Re: [PATCH] usb: gadget: u_serial: Add null pointer check in gserial_resume

[Greg Kroah-Hartman]

On Thu, Feb 09, 2023 at 10:31:50AM +0530, Prashanth K wrote:
> 
> 
> On 09-02-23 01:51 am, Alan Stern wrote:
> > On Wed, Feb 08, 2023 at 09:15:54PM +0530, Prashanth K wrote:
> > > 
> > > 
> > > On 08-02-23 08:24 pm, Greg Kroah-Hartman wrote:
> > > > On Wed, Feb 08, 2023 at 07:24:47PM +0530, Prashanth K wrote:
> > > > > Consider a case where gserial_disconnect has already cleared
> > > > > gser->ioport. And if a wakeup interrupt triggers afterwards,
> > > > > gserial_resume gets called, which will lead to accessing of
> > > > > gserial->port and thus causing null pointer dereference.Add
> > > > > a null pointer check to prevent this.
> > > > > 
> > > > > Fixes: aba3a8d01d62 (" usb: gadget: u_serial: add suspend resume callbacks")
> > > > 
> > > > Nit, and our tools will complain, no " " before the "usb:" string here,
> > > > right?
> > > > 
> > > Will fix it in next patch.
> > > > 
> > > > 
> > > > > Signed-off-by: Prashanth K <quic_prashk@xxxxxxxxxxx>
> > > > > ---
> > > > >    drivers/usb/gadget/function/u_serial.c | 3 +++
> > > > >    1 file changed, 3 insertions(+)
> > > > > 
> > > > > diff --git a/drivers/usb/gadget/function/u_serial.c b/drivers/usb/gadget/function/u_serial.c
> > > > > index 840626e..98be2b8 100644
> > > > > --- a/drivers/usb/gadget/function/u_serial.c
> > > > > +++ b/drivers/usb/gadget/function/u_serial.c
> > > > > @@ -1428,6 +1428,9 @@ void gserial_resume(struct gserial *gser)
> > > > >    	struct gs_port *port = gser->ioport;
> > > > >    	unsigned long	flags;
> > > > > +	if (!port)
> > > > > +		return;
> > > > > +
> > > > 
> > > > What prevents port from going to NULL right after this check?
> > > In our case we got a null pointer de-reference while performing USB
> > > compliance tests, as the gser->port was null. Because in gserial_resume,
> > > spinlock_irq_save(&port->port_lock) accesses a null-pointer as port was
> > > already marked null by gserial_disconnect.
> > > 
> > > And after gserial_resume acquires the spinlock, gserial_disconnect cant mark
> > > it null until the spinlock is released. We need to check if the port->lock
> > > is valid before accessing it, otherwise it can lead to the above mentioned
> > > scenario
> > 
> > What happens if gserial_disconnect sets gser->port to NULL immediately
> > after your new check occurs, but before
> > spinlock_irq_save(&port->port_lock) gets called?
> > 
> > You may need to add a static spinlock to prevent this from happening.
> > 
> > Alan Stern
> In that case i guess we have to make port_lock a global variable and take it
> out of gs_port structure.
> 
> + static DEFINE_SPINLOCK(port_lock);
> 
> struct gs_port {
> 	struct tty_port port;
> -	spinlock_t port_lock;
> 
> This will require us to change all the spinlock(port->port_lock) used in
> u_serial.c, what do you suggest?

Yes, that would be the correct thing to do.