netdev - Re: [PATCH net-next v4 01/13] net: phy: Introduce ethernet link topology representation

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: <20231218094956.676e7ed0@device-28.home>
Date: Mon, 18 Dec 2023 09:49:56 +0100
From: Maxime Chevallier <maxime.chevallier@...tlin.com>
To: Andrew Lunn <andrew@...n.ch>
Cc: Vladimir Oltean <vladimir.oltean@....com>, davem@...emloft.net,
 netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
 thomas.petazzoni@...tlin.com, Jakub Kicinski <kuba@...nel.org>, Eric
 Dumazet <edumazet@...gle.com>, Paolo Abeni <pabeni@...hat.com>, Russell
 King <linux@...linux.org.uk>, linux-arm-kernel@...ts.infradead.org,
 Christophe Leroy <christophe.leroy@...roup.eu>, Herve Codina
 <herve.codina@...tlin.com>, Florian Fainelli <f.fainelli@...il.com>, Heiner
 Kallweit <hkallweit1@...il.com>, Köry Maincent
 <kory.maincent@...tlin.com>, Jesse Brandeburg <jesse.brandeburg@...el.com>,
 Jonathan Corbet <corbet@....net>, Marek Behún
 <kabel@...nel.org>, Piergiorgio Beruto <piergiorgio.beruto@...il.com>,
 Oleksij Rempel <o.rempel@...gutronix.de>, Nicolò Veronese
 <nicveronese@...il.com>
Subject: Re: [PATCH net-next v4 01/13] net: phy: Introduce ethernet link
 topology representation

Hello Andrew,

Thanks for the review,

On Sun, 17 Dec 2023 17:57:10 +0100
Andrew Lunn <andrew@...n.ch> wrote:

> On Fri, Dec 15, 2023 at 11:45:23PM +0200, Vladimir Oltean wrote:
> > On Fri, Dec 15, 2023 at 06:12:23PM +0100, Maxime Chevallier wrote:  
> > > diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c
> > > index d8e9335d415c..89daaccc9276 100644
> > > --- a/drivers/net/phy/phy_device.c
> > > +++ b/drivers/net/phy/phy_device.c
> > > @@ -1491,6 +1500,11 @@ int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
> > >  
> > >  		if (phydev->sfp_bus_attached)
> > >  			dev->sfp_bus = phydev->sfp_bus;
> > > +
> > > +		err = phy_link_topo_add_phy(&dev->link_topo, phydev,
> > > +					    PHY_UPSTREAM_MAC, dev);
> > > +		if (err)
> > > +			goto error;
> > >  	}
> > >  
> > >  	/* Some Ethernet drivers try to connect to a PHY device before
> > > @@ -1816,6 +1830,7 @@ void phy_detach(struct phy_device *phydev)
> > >  	if (dev) {
> > >  		phydev->attached_dev->phydev = NULL;
> > >  		phydev->attached_dev = NULL;
> > > +		phy_link_topo_del_phy(&dev->link_topo, phydev);
> > >  	}
> > >  	phydev->phylink = NULL;
> > >  
> > > diff --git a/drivers/net/phy/phy_link_topology.c b/drivers/net/phy/phy_link_topology.c
> > > new file mode 100644
> > > index 000000000000..22f6372d002c
> > > --- /dev/null
> > > +++ b/drivers/net/phy/phy_link_topology.c
> > > +int phy_link_topo_add_phy(struct phy_link_topology *topo,
> > > +			  struct phy_device *phy,
> > > +			  enum phy_upstream upt, void *upstream)
> > > +{
> > > +	struct phy_device_node *pdn;
> > > +	int ret;
> > > +
> > > +	/* Protects phy and upstream */
> > > +	ASSERT_RTNL();  
> > 
> > Something to think for the PHY library maintainers. This is probably
> > the first time when the rtnl_lock() requirement is asserted at
> > phy_attach_direct() time.  
> 
> There are two use cases here for plain MAC drivers.
> 
> 1) phy_attach_direct() is called from probe. RTNL is normally not
> held, the driver would have to take it before making the call.
> 
> 2) phy_attach_direct() is called from ndo_open. In that case,
> __dev_open() has a ASSERT_RTNL() so we can assume RTNL has been taken.
> 
> So i don't think we can assume RTNL is held, but it might be held.
> 
> We need a better understanding what is being protected here.

I'm protecting the struct phy_device and the *upstream pointer (which
can be a net_device or a struct phy_device as well). In particular, I'm
protecting the phy_device->phyindex field. While I don't see any case
where we would concurrently write to it, given the weird topologies
that might be implemented in the future, I guess better safe than sorry.

For the rest, the actual phy list and the next_index pointer from the
topology should be already protected by the xarray mechanism.

Thanks,

Maxime