netdev - Re: [PATCH net 2/4] net: ethernet: ti: cpsw: fix device and of

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] [day] [month] [year] [list]

Message-ID: <20161101170407.GD2664@localhost>
Date:   Tue, 1 Nov 2016 18:04:07 +0100
From:   Johan Hovold <johan@...nel.org>
To:     David Miller <davem@...emloft.net>
Cc:     johan@...nel.org, f.fainelli@...il.com, mugunthanvnm@...com,
        yisen.zhuang@...wei.com, salil.mehta@...wei.com,
        netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH net 2/4] net: ethernet: ti: cpsw: fix device and of_node
 leaks

On Tue, Nov 01, 2016 at 12:48:48PM -0400, David Miller wrote:
> From: Johan Hovold <johan@...nel.org>
> Date: Tue, 1 Nov 2016 17:42:25 +0100
> 
> > On Tue, Nov 01, 2016 at 12:27:11PM -0400, David Miller wrote:
> >> From: Johan Hovold <johan@...nel.org>
> >> Date: Tue,  1 Nov 2016 12:03:35 +0100
> >> 
> >> > diff --git a/drivers/net/ethernet/ti/cpsw-phy-sel.c b/drivers/net/ethernet/ti/cpsw-phy-sel.c
> >> > index 054a8dd23dae..589beb843f56 100644
> >> > --- a/drivers/net/ethernet/ti/cpsw-phy-sel.c
> >> > +++ b/drivers/net/ethernet/ti/cpsw-phy-sel.c
> >> > @@ -176,8 +176,11 @@ void cpsw_phy_sel(struct device *dev, phy_interface_t phy_mode, int slave)
> >> >  	}
> >> >  
> >> >  	dev = bus_find_device(&platform_bus_type, NULL, node, match);
> >> > +	of_node_put(node);
> >> >  	priv = dev_get_drvdata(dev);
> >> >  
> >> > +	put_device(dev);
> >> > +
> >> >  	priv->cpsw_phy_sel(priv, phy_mode, slave);
> >> >  }
> >> >  EXPORT_SYMBOL_GPL(cpsw_phy_sel);
> >> 
> >> The only reference you have to 'dev' is the one obtained from the
> >> bus_find_device() call, therefore you must at least hold onto
> >> 'dev' until after the priv->cpsw_phy_sel(priv, phy_mode, slave); call.
> > 
> > As I mentioned in the commit message "...there is no guarantee that the
> > devres-managed struct cpsw_phy_sel_priv will continue to be valid until
> > this function returns regardless of this change".
> > 
> > Specifically, holding a reference to dev does not prevent the
> > cpsw_phy_sel driver from being unbound and priv from being freed.
> 
> But you should at least hold onto the object while you call a function
> pointer embedded in a data structure referred by it.

But there is no such reference to be held (currently). While priv is
valid (i.e. dev has a driver bound), driver core will hold a reference
to dev. If someone unbinds the driver, priv will be gone long before dev
and all bets are off anyway.

So I released the reference to dev before dereferencing priv on purpose
to avoid having someone make false assumptions about the lifetime of priv.

Johan