Re: [PATCH mellanox-tree] net/mlx5: prevent an integer underflow in mlx5_perout_configure()

[Dan Carpenter <dan.carpenter@oracle.com>]

On Mon, Mar 01, 2021 at 12:12:34PM +0200, Eran Ben Elisha wrote:
> 
> 
> On 2/19/2021 11:57 AM, Dan Carpenter wrote:
> > The value of "sec" comes from the user.  Negative values will lead to
> > shift wrapping inside the perout_conf_real_time() function and triggger
> > a UBSan warning.
> > 
> > Add a check and return -EINVAL to prevent that from happening.
> > 
> > Fixes: 432119de33d9 ("net/mlx5: Add cyc2time HW translation mode support")
> > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> > ---
> > Saeed, I think this goes through your git tree and you will send a pull
> > request to the networking?
> > 
> >  From static analysis.  Not tested.
> > 
> >   drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
> > index b0e129d0f6d8..286824ca62b5 100644
> > --- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
> > +++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
> > @@ -516,7 +516,7 @@ static int mlx5_perout_configure(struct ptp_clock_info *ptp,
> >   		nsec = rq->perout.start.nsec;
> >   		sec = rq->perout.start.sec;
> > -		if (rt_mode && sec > U32_MAX)
> 
> This if clause was set to reject perout time start sec bigger than U32_MAX,
> as rt mode specifically doesn't support it.
> 
> A user negative values protection should be generic for all netdev drivers,
> inside the caller ioctl func, and not part of any driver code.
> 

I'm not a networking expert...  :/  It's easier for me to see that this
code will trigger a syzbot splat vs saying that there is no valid use
case for negative seconds any driver.

What you're saying sounds reasonable enough to me, but I don't know
enough about networking to comment one way or the other.  Maybe the
other drivers have a use for negative seconds?

regards,
dan carpenter

> > +		if (rt_mode && (sec < 0 || sec > U32_MAX))
> >   			return -EINVAL;
> >   		time_stamp = rt_mode ? perout_conf_real_time(sec, nsec) :
> >