RE: [PATCH] mm/usercopy: Use memory range to be accessed for wraparound check

From: David Laight
Date: Wed Nov 14 2018 - 06:09:25 EST


From: William Kucharski
> Sent: 14 November 2018 10:35
>
> > On Nov 13, 2018, at 5:51 PM, Isaac J. Manjarres <isaacm@xxxxxxxxxxxxxx> wrote:
> >
> > diff --git a/mm/usercopy.c b/mm/usercopy.c
> > index 852eb4e..0293645 100644
> > --- a/mm/usercopy.c
> > +++ b/mm/usercopy.c
> > @@ -151,7 +151,7 @@ static inline void check_bogus_address(const unsigned long ptr, unsigned long n,
> > bool to_user)
> > {
> > /* Reject if object wraps past end of memory. */
> > - if (ptr + n < ptr)
> > + if (ptr + (n - 1) < ptr)
> > usercopy_abort("wrapped address", NULL, to_user, 0, ptr + n);
>
> I'm being paranoid, but is it possible this routine could ever be passed "n" set to zero?
>
> If so, it will erroneously abort indicating a wrapped address as (n - 1) wraps to ULONG_MAX.
>
> Easily fixed via:
>
> if ((n != 0) && (ptr + (n - 1) < ptr))

Ugg... you don't want a double test.

I'd guess that a length of zero is likely, but a usercopy that includes
the highest address is going to be invalid because it is a kernel address
(on most archs, and probably illegal on others).
What you really want to do is add 'ptr + len' and check the carry flag.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)