linux-kernel - Re: [PATCH v2] RO/NX protection for loadable kernel modules

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <817ecb6f0906300611s3b21b85by54e689e073bd2012@mail.gmail.com>
Date:	Tue, 30 Jun 2009 09:11:33 -0400
From:	Siarhei Liakh <sliakh.lkml@...il.com>
To:	Arjan van de Ven <arjan@...radead.org>
Cc:	James Morris <jmorris@...ei.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	linux-kernel@...r.kernel.org,
	linux-security-module@...r.kernel.org, Andi Kleen <ak@....de>,
	Rusty Russell <rusty@...tcorp.com.au>,
	Thomas Gleixner <tglx@...utronix.de>,
	"H. Peter Anvin" <hpa@...or.com>, Ingo Molnar <mingo@...e.hu>
Subject: Re: [PATCH v2] RO/NX protection for loadable kernel modules

On Mon, Jun 29, 2009 at 11:30 AM, Arjan van de Ven<arjan@...radead.org> wrote:
> On Mon, 29 Jun 2009 11:16:40 -0400
> Siarhei Liakh <sliakh.lkml@...il.com> wrote:
>
>> This patch is a logical extension of the protection provided by
>> CONFIG_DEBUG_RODATA to LKMs. The protection is provided by splitting
>> module_core and module_init into three logical parts each and setting
>> appropriate page access permissions for each individual section:
>>
>>   1. Code: RO+X
>>   2. RO data: RO+NX
>>   3. RW data: RW+NX
>>
>> In order to achieve proper protection, layout_sections() have been
>> modified to align each of the three parts mentioned above onto page
>> boundary. Next, the corresponding page access permissions are set
>> right before successful exit from load_module(). Further,
>> module_free() have been modified to set module_core or module_init as
>> RW+NX right before calling vfree(). Functionality of this patch is
>> enabled only when CONFIG_DEBUG_RODATA defined at compile time.
>>
>> This is the second revision of the patch: it have been re-written to
>> reduce the number of #ifdefs and to make it architecture-agnostic.
>> Code formatting have been corrected also.
>>
>
> you can still go one step further....
> there is no downside to doing NX at all for modules, except for the 3
> sections now each being page aligned thing. So in principle NX should
> just not be part of any ifdef, only the alignment has any justification
> for being so.
> What you can do in the !CONFIG_OPTION case, is treating the "overlap"
> pages as "most permissive goes"..... if you do that you should have 1
> ifdef in total.
>
> (and one can still argue that making this an option is not even worth
> that, and just always do it unconditional)
>

I can make NX unconditional. However, it will not reduce the number of #ifdefs.
There are two of them in the patch right now: one controls the inclusion of two
extra fields (init_ro_size, core_ro_size) in struct module, and the other one
controls the inclusion of ALL patch code. The *_ro_size fields are used only
for RO, and are not used to set NX. Therefore, this #ifdef will stay even if
NX is unconditional. Since the second #ifdef controls ALL of the patch's code
it will also stay (to control RO part) when NX becomes unconditional.

Given that it will not reduce the number of #ifdefs, do you still think that NX
should be made unconditional?

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