Re: rwx mapping between ex_table and rodata

[Kees Cook]

On Thu, Oct 1, 2015 at 2:12 AM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
> * Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>
>> On Mon, 28 Sep 2015, Kees Cook wrote:
>> > > --- a/arch/x86/mm/init_64.c
>> > > +++ b/arch/x86/mm/init_64.c
>> > > @@ -1132,7 +1132,7 @@ void mark_rodata_ro(void)
>> > >          * has been zapped already via cleanup_highmem().
>> > >          */
>> > >         all_end = roundup((unsigned long)_brk_end, PMD_SIZE);
>> > > -       set_memory_nx(rodata_start, (all_end - rodata_start) >> PAGE_SHIFT);
>> > > +       set_memory_nx(text_end, (all_end - text_end) >> PAGE_SHIFT);
>> > >
>> > >         rodata_test();
>> > >
>> > >
>> >
>> > That should work, yeah. I'd still like to find the default permissions and
>> > make them W+nx, though. Regardless, let's get the above added.
>>
>> The default permissions are set at boot time when setting up the early page
>> tables. When we split them up later on we inherit the PTE bits and then we do
>> that _ro/nx cleanup after the overall layout has been settled.
>>
>> We can't make them W+nx in the early setup without shooting ourself in the foot,
>> because we only set up at the pud/pmd level.
>
> So I think at minimum we should do a (debug) scan in late init, of the whole
> range, for any leftover WX permissions? That would have caught this bug. (and
> might catch other existing bugs that might occur with various configs/hw-layouts.)

I think this would be great. I'd like to disassociate it from PTDUMP,
though, since that exposes kernel address to userspace. It'd be nice
to have the check without also the debugfs entry.

-Kees

-- 
Kees Cook
Chrome OS Security
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/