Re: [PATCH] lkdtm: Test KUAP directional user access unlocks on powerpc

[Russell Currey <ruscur@russell.cc>]

On Fri, 2020-01-31 at 07:58 +0100, Christophe Leroy wrote:
> 
> Le 31/01/2020 à 07:53, Russell Currey a écrit :
> > On Fri, 2020-01-31 at 07:44 +0100, Christophe Leroy wrote:
> > > Le 31/01/2020 à 06:31, Russell Currey a écrit :
> > > > +	pr_info("attempting bad read at %px with write
> > > > allowed\n",
> > > > ptr);
> > > > +	tmp = *ptr;
> > > > +	tmp += 0xc0dec0de;
> > > > +	prevent_write_to_user(ptr, sizeof(unsigned long));
> > > 
> > > Does it work ? I would have thought that if the read fails the
> > > process
> > > will die and the following test won't be performed.
> > 
> > Correct, the ACCESS_USERSPACE test does the same thing.  Splitting
> > this
> > into separate R and W tests makes sense, even if it is unlikely
> > that
> > one would be broken without the other.
> > 
> 
> Or once we are using user_access_begin() stuff, we can use 
> unsafe_put_user() and unsafe_get_user() which should return an error 
> instead of killing the caller.

Even better, and thanks for your work on all this stuff.

- Russell

> 
> Christophe