Re: [PATCH net-next] net: macb: allow MTU changes while the interface is running

["Théo Lebrun" <theo.lebrun@bootlin.com>]

Hello Nicolai,

On Mon Mar 16, 2026 at 10:27 AM CET, Nicolai Buchwitz wrote:
> macb_change_mtu() currently returns -EBUSY if the interface is running,
> requiring users to bring the interface down before changing the MTU. This
> is unnecessarily restrictive.

One valid reasoning for making the operation return -EBUSY is that
macb_close() will lead to packet loss for a few seconds. Users might
not expect an MTU change to trigger that. -EBUSY means user (be it
human or software) becomes aware and has time for a second thought.

But clearly, it becomes annoying as the user in the long run, when you
own your platform and know the implications.

> Instead, close and reopen the interface around the MTU change so that RX
> DMA buffers are reallocated for the new MTU. This is the same approach
> used by many other network drivers (e.g. igb, tg3, stmmac).
>
> Signed-off-by: Nicolai Buchwitz <nb@tipi-net.de>
> ---
>  drivers/net/ethernet/cadence/macb_main.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
> index 5e27e0e87a55..8dd01031250d 100644
> --- a/drivers/net/ethernet/cadence/macb_main.c
> +++ b/drivers/net/ethernet/cadence/macb_main.c
> @@ -3262,11 +3262,16 @@ static int macb_close(struct net_device *dev)
>  
>  static int macb_change_mtu(struct net_device *dev, int new_mtu)
>  {
> -	if (netif_running(dev))
> -		return -EBUSY;
> +	bool was_running = netif_running(dev);
> +
> +	if (was_running)
> +		macb_close(dev);
>  
>  	WRITE_ONCE(dev->mtu, new_mtu);
>  
> +	if (was_running)
> +		return macb_open(dev);
> +
>  	return 0;
>  }

Reviewed-by: Théo Lebrun <theo.lebrun@bootlin.com>
Tested-by: Théo Lebrun <theo.lebrun@bootlin.com> # on eyeq5-epm

Thanks,

--
Théo Lebrun, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com