[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aLhkKVsbrkXmFbgK@shell.armlinux.org.uk>
Date: Wed, 3 Sep 2025 16:52:09 +0100
From: "Russell King (Oracle)" <linux@...linux.org.uk>
To: Vladimir Oltean <vladimir.oltean@....com>
Cc: netdev@...r.kernel.org, Andrew Lunn <andrew@...n.ch>,
Heiner Kallweit <hkallweit1@...il.com>,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 net 1/2] net: phylink: add lock for serializing
concurrent pl->phydev writes with resolver
On Wed, Sep 03, 2025 at 06:31:20PM +0300, Vladimir Oltean wrote:
> On Wed, Sep 03, 2025 at 04:26:35PM +0100, Russell King (Oracle) wrote:
> > On Wed, Sep 03, 2025 at 06:23:47PM +0300, Vladimir Oltean wrote:
> > > @@ -2305,6 +2314,7 @@ void phylink_disconnect_phy(struct phylink *pl)
> > >
> > > phy = pl->phydev;
> > > if (phy) {
> > > + mutex_lock(&pl->phy_lock);
> >
> > If we can, I think it would be better to place this a couple of lines
> > above and move the unlock.
>
> Sorry for potentially misunderstanding, do you mean like this?
>
> mutex_lock(&pl->phy_lock);
> phy = pl->phydev;
> if (phy) {
> mutex_lock(&phy->lock);
> mutex_lock(&pl->state_mutex);
> pl->phydev = NULL;
> pl->phy_enable_tx_lpi = false;
> pl->mac_tx_clk_stop = false;
> mutex_unlock(&pl->state_mutex);
> mutex_unlock(&phy->lock);
> mutex_unlock(&pl->phy_lock);
> flush_work(&pl->resolve);
>
> phy_disconnect(phy);
> } else {
> mutex_unlock(&pl->phy_lock);
> }
>
> move the unlock where? because flush_work(&pl->resolve) needs to happen
> unlocked, otherwise we'll deadlock with phylink_resolve().
>
> Additionally, dereferincing pl->phydev under rtnl_lock() is already safe,
> and doesn't need the secondary clock.
The reason I'm making the suggestion is for consistency. If the lock
is there to ensure that reading pl->phydev is done safely, having one
site where we read it and then take the lock makes it look confusing.
I've also been thinking that it should be called pl->phydev_mutex
(note that phylink uses _mutex for mutexes.)
To avoid it looking weird, what about this:
mutex_lock(&pl->phy_lock);
phy = pl->phydev;
if (phy) {
mutex_lock(&phy->lock);
mutex_lock(&pl->state_mutex);
pl->phydev = NULL;
pl->phy_enable_tx_lpi = false;
pl->mac_tx_clk_stop = false;
mutex_unlock(&pl->state_mutex);
mutex_unlock(&phy->lock);
}
mutex_unlock(&pl->phy_lock);
if (phy)
flush_work(&pl->resolve);
phy_disconnect(phy);
}
--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
Powered by blists - more mailing lists