On Tue, Mar 24, 2020 at 01:12:36PM +0100, Greg KH wrote:
> On Mon, Mar 23, 2020 at 04:44:31PM -0400, Kyungtae Kim wrote:
> > On Mon, Mar 23, 2020 at 07:46:16AM +0100, Greg KH wrote:
> > > On Sun, Mar 22, 2020 at 11:34:01PM -0400, Kyungtae Kim wrote:
> > > > We report a bug (in linux-5.5.11) found by FuzzUSB (modified version
> > > > of syzkaller)
> > > >
> > > > Seems the variable "npadch" has a very large value (i.e., 333333333)
> > > > as a result of multiple executions of the function "k_ascii" (keyboard.c:888)
> > > > while the variable "base" has 10.
> > > > So their multiplication at line 888 in "k_ascii" will become
> > > > larger than the max of type int, causing such an integer overflow.
> > > >
> > > > I believe this can be solved by checking for overflow ahead of operations
> > > > e.g., using check_mul_overflow().
> > > >
> > > > kernel config: https://kt0755.github.io/etc/config_v5.5.11
> > >
> > > Great, can you send a patch for this?
> > >
> > > thanks,
> > >
> > > greg k-h
> >
> > I'm not sure the following works best.
> >
> > diff --git a/drivers/tty/vt/keyboard.c b/drivers/tty/vt/keyboard.c
> > index 15d33fa0c925..c1ae9d2e6970 100644
> > --- a/drivers/tty/vt/keyboard.c
> > +++ b/drivers/tty/vt/keyboard.c
> > @@ -869,6 +869,7 @@ static void k_meta(struct vc_data *vc, unsigned
> > char value, char up_flag)
> >  static void k_ascii(struct vc_data *vc, unsigned char value, char up_flag)
> >  {
> >         int base;
> > +       int bytes, res;
> >
> >         if (up_flag)
> >                 return;
> > @@ -884,6 +885,8 @@ static void k_ascii(struct vc_data *vc, unsigned
> > char value, char up_flag)
> >
> >         if (npadch == -1)
> >                 npadch = value;
> > +       else if (check_mul_overflow(npadch, base, &bytes) ||
> > check_add_overflow(bytes, value, &res))
> > +               return;
> >         else
> >                 npadch = npadch * base + value;
> >  }
>
> Does this solve the issue for you?  If so, can you fix it up and resend
> it in a format that I can apply it in?
>
> thanks,
>
> greg k-h

I tested the fixed code using a repro (attached), and observed no such
a crash occurred anymore.
Anyway I will send a patch in a separate email.

Thanks,
Kyungtae Kim