Re: [PATCH net] net: bcmgenet: add hwtstamp callbacks to restore PHY timestamping

[Nicolai Buchwitz <nb@tipi-net.de>]

Hi James

On 18.9.2026 05:01, James Clark wrote:
> Hardware timestamp configuration fails with EOPNOTSUPP on BCMGENET
> interfaces connected to a timestamp-capable PHY. On a Raspberry Pi
> CM4, "hwstamp_ctl -i eth0 -t 1 -r 12" fails with "SIOCSHWTSTAMP failed:
> Operation not supported", preventing ptp4l from using hardware
> timestamping.
> 
> Commit 5062245a5a7f ("net: remove legacy way to get/set HW timestamp
> config") removed the legacy ioctl fallback, assuming that all MAC
> drivers supporting hardware timestamping implement ndo_hwtstamp_get()
> and ndo_hwtstamp_set(), even when timestamping is provided only by
> an attached PHY. BCMGENET does not implement these callbacks, so the
> core rejects timestamp configuration requests before they reach
> the PHY.
> 
> Add ndo_hwtstamp_get() and ndo_hwtstamp_set() callbacks so the core
> can dispatch timestamp configuration requests through its phylib
> helpers. Both callbacks return -EOPNOTSUPP because the MAC itself
> does not support hardware timestamping; requests for the PHY are
> handled by the core without invoking these callbacks.

I've tested with and without your patch and can reproduce the issue / 
fix.
IMHO the commit message should mention that 'get' isn't restored - it 
was
already broken as the legacy path through phy_mii_ioctl() only handled
SIOCSHWTSTAMP.

> 
> Fixes: 5062245a5a7f ("net: remove legacy way to get/set HW timestamp 
> config")
> Assisted-by: LLM
> Signed-off-by: James Clark <jjc@jclark.com>
> ---
>  drivers/net/ethernet/broadcom/genet/bcmgenet.c | 15 +++++++++++++++
>  1 file changed, 15 insertions(+)
> 
> diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c 
> b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> index a2305e642..8ead37ff9 100644
> --- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> +++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c
> @@ -3712,6 +3712,19 @@ static int bcmgenet_change_carrier(struct 
> net_device *dev, bool new_carrier)
>  	return 0;
>  }
> 
> +static int bcmgenet_hwtstamp_get(struct net_device *dev,
> +				 struct kernel_hwtstamp_config *cfg)
> +{
> +	return -EOPNOTSUPP;
> +}
> +
> +static int bcmgenet_hwtstamp_set(struct net_device *dev,
> +				 struct kernel_hwtstamp_config *cfg,
> +				 struct netlink_ext_ack *extack)
> +{
> +	return -EOPNOTSUPP;
> +}

The check that rejects the request is in dev_set_hwtstamp():

         if (!ops->ndo_hwtstamp_set)
                 return -EOPNOTSUPP;

It runs before dev_set_hwtstamp_phylib() checks 
phy_is_default_hwtstamp()
and nothing about this is specific to bcmgenet. Personally, I think this 
should
be better handled by the core than adding -EOPNOTSUPP stubs to the 
drivers.

I will send a patch for this.

> +
>  static const struct net_device_ops bcmgenet_netdev_ops = {
>  	.ndo_open		= bcmgenet_open,
>  	.ndo_stop		= bcmgenet_close,
> @@ -3723,6 +3736,8 @@ static const struct net_device_ops 
> bcmgenet_netdev_ops = {
>  	.ndo_set_features	= bcmgenet_set_features,
>  	.ndo_get_stats64	= bcmgenet_get_stats64,
>  	.ndo_change_carrier	= bcmgenet_change_carrier,
> +	.ndo_hwtstamp_get	= bcmgenet_hwtstamp_get,
> +	.ndo_hwtstamp_set	= bcmgenet_hwtstamp_set,
>  };
> 
>  /* GENET hardware parameters/characteristics */

Thanks,
Nicolai