Re: next-20170515: WARNING: CPU: 0 PID: 1 at arch/x86/mm/dump_pagetables.c:236 note_page+0x630/0x7e0

[Kees Cook]

On Fri, May 19, 2017 at 12:18 PM, Andy Lutomirski <luto@xxxxxxxxxx> wrote:
> On Fri, May 19, 2017 at 12:16 PM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>> On Fri, May 19, 2017 at 11:27 AM, Andy Lutomirski <luto@xxxxxxxxxx> wrote:
>>> One thing I've pondered: can we make some debugging mode (kmemleak,
>>> perhaps?) check that freed memory is RW at the time it's freed?  I
>>> once wrote some buggy code that freed an R page and caused an OOPS
>>> much later, and this bug here seems likely to be some code that frees
>>> RWX memory.
>>
>> Which begs for even more checks: nothing should ever make a page RWX.
>> Either R, RW, or RX only... (or X too I guess, in the future).
>
> I could see pages being RWX temporarily during boot.  OTOH if we ban
> RWX outright (after very early boot, anyway), then catching code that
> messes up and leaves pages RWX gets much easier.

Right, early boot is kind of special. It'd be nice to have there, but
I meant during normal runtime. We'd probably need to adjust
set_memory_rw/ro/nx/x around to have the correct side-effects, instead
of just controlling specific bits:

set_memory_rw() (RW_)
set_memory_ro() (R__)
set_memory_rx() (R_X)
set_memory_x() (__X)

That kind of refactoring might be not _too_ bad:

- add set_memory_rx()
- s/\bset_memory_x\b/set_memory_rx/g
- fix what breaks from expecting writable-executable memory
- adjust set_memory_rw() to drop x
- fix what breaks from expecting writable-executable memory
- adjust set_memory_ro() to drop x
- fix what breaks from expecting executable memory
- add set_memory_x() some day...

-Kees

-- 
Kees Cook
Pixel Security