Re: [patch] USB: usbtest: prevent a divide by zero bug

[Sebastian Andrzej Siewior <sebastian@breakpoint.cc>]

On Sat, Nov 17, 2012 at 09:10:56PM +0300, Dan Carpenter wrote:
> On Sat, Nov 17, 2012 at 06:48:55PM +0100, walter harms wrote:
> > 
> > 
> > Am 17.11.2012 16:06, schrieb Dan Carpenter:
> > > If param->length is zero, then this could lead to a divide by zero bug
> > > later in the function when we do: size %= max;
> > > 
> > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> > > 
> > > diff --git a/drivers/usb/misc/usbtest.c b/drivers/usb/misc/usbtest.c
> > > index f10bd97..7667b12 100644
> > > --- a/drivers/usb/misc/usbtest.c
> > > +++ b/drivers/usb/misc/usbtest.c
> > > @@ -423,6 +423,9 @@ alloc_sglist(int nents, int max, int vary)
> > >  	unsigned		i;
> > >  	unsigned		size = max;
> > >  
> > > +	if (max = 0)
> > > +		return NULL;
> > > +
> > 
> >   maybe you should be more defensive and check from (max <= 0)
> > 
> 
> Nah...  Testing for = 0 is ok.

The parameter comes from user. -1 is hardly possible because the parameter is
defined as unsigned and only alloc_sglist() parameters are signed. Could you
please convert the int to unsigned so it matches the original source of the
parameter?

Passing -1 from user space leads to 

|WARNING: at /home/bigeasy/work/new/TI/linux/mm/page_alloc.c:2403
|__alloc_pages_nodemask+0x24d/0x6d0()

aka ENOMEM so it is not that big of deal.

0 on the hand is more critical.

> regards,
> dan carpenter

Sebastian