Re: [PATCH net-next 2/2] net: bcmgenet: Support calling set_pauseparam from panic context

[Florian Fainelli <florian.fainelli@broadcom.com>]

On 11/3/25 14:19, Andrew Lunn wrote:
>> @@ -139,7 +141,8 @@ void bcmgenet_phy_pause_set(struct net_device *dev, bool rx, bool tx)
>>   	linkmode_mod_bit(ETHTOOL_LINK_MODE_Pause_BIT, phydev->advertising, rx);
>>   	linkmode_mod_bit(ETHTOOL_LINK_MODE_Asym_Pause_BIT, phydev->advertising,
>>   			 rx | tx);
>> -	phy_start_aneg(phydev);
>> +	if (!panic_in_progress())
>> +		phy_start_aneg(phydev);
> 
> 
> That does not look correct. If pause autoneg is off, there is no need
> to trigger an autoneg.
> 
> This all looks pretty messy.

That is pre-existing code, so it would be a separate path in order to 
fix, though point taken.

> 
> Maybe rather than overload set_pauseparams, maybe add a new ethtool
> call to force pause off?

Yes, I like that idea, that way it is clear which drivers support 
disabling pause from a panic context.

> 
> It looks like it would be something like:
> 
> struct bcmgenet_priv *priv = netdev_priv(dev);
> u32 reg;
> 
> reg = bcmgenet_umac_readl(priv, UMAC_CMD);
> reg &= ~(CMD_RX_PAUSE_IGNORE| CMD_TX_PAUSE_IGNORE);
> bcmgenet_umac_writel(priv, reg, UMAC_CMD);

Yes, that is essentially what needs to be done.

Thanks!
-- 
Florian