethtool: copy the rxfh flow handling
    https://git.kernel.org/netdev/net-next/c/f4f126535546
  - [net-next,2/9] net: ethtool: remove the duplicated handling from rxfh and rxnfc
    https://git.kernel.org/netdev/net-next/c/2a644c5cecc0
  - [net-next,3/9] net: ethtool: require drivers to opt into the per-RSS ctx RXFH
    https://git.kernel.org/netdev/net-next/c/fac4b41741b5
  - [net-next,4/9] net: ethtool: add dedicated callbacks for getting and setting rxfh fields
    https://git.kernel.org/netdev/net-next/c/9bb00786fc61
  - [net-next,5/9] eth: remove empty RXFH handling from drivers
    https://git.kernel.org/netdev/net-next/c/86b2315e7041
  - [net-next,6/9] eth: fbnic: migrate to new RXFH callbacks
    https://git.kernel.org/netdev/net-next/c/2a34007ba977
  - [net-next,7/9] net: drv: vmxnet3: migrate to new RXFH callbacks
    https://git.kernel.org/netdev/net-next/c/2f14765d6397
  - [net-next,8/9] net: drv: virtio: migrate to new RXFH callbacks
    https://git.kernel.org/netdev/net-next/c/63d474cfb596
  - [net-next,9/9] net: drv: hyperv: migrate to new RXFH callbacks
    https://git.kernel.org/netdev/net-next/c/6867fbe3a9f4

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

Re: [PATCH net-next 0/9] net: ethtool: add dedicated RXFH driver callbacks

[Gal Pressman]

Hi Jakub,

On 11/06/2025 17:59, Jakub Kicinski wrote:
> The future of n-tuple filters is uncertain within netlink.

What does that mean exactly?

Re: [PATCH net-next 0/9] net: ethtool: add dedicated RXFH driver callbacks

[Jakub Kicinski]

On Wed, 25 Jun 2025 09:44:04 +0300 Gal Pressman wrote:
> On 11/06/2025 17:59, Jakub Kicinski wrote:
> > The future of n-tuple filters is uncertain within netlink.  
> 
> What does that mean exactly?

Just that I don't have a clear idea of where it should go.
In this series I was refactoring rxnfc code - one could argue
I should also add dedicated callbacks for n-tuple ioctls
which are also muxed into the rxnfc driver op. But at some
point some people were pushing for n-tuple filters to be
deprecated in favor of cls_flower. Also Jamal's P4-ish
proposal could become n-tuple-next-gen. Or we could lift
existing ntuple filters into netlink.. 🤷️ 

IDK where we are on that, so I'm not refactoring the ntuple callbacks.