[PATCH net-next] net: reject a netdev that implements only one hwtstamp NDO

[PATCH net-next] net: reject a netdev that implements only one hwtstamp NDO

[Nicolai Buchwitz]

ndo_hwtstamp_get and ndo_hwtstamp_set only make sense as a pair. A
driver providing just one gets no complaint today, and the mismatch
turns up in user space, where a set can succeed while the matching get
fails.

Catch it at registration instead.

Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Nicolai Buchwitz <nb@tipi-net.de>
---
 net/core/dev.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/net/core/dev.c b/net/core/dev.c
index f660fccfc0db..574fd0ba9080 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -11447,6 +11447,13 @@ int register_netdevice(struct net_device *dev)
 		goto err_uninit;
 	}
 
+	if (!dev->netdev_ops->ndo_hwtstamp_get !=
+	    !dev->netdev_ops->ndo_hwtstamp_set) {
+		netdev_WARN(dev, "driver implements only one hwtstamp NDO\n");
+		ret = -EINVAL;
+		goto err_uninit;
+	}
+
 	if (netdev_need_ops_lock(dev) &&
 	    dev->netdev_ops->ndo_set_rx_mode &&
 	    !dev->netdev_ops->ndo_set_rx_mode_async)

base-commit: 014d795c73837ea2339a4ea8e8f82c6e959b845d
-- 
2.53.0

Re: [PATCH net-next] net: reject a netdev that implements only one hwtstamp NDO

[Jakub Kicinski]

On Tue, 29 Sep 2026 23:32:32 +0200 Nicolai Buchwitz wrote:
> @@ -11447,6 +11447,13 @@ int register_netdevice(struct net_device *dev)
>  		goto err_uninit;
>  	}
>  
> +	if (!dev->netdev_ops->ndo_hwtstamp_get !=
> +	    !dev->netdev_ops->ndo_hwtstamp_set) {
> +		netdev_WARN(dev, "driver implements only one hwtstamp NDO\n");
> +		ret = -EINVAL;
> +		goto err_uninit;
> +	}
> +
>  	if (netdev_need_ops_lock(dev) &&
>  	    dev->netdev_ops->ndo_set_rx_mode &&
>  	    !dev->netdev_ops->ndo_set_rx_mode_async)

Third such check added here.. we should probably factor all three
out of register_netdevice(). Hardest part is probably naming the
new helper. netdev_check_ops() ? We already have ethtool_check_ops()
We can't move it earlier because some drivers set features form
ndo_init And while you move it you can also make the warning
about not using _async return an error. It was "soft" to avoid
bricking people's systems on boot, but it's been there since v7.1
people will have enough time to spot offenders by v7.4
-- 
pw-bot: cr