Re: [PATCH] usb: phy: Initialize struct usb_phy list_head

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Fri, Nov 21, 2025 at 04:39:58PM +0000, Diogo Ivo wrote:
> 
> 
> On 11/21/25 15:03, Greg Kroah-Hartman wrote:
> > On Fri, Nov 21, 2025 at 02:55:35PM +0000, Diogo Ivo wrote:
> > > 
> > > 
> > > On 11/21/25 14:09, Greg Kroah-Hartman wrote:
> > > > On Thu, Nov 13, 2025 at 02:59:06PM +0000, Diogo Ivo wrote:
> > > > > When executing usb_add_phy() and usb_add_phy_dev() it is possible that
> > > > > usb_add_extcon() fails (for example with -EPROBE_DEFER), in which case
> > > > > the usb_phy does not get added to phy_list via
> > > > > list_add_tail(&x->head, phy_list).
> > > > > 
> > > > > Then, when the driver that tried to add the phy receives the error
> > > > > propagated from usb_add_extcon() and calls into usb_remove_phy() to
> > > > > undo the partial registration there will be an unconditional call to
> > > > > list_del(&x->head) which is notinitialized and leads to a NULL pointer
> > > > > dereference.
> > > > > 
> > > > > Fix this by initializing x->head before usb_add_extcon() has a chance to
> > > > > fail.
> > > > > 
> > > > > Fixes: 7d21114dc6a2d53 ("usb: phy: Introduce one extcon device into usb phy")
> > > > > Cc: stable@vger.kernel.org
> > > > > Signed-off-by: Diogo Ivo <diogo.ivo@tecnico.ulisboa.pt>
> > > > > ---
> > > > >    drivers/usb/phy/phy.c | 4 ++++
> > > > >    1 file changed, 4 insertions(+)
> > > > > 
> > > > > diff --git a/drivers/usb/phy/phy.c b/drivers/usb/phy/phy.c
> > > > > index e1435bc59662..5a9b9353f343 100644
> > > > > --- a/drivers/usb/phy/phy.c
> > > > > +++ b/drivers/usb/phy/phy.c
> > > > > @@ -646,6 +646,8 @@ int usb_add_phy(struct usb_phy *x, enum usb_phy_type type)
> > > > >    		return -EINVAL;
> > > > >    	}
> > > > > +	INIT_LIST_HEAD(&x->head);
> > > > > +
> > > > >    	usb_charger_init(x);
> > > > >    	ret = usb_add_extcon(x);
> > > > >    	if (ret)
> > > > > @@ -696,6 +698,8 @@ int usb_add_phy_dev(struct usb_phy *x)
> > > > >    		return -EINVAL;
> > > > >    	}
> > > > > +	INIT_LIST_HEAD(&x->head);
> > > > > +
> > > > >    	usb_charger_init(x);
> > > > >    	ret = usb_add_extcon(x);
> > > > >    	if (ret)
> > > > > 
> > > > 
> > > > Shouldn't you be also removing an existing call to INIT_LIST_HEAD()
> > > > somewhere?  This is not "moving" the code, it is adding it.
> > > 
> > >  From my understanding that's exactly the problem, currently there is no
> > > call to INIT_LIST_HEAD() anywhere on these code paths, meaning that if
> > > we do not reach the point of calling list_add_tail() at the end of
> > > usb_add_phy() and usb_phy_add_dev() then x->head will remain uninitialized
> > > and fault when running usb_remove_phy().
> > 
> > Then how does this work at all if the list is never initialized?
> 
> In this case in drivers/usb/phy/phy.c a static LIST_HEAD(phy_list) is
> declared and then for each new 'struct usb_phy *x' the x->head entry
> will get added to this list by calling list_add_tail(&x->head, &phy_list).

Great, can you document this in the changelog text so that it makes more
sense (and properly quote the fixes: sha1, I think you have too many
digits there...) and resend a v2.

thanks,

greg k-h