net: stmmac: weirdness in stmmac_hw_setup()

[Furong Xu]

On Thu, 20 Feb 2025 16:17:43 +0000, "Russell King (Oracle)" wrote:

> While looking through the stmmac driver, I've come across some
> weirdness in stmmac_hw_setup() which looks completely barmy to me.
> 
> It seems that it follows the initialisation suggested by Synopsys
> (as best I can determine from the iMX8MP documentation), even going
> as far as to *enable* transmit and receive *before* the network
> device has been administratively brought up. stmmac_hw_setup() does
> this:
> 
>         /* Enable the MAC Rx/Tx */
>         stmmac_mac_set(priv, priv->ioaddr, true);
> 
> which sets the TE and RE bits in the MAC configuration register.
> 
> This means that if the network link is active, packets will start
> to be received and will be placed into the receive descriptors.
> 
> We won't transmit anything because we won't be placing packets in
> the transmit descriptors to be transmitted.
> 
> However, this in stmmac_hw_setup() is just wrong. Can it be deleted
> as per the below?
> Tested-by: Thierry Reding <treding at nvidia.com>
> 
> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> index c2913f003fe6..d6e492f523f5 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> @@ -3493,9 +3493,6 @@ static int stmmac_hw_setup(struct net_device *dev, bool ptp_register)
>  		priv->hw->rx_csum = 0;
>  	}
>  
> -	/* Enable the MAC Rx/Tx */
> -	stmmac_mac_set(priv, priv->ioaddr, true);
> -
>  	/* Set the HW DMA mode and the COE */
>  	stmmac_dma_operation_mode(priv);
>  

A better fix here:
https://lore.kernel.org/all/tencent_CCC29C4F562F2DEFE48289DB52F4D91BDE05@qq.com/

I can reproduce and confirm that patch works well.
I was going to leave a Tested-by: tag on that patch if everything looks good enough,
but the author (now copied) never come back.