linux-kernel - Re: [PATCH v4 1/3] security: Refactor LSM hooks into an array and enum

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] [day] [month] [year] [list]

Message-ID: <CAMp4zn-4Gkk=sopA52CBFBZvLQLweDOhZ9fvnyQYiE0g0Y3tsg@mail.gmail.com>
Date:   Wed, 7 Mar 2018 14:22:14 -0800
From:   Sargun Dhillon <sargun@...gun.me>
To:     Casey Schaufler <casey@...aufler-ca.com>
Cc:     LSM <linux-security-module@...r.kernel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>,
        Kees Cook <keescook@...omium.org>,
        Igor Stoppa <igor.stoppa@...wei.com>
Subject: Re: [PATCH v4 1/3] security: Refactor LSM hooks into an array and enum

On Wed, Mar 7, 2018 at 12:23 PM, Casey Schaufler <casey@...aufler-ca.com> wrote:
> On 3/7/2018 11:18 AM, Sargun Dhillon wrote:
>> On Wed, Mar 7, 2018 at 9:45 AM, Casey Schaufler <casey@...aufler-ca.com> wrote:
>>> On 3/6/2018 11:23 PM, Sargun Dhillon wrote:
>>>> This commit should have no functional change. It changes the security hook
>>>> list heads struct into an array. Additionally, it exposes all of the hooks
>>>> via an enum. This loses memory layout randomization as the enum is not
>>>> randomized.
>>> Please explain why you want to do this. I still dislike it.
>>>
>> Do you dislike it because of the loss of randomization, or some other reason?
>
> I dislike a huge array of untyped function pointers.
> I dislike the loss of type checking in security.c
>
Ok, I can go back to that. My previous approach was using two
list_heads, but people suggested otherwise. Would you be okay with the

>> The reason for not just having a second list_heads is that it's
>> somewhat ugly having to replicate that structure twice -- once for
>> dynamic hooks, and once for 'static' hooks.
>
> There was discussion about this some time ago. In the case
> where you don't allow dynamic hooks, you mark the lists ro_after_init
> whereas in the case with them you don't, but use the locking.
>
>
Why expose the compiled-in hooks to being RW? To me, having two
list_heads seems better, because you have the built-in ones be read
only, and the ones at load time be read/write. In fact, we could
protect the r/w with pmalloc?

>> Instead, we have one enum that LSMs can use and two arrays of heads
>> rather than an entire unrolled set of list_heads.
>
> But how is this better? What is the advantage?
>
>>
>> If we had a way to randomize this, would it make you comfortable?
>>
>