linux-kernel - Re: [PATCH] net: dsa: Fix conditional handling of Wake-on-Lan configuration in dsa_user_set

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <ZwOiNQSNJ7CzqbO1@shell.armlinux.org.uk>
Date: Mon, 7 Oct 2024 09:56:21 +0100
From: "Russell King (Oracle)" <linux@...linux.org.uk>
To: Andrew Lunn <andrew@...n.ch>
Cc: Mohammed Anees <pvmohammedanees2003@...il.com>, davem@...emloft.net,
	edumazet@...gle.com, f.fainelli@...il.com, kuba@...nel.org,
	linux-kernel@...r.kernel.org, netdev@...r.kernel.org,
	olteanv@...il.com, pabeni@...hat.com
Subject: Re: [PATCH] net: dsa: Fix conditional handling of Wake-on-Lan
 configuration in dsa_user_set_wol

On Sun, Oct 06, 2024 at 09:57:26PM +0200, Andrew Lunn wrote:
> On Sun, Oct 06, 2024 at 09:40:32PM +0530, Mohammed Anees wrote:
> > Considering the insight you've provided, I've written the code below
> > 
> > static int dsa_user_set_wol(struct net_device *dev, struct ethtool_wolinfo *w)
> > {
> > 	struct dsa_port *dp = dsa_user_to_port(dev);
> > 	struct dsa_switch *ds = dp->ds;
> > 	int ret;
> > 
> > 	ret = phylink_ethtool_set_wol(dp->pl, w);
> > 
> > 	if (ret != -EOPNOTSUPP)
> > 		return ret;
> > 
> > 	if (ds->ops->set_wol)
> > 		ret = ds->ops->set_wol(ds, dp->index, w);
> > 		if (ret != -EOPNOTSUPP)
> > 			return ret;
> 
> This can be simplified to just:
> 
> > 	if (ds->ops->set_wol)
> > 		return ds->ops->set_wol(ds, dp->index, w);
> > 
> > 	return -EOPNOTSUPP;
> > }

I don't think the above is correct. While the simplification is, the
overall logic is not.

Let's go back to what Andrew said in his previous reply:

"So userspace could say pumbagsf, with the PHY supporting pmub and the
MAC supporting agsf, and the two need to cooperate."

The above does not do this. Let's go back further:

        phylink_ethtool_set_wol(dp->pl, w);

        if (ds->ops->set_wol)
                ret = ds->ops->set_wol(ds, dp->index, w);

The original code _does_ do this, allowing the PHY and MAC to set
their modes, although the return code is not correct.

I notice V2 of the patch has been posted - in my opinion prematurely
because there's clearly the discussion on the first version has not
reached a conclusion yet.

What I would propose is the following:

	int phy_ret, mac_ret;

	phy_ret = phylink_ethtool_set_wol(dp->pl, w);
	if (phy_ret != 0 && phy_ret != -EOPNOTSUPP)
		return phy_ret;

	if (ds->ops->set_wol)
		mac_ret = ds->ops->set_wol(ds, dp->index, w);
	else
		mac_ret = -EOPNOTSUPP;

	if (mac_ret != 0 && mac_ret != -EOPNOTSUPP)
		return mac_ret;

	/* Combine the two return codes. If either returned zero,
	 * then we have been successful.
	 */
	if (phy_ret == 0 || mac_ret == 0)
		return 0;
	
	return -EOPNOTSUPP;

Which I think is the closest one can get to - there is the possibility
for phylink_ethtool_set_wol() to have modified the WoL state, but
ds->ops->set_wol() to fail with an error code, causing this to return
failure, but I don't see that as being avoidable without yet more
complexity.

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