Re: [PATCH v8 net-next 10/12] net: transfer rtnl_lock() requirement from ethtool_set_ethtool_phy_ops() to caller

["Russell King (Oracle)" <linux@armlinux.org.uk>]

On Mon, Jul 17, 2023 at 06:27:07PM +0300, Vladimir Oltean wrote:
> phy_init() and phy_exit() will have to do more stuff under rtnl_lock()
> in a future change. Since rtnl_unlock() -> netdev_run_todo() does a lot
> of stuff under the hood, it's a pity to lock and unlock the rtnetlink
> mutex twice in a row.
> 
> Change the calling convention such that the only caller of
> ethtool_set_ethtool_phy_ops(), phy_device.c, provides a context where
> the rtnl_mutex is already acquired.
> 
> Note that phy_exit() wasn't performing the opposite teardown of
> phy_init(). Reverse mdio_bus_init() with ethtool_set_ethtool_phy_ops(),
> so that this is now the case.

To me, this looks buggy.

> @@ -3451,11 +3452,14 @@ static int __init phy_init(void)
>  {
>  	int rc;
>  
> +	rtnl_lock();
> +	ethtool_set_ethtool_phy_ops(&phy_ethtool_phy_ops);
> +	rtnl_unlock();
> +
>  	rc = mdio_bus_init();
>  	if (rc)
>  		return rc;

If mdio_bus_init() fails, and phylib is built as a module, then we
leave ethtool_phy_ops pointing into module space that has potentially
been freed or re-used for another module. This error path needs to
properly clean up.

The same is also true for the other failure paths in phy_init() which
already do not cater for their failures leaving a dangling pointer in
ethtool_phy_ops. This should probably be fixed first in a separate
patch for the net tree.

Thanks.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!