Re: [kernel-hardening] Re: [PATCH v9 1/4] syscalls: Verify address limit before returning to user-mode

From: Thomas Garnier
Date: Mon May 08 2017 - 15:51:41 EST


On Mon, May 8, 2017 at 8:26 AM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> On Mon, May 8, 2017 at 8:22 AM, Daniel Micay <danielmicay@xxxxxxxxx> wrote:
>> On Mon, 2017-05-08 at 09:52 +0200, Ingo Molnar wrote:
>>>
>>> ... it's just not usable in that form for a regular maintenance flow.
>>>
>>> So what would be more useful is to add a specific Sparse check that
>>> only checks
>>> KERNEL_DS, to add it as a regular (.config driven) build option and
>>> make sure the
>>> kernel build has zero warnings.
>>>
>>> From that point on we can declare that this kind of bug won't occur
>>> anymore, if
>>> the Sparse implementation of the check is correct.
>>>
>>> But there's a (big) problem with that development model: Sparse is not
>>> part of the
>>> kernel tree and adding a feature to it while making the kernel depend
>>> on that
>>> brand new feature is a logistical nightmare. The overhead is quite
>>> similar to
>>> adding new features to a compiler - it happens at a glacial pace and
>>> is only done
>>> for major features really, at considerable expense. I don't think this
>>> is an
>>> adequate model for 'extended syntax checking' of the kernel,
>>> especially when it
>>> comes to correctness that has such obvious security impact.
>>>
>>> Thanks,
>>>
>>> Ingo
>>
>> There's the option of using GCC plugins now that the infrastructure was
>> upstreamed from grsecurity. It can be used as part of the regular build
>> process and as long as the analysis is pretty simple it shouldn't hurt
>> compile time much.
>
> Well, and that the situation may arise due to memory corruption, not
> from poorly-matched set_fs() calls, which static analysis won't help
> solve. We need to catch this bad kernel state because it is a very bad
> state to run in.

Of course, I agree with Kees points on this and previous emails.

A static analysis solution is hard to scale across functions and build
time can suffer. I don't think the coverage will be good enough to
consider this change and static analysis as similar.

>
> -Kees
>
> --
> Kees Cook
> Pixel Security



--
Thomas